SimForums.com Homepage
Forum Home Forum Home > General Discussion Forums > Hardware, Software, and Computer Technology
  New Posts New Posts RSS Feed - Haswell: 4.8GHz on Air.. Building A Haswell System
  FAQ FAQ  Forum Search   Register Register  Login Login

Topic ClosedHaswell: 4.8GHz on Air.. Building A Haswell System

 Post Reply Post Reply
Author
Message
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 13503
Direct Link To This Post Topic: Haswell: 4.8GHz on Air.. Building A Haswell System
    Posted: June-27-2013 at 10:17am
 
 THIS POST IS NOW COMPLETE....
 
 
Youbetch'a! 
 
 
This thread will cover a Haswell computer system for FSX use and will touch on everything from parts to clocking. Even if you do not intend to clock, this thread will guide you in building a Haswell system.
 
Don't skip through it, READ!
 

BEFORE you crack a processor open, make the INFORMED DECISION to open that chip! You may decide you like what you see on the screen at the clock you are able to achieve without any modification of the CPU since Haswell is a FAR better chip that Sandy Bridge clock-per-clock. A Haswell at ANY clock speed, including default, is going to outrun a Sandy Bridge of equal clock!

@ 4.1 you can expect the same performance as a SB running 4500, no problem!

@ 4.5 you can expect the same performance as a SB running 5GHz, no problem!

@ 4.8 you can expect the same performance as a SB running 5.3+GHz, no problem!

It only goes up from there.
 
I will post BIOS setups for clocking Haswell for basic clocks of 4100-4300 that do not require any special cooling or mod, 43-4400 which can be done on air but carefully...  4500 (which can be done without the mod IF the processor is stable @ 1.20-122v) and 4800 which require modification of the IHS thermal connection in this thread.
 
Of course with liquid, and the right liquid system, Haswell can be clocked above 4.4 with or without any modification of the CPU IHS thermal contact. How far and with what limits will depend on several factors which all comes down to: There is NO set way to define 'exactly' what temperatures will be with this CPU running above 4.1-4.2. Even with the IHS mod one chip may need more voltage that another once you pass the 4.3 mark.
 
 
The big news is, you can run 4800 (or less) on air with a modification and the right parts behind the clock...
 
4800 and 4500 on air and be 100% FPU load test stable requires the cap of the processor be removed, the thermal TIM cleaned from the IHS and replaced with ONE PRODUCT and one product ONLY: Coollaboratory Liquid Pro, not the Ultra! and not anything else!  
 
USA:
 
 
Europe:
 
 
 
Proof:
 
Processor: 4770K @ 1.40 CPU Voltage  4800MHz
Memory: G.Skill DDR3 2400 9-11-11-31
Motherboard: Asus Z87 Sabertooth - any decent Z87 board will do as long as the board has good heatsinks for the voltage regulators
HSF: Thermalright Archon SB-E x2 - with the original fans
Tower: Cooler Master HAF 932 with upgraded fans for ultra-quiet operation, higher CFM and the rear fan perf-metal port cut out for airflow and replaced with thin wire fan grill. BitFenix Spectre Pro 230mm (top/side fans) and two Noiseblocker NB-BlackSilentPro PK-3 140mm (1700RPM front/rear fans)
 
Ambient Temp of room: Room is climate controlled with zone control. Normal room temp 78F @ 45% humidity. For 100% FPU stress test the room zone temp was raised to 81-82F and verified before starting the test and during to confirm linear results at the higher ambient temperatures.
 
Load: 100% FPU STRESS TEST AIDA64 Extreme Edition v3.00.25 or above which comes with HASWELL instruction load tests for well over 1 hour followed by OCCT: Linpack 64bit AVX enabled, 1hr.
 
CPUz: v1.64.2 or above
 
Intel Hyper Threading: MUST BE DISABLED for these clock speeds under AIR and while testing full FPU Stress however it CAN be re-enabled and run with normal load tests as well as normal operation. FPU Stress of this nature will never occur with normal load tests or with typical normal software use. HT will have NO EFFECT on FSX performance so running with it ON is a waste of temps for a FSX tower!
 
IMPORTANT: 100% FPU Stress is the highest load you can place on a processor. We normally do not run these load tests when clocking to check stability but in this case its important to prove the fix works and the system is stable. This type of testing has been known to force even unclocked systems to show instability so this test without a doubt proves that running Haswell on air at 4.8GHz is possible.
 
 
NOTE: 100% FPU Stress can only be performed by disabling all other stress tests except FPU in AIDA64 as shown
 
 
 
When running Haswell instruction FPU tests, 88-90c is the MAX acceptable TOP END. Since we wont ever see that temp in any other load test the top end is allowable AND we do not shut down the Intel thermal features of the BIOS which will insure no damage to the CPU under these conditions.
 
A Haswell clock is extremely simple as compared to clocking in the past! There are very few changes required as Asus got this right with their BIOS and motherboards.
 
OTHER FULL LOAD TESTS: Alternately OCCT v4.40 or above can be used by setting the test to CPU: Linpack and enabling 64bit and enable 'AVX' however that test, although good to use, will not load the processor quite as high as AIDA64.
 
On the other hand, OCCT WILL use far more memory and in that the combination of the two are a good pair to verify the IHS modification is working and is 100% stable.
 
The memory controller and the memory can become unstable when a CPU is pushed this hard, therefore the OCCT: Linpack 64bit AVX test is run to back up the results from AIDA64.
 
 

==================================================================


OK, so this has been successfully accomplished and it took about 45min to 1hr to complete.

A 4770K @ 4.5GHz is equal to a Sandy Bridge @ 5.0 or slightly above. A Haswell 4770K @ 4.8GHz is far better than a Sandy Bridge @ 5.3+GHz

Intel found it in their infinite wisdom to not use a solder based IHS contact between the processor itself and the inside of the cap it covers. Instead they use a CHEAP and DRY thermal compound pad which insulates the CPU from the cap, and from there its all downhill.


If you have fairly steady hands and can follow directions, YOU CAN FIX THIS. I will show you how, but be warned.. what I am about to show you means you BOUGHT the processor, there is NO WARRANTY anymore and if you screw-up by damaging the SMD resistors under the cap, or bend pins in your motherboard CPU socket removing and reinstalling the CPU: you ATE that processor or motherboard.

Now, I HAD TO SAY THAT. IF you do this, its ON YOU, not me, not Flight1 and not anyone else.

That being said, this operation is really very simple.. you just need to know where those SMD resistors are, what tools to use and how to get that cap off, clean underneath, prep and replace that TIM the right way and with the right compound.

This is not hard but it does take a little control with a hobby or utility knife. If you are 'fairly' good with a hobby excato knife and you are not IMPATIENT, this is easy to do.

 

NOTE: You will find a lot of guides on the net as well as videos about this subject..  be very cautious as I have seen some of those videos and guides can lead to disaster or failure the way they show how this is done. Some are good, many are bad and posted by amateurs who come very close to eating a processor and some of them are down-right IGNORANT in how they leave the silicone on the cap/wafer and most important the thermal compound they use to replace the TIM and then cry this doesn't work or does not work very well. BE CAUTIOUS of the internet tech expert with a video camera and a desire to display the size of his internet 'pee-pee'.
 

I will post the entire process here as I get time. Check back every so many days for a update. I will post the word: COMPLETE GUIDE when the full guide is ready here.
 
This guide will be for everyone as it will leave no stone unturned and show every step with all the 'gotcha's'.
 

DO NOTE: Ivy Bridge suffers from this same issue. This can be done with Ivy Bridge (certain models exempt) as well. IB will outperform SB as well but will not quite outperform Haswell at the same CPU speed.

If this is done right, and, your CPU cooler and tower are correctly specified for clocking, you can expect a 15-20c drop in UPPER END temps. I was able to accomplish a drop of 22-25c respectfully.

This mod does not mean you have to run 4800! Haswell @ 4.5 and 1.25-1.26 CPU Voltage is a SNAP using this mod and you wont crack 70c on a normal load test, heck you probably wont see higher than 58-65c! LOL
 
 
If you run high-end liquid and know what you are doing in a BIOS and with a clock,.. this mod opens the door to 5Ghz+ without extreme cooling such as LN2 or phase change.
 

More later.....

 
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 13503
Direct Link To This Post Posted: June-28-2013 at 1:40pm
 
 
THIS POST IS NOW COMPLETE....
 
 
 
 
THE ISSUES AND THE PARTS FOR FSX: Do you need this mod?
 
 
Regardless of making a mod to the CPU IHS or not, clocking or not, everyone should start here and read this post...
 
 
To start this off I would be remiss and a complete nincompoop of a engineer if I did not preface this entire subject with a very detailed explanation of the issue and the possibilities around it!

...and then add in this question:


DO YOU NEED THE IHS MOD AND WHY?

Any self-crowned technical wizard on the internet who posts:

"Do this mod, you wont regret it.. here is how its done but you do it at your own risk"

..and then does not preface that statement with what I am about to outline in this entire post is NOT anyone you should be taking technical advice from!

 
We are not talking about swapping CPU coolers or changing the thermal paste brand here we are entering a area that modifies a CPU whereby the warranty is gone and there is no recovery of cost. Although a simplistic operation to accomplish with a little patience with a blade this may or may not be right for you to do, and, you need to know the facts before you make that decision.
 

Some users like myself who are experienced and not short on the wallet will do this mod regardless of anything else ONCE WE CONFIRM that there is a valid suspicion that there is a problem under that cap, NOT BEFORE and even then we take a chance as to what we will find under that cap.

But before you personally commit to something like this there are few items you should consider first, and so you can make an INFORMED DECISION I will open this topic with the following information...


First, and foremost, although I have confirmed the first production run from one facility from industry sources that the Haswell 4770k does in fact include a very poorly designed IHS thermal connection between the top of the processor and the inside of the cap, that DOES NOT MEAN Intel did not split the design change and produce some with solder nor does it mean Intel wont switch up later and correct this on later steps of the same processor!

Just because I might have connections into Intel or in the industry does not mean that I get all the information. There are sometimes limits to information that my contacts have access to as they are not privileged in all facets of the business dealings nor are they involved with every aspect of production or decisions around production. I get good information but I also know that information comes with limitations which include non-disclosure issues.

Furthermore, although many of the Ivy Bridge designs are the same as Haswell, there have been confirmed differences in models, but again, there is no way to confirm all models were produced the same way therefore it is not possible to know if there is TIM or fluxless solder under all those caps. Intel appears to be experimenting and started that with IB.

My take on this is Intel pulled this stunt to cut costs since PC sales are down and they are watching the results. That means they could switch-up in production runs and completely return to fluxless solder, OR, decide that their flagship or hex core products have the better IHS and leave the standard models with TIM. There is no way to predict or know what they will do.

The problem with using temps alone to define if a issue may be present with a 6-8 core processor is a hex core will always run hotter than a quad, simple physics, so trying to guess if that chip has a TIM or solder can be a crap-shoot.


THEREFORE before you go committing to doing anything with a processor cap and removing it for inspection you should first consider the fact that there is a possibility you could open that cap and find out that solder was used instead of TIM, and second, you should ask yourself: Do I need this modification and if so, why.


To help you make that decision I am going to list facts here.. some people just do not clock or if they do they do not look for top-end, others like cooler temps regardless of the CPU speed or if they clock or not. Sill others want to clock but do not have the nerve to open a CPU cap be it a financial loss fear or lack of confidence in their ability to do it right. Because of all these differences you should know the facts before making the decision to crack that cap..

 
 
 
 

FACT 1:
 
DON'T BE A DOPE.... TEST THE PARTS FIRST!

Unless you have far more 'cojones' than brains, NEVER crack a CPU cap (or make any other mods to hardware) before that CPU, motherboard, video card and memory have been fully tested for a period of time running unclocked.

We don't modify anything until we have fully installed a system all the way up through Windows, tuned, trimmed and verified with stability and defect tests.

To do so is outrageously ignorant as we have no idea when those parts arrive if the motherboard, CPU, video card or memory or any other components have defects.

Once the hardware is confirmed stable and free of defect, then we move to the next level.

 
 
 

FACT 2:
 
NOT EVERYONE WILL CLOCK OR WANT TO MOD THE PROCESSOR.. WHAT DO I NEED TO ACCOMPLISH MY GOALS:
 
 
Since we have no idea at this point if we need this modification, in order to make that call you need to first ask yourself:

Do I want to clock and if I do how high do I want to clock, and, is it worth it to me personally to run a high-end clock. If I do want to clock, what should I order for parts to clock Haswell?

 

Of course there are different levels of this. Lets look at each and what is required:

 
 
 
* I do not want to clock, or, I do not want to clock but I prefer a better cooling system than the Intel provided heatsink.

- Do not mod with the CPU or IHS   DONE!  If you prefer a better cooling solution then a lower grade liquid such as a Corsair H60 or a medium duty larger air heatsink will do the job.

You can also work with the cheaper budget boards if they meet your needs for features over buying the more expensive and better made products that target high overclock performance
 

--------------------
 
 
* I want to clock but I don't really know how and I am not technically inclined enough to learn or I just do not want to learn

 - Do not mod with the CPU or IHS. With the Asus motherboard you can after confirming the parts as defined in FACT 1, simply enter the ADVANCED BIOS settings under the AI Tweaker menu click the setting "OC TUNER and set it to "RATIO ONLY".

 
I will discuss the details about how to do that later but here is the gist... The BIOS will automatically reboot, test, and set itself to run a automated CPU multiplier and CPU voltage change as it loads the system. Asus got this system right and depending on the test on reboot you will be running right around 4.1GHz automatically.

Do be aware that for best TEMP results a decent heatsink replacement would be best such as the Corsair H60 or a medium duty larger air heatsink will do the job but this automated clock can be done with the stock Intel heatsink you will simply run warmer on the Intel HSF but not dangerous. I would still suggest a medium duty replacement just to be sure and if you prefer cooler results. 

 The budget boards will most likely support this kind of low-end clocking.. you can check that at the Asus website or download the motherboard manual and review it first to confirm.

 
 
-----------------------
 
* I want to clock manually in the BIOS with locked voltages for max stability to about 4.2-4.4GHz, perhaps a bit higher, and, I do not want to mod the CPU.


First, it is important to understand that at this point WE HAVE NO IDEA what quality of IHS thermal conductivity is in play, and, we have no clue as it if we have a processor that will run stable at LOW CPU voltage.

The SAFE BET is to always assume you will NOT GET a "golden chip" and you will receive a lousy one. 'Golden chip' processors are the processors that will run high CPU speeds on very low CPU voltage. They are few and far between! If you work from the assumption you will be in the worse situation for clocking when ordering the processor YOU WILL NEVER LOSE. It will only get better if you do happen to get to hitch a ride on the golden goose.

So under this scenario defined by the statement above;

- Do not mod with the CPU or IHS. In this case you will need a 'super-cooler' class air cooler such as the Thermalright Archon SB-E x2 or the Thermalright Silver Arrow SB-E Extreme  (or equivalent) OR you will need a liquid system that can handle the heat. With a air cooler you WILL most likely be limited to 4.3-4.5 and that assumes the chip will allow it!

BE VERY AWARE that some air coolers can cover the memory slots (The Thermalright Archon does not) and 'high-back' memory heatsinks wont fit under the fan! Some of those liquid systems on the market are no better at cooling than the large heatsinks I defined and they are far louder.

 
OPTIONAL: You may want to consider upgrading the fans that come with the heatsink for higher volume. For 140mm fans I can highly suggest the BitFenix 'pressure rated' fans  http://www.frozencpu.com/products/19769/fan-1248/BitFenix_Spectre_Pro_All_Black_140mm_PWM_Fan_BFF-SPRO-P14025KK-RP.html?tl=g36c365s1507  
 
Those BitFenix fans are case fans and they can be loud because of that but if you wish to use them to replace the fans on the heatsink and the sleeves between the holes are not removable, then your choice is to modify and cut those sleeves out which is really not necessary.. or far more simplistically:    use zip ties. Simply slip 2 - 12" long zip ties through the holes at opposing corners (diagonal) and pass then through the fins of the heatsink and back around, zip closed (don't over torque) and cut off the excess

I would highly suggest for liquid and for any clocking above 4.3 without any mod should be a Corsair H100 system (latest revision) and be aware that your limit may very well be 4.5 with that cooler if the IHS contact is extremely poor. Most likely you will be able to achieve 4.6 without issues on the H100, possibly a bit more depending on the chip. On the other hand you MIGHT be able to run 4.8. Its a crap-shoot!

 

A better motherboard is suggested for any clock over 4.2Ghz. The budget boards are not made for higher clocking.
 
 
----------------------
 
 
* I'm all in for 4.5 to 4.6GHz but I do not want to use liquid and I don't want to mod a processor..

- This is a crap-shoot you will probably lose.

You will need to first clock your system up slowly and in increments to define if your IHS thermal connection is of poor quality or, assume that will be the case when purchasing parts but in this case you must provide yourself with the temp data in load testing. If you require 1.24 for 4.5 and 1.27 for 4.6 and your temps are very high using OCCT:CPU test alone, then you MUST take a lower CPU speed. 

Keep this in mind, the standard OCCT: CPU test WILL NOT LOAD that processor to the max. If you are seeing 80-85c max temps under that test then you ARE NOT loading that CPU with maximum instruction and you will hit 100c+ under the right conditions! You can be stable in OCCT but unstable in CPU load spikes.. that is a chance you will take without the IHS mod at those temperatures.

IF you can run OCCT: CPU test and not exceed 72-75c then your IHS is far better than many, and, that could be because you have a golden chip that does not require 1.27v for 4600. MOST processor will NOT run 4.6 at lower voltages and require 1.3 or above. If  you can pass that test at that temp then you do not need to mod the IHS and can run 45-4600 safely on air however I seriously doubt ANYONE will get to 4600 on air without the IHS mod even with a golden chip.

 

Purchase a good mid-range price motherboard, 180-250 USD for clocking above 4.3
 
 
----------------
 
 
* I'm all in for 4.8-5GHz and higher if possible and I don't mind liquid or the cost but I don't want to mod the processor.

-Although you might get a golden chip and the Corsair H100 can deal with it, the ODDS ARE YOU WONT. Therefore if you are thinking of anything above 4.8 you need to plan to either BUILD a custom liquid system, (you better know what you are doing), OR, invest in a very powerful system that is far better rated that the latest Corsair H100 or H110! You will have to load test and I suggest you use the AIDA64 FPU Stress test alone to define your limits as you increase CPU speed and CPU voltage.

You can try it with a H100 or H110 liquid system but do not be disappointed if it does not happen. It might!   or it might not!

 

 
 
------------
 
* I'm all in for 4.6-4.8GHz and higher if possible and I don't want to use liquid but I don't want to mod the processor. 

- I'm all in for a harem and palace in Saudi Arabia... Wacko Keep dreaming...    NEXT!

 
Ok, all jokes aside..  With the current production Haswell this is like walking into a casino and putting a 30 year earned IRA down on one roll. You might walk out rich but most likely you just bought the farm for financial ruin.

 

-----------------------------

 
* I'm all in for 4.6-4.8GHz and higher if possible, I don't want to use liquid and I am willing to try the mod, or, I will mod the processor understanding the warnings you posted.

- In this case you will need a 'super-cooler' class air cooler such as the Thermalright Archon SB-E x2 or the Thermalright Silver Arrow SB-E Extreme (or equivalent). With a air cooler you WILL most likely be limited to 4.7-4.8 and that assumes the chip will allow it!

BE VERY AWARE that some air coolers can cover the memory slots (The Thermalright Archon does not) and 'high-back' memory heatsinks wont fit under the fan!

Choose your air cooler wisely and don't think for one minute a Cooler Master or a Zalman will cool that proc like a 'super-cooler' class heatsink will!

 

-----------------------------------
 
* I have the cooling system parts all under control (I know what I am doing).. It looks like I may be able to clock higher from the temps I am seeing under the load tests and I just want to clock as high as I can.. My 'cojones' and wallet are ready for anything.

-Ok.. its your nickel,.. so then lets rock with a mod.

 

 
 
 
FACT 3:
 
Unless you are simply 'all in at any cost' The final decision should rest with load testing and performance results.

I have outline different scenarios above, but one thing is for certain, in all circumstances regardless of clocking, mods or not, you need to LOAD AND STABILITY TEST your system before making the decision. You may find out you just happened to get a golden chip, or, you may have received a processor that for what-ever-reason the IHS thermal connection is better than most.

BEFORE you crack a processor, make the INFORMED DECISION to open that chip! You may decide you like what you see on the screen at the clock you are able to achieve since Haswell is a FAR better chip that Sandy Bridge clock-per-clock. A Haswell at ANY clock speed, including default, is going to outrun a Sandy Bridge of equal clock!

@ 4.1 you can expect the same performance as a SB running 4500, no problem!

@ 4.5 you can expect the same performance as a SB running 5GHz, no problem!

@ 4.8 you can expect the same performance as a SB running 5.3+GHz, no problem!

It only goes up from there.
 
So make your decision before ordering parts and cover yourself in case you change your mind.. at the same time there is no such thing as too cool unless we are talking liquid nitrogen or extreme phase change.
 
 
 

FACT 4:

HASWELL LOVES MEMORY SPEED AND LOW TIMING

REGARDLESS OF the CPU speed, you will want, at minimum, DDR3 2400 9-11-11-31 memory and if the sticks are decent quality they will run a 1T command rate with no additional voltages required (unless we are talking 5GHz+ then some voltage tuning may be required)

BUDGET: You can use 2133 9-11-10-28 memory which is typically cheaper but I would highly suggest the investment in the 2400 C9 memory product.
 
 
DDR3 2800: This is completely in range and not a problem assuming you are not running COLD on CPU settings. Same conditions as outlined for DDR3 2400. Your timing wants to be CAS 11 or 12 MAX, lower if possible. Do not spend the money on DDR3 2800 memory that runs a CAS timing above C12
 

ABOVE 2800 - You are now getting into a grey area. YES, unlike platforms of the past, Haswell can run these speeds but stability will be based on the stability of the CPU and the temps and may very well require additional knowledge of DRAM settings in the BIOS for both sub-timing and voltage control. The only way to know is test it all in combat at the CPU speed you wish to run. The stick may run their default speed (in example DDR3 3000 12-14-14-38 for a fat wallet of 750 dollars LOL ) and not require any additional settings, OR, you may have to boost the DRAM Voltage, or DRAM POWER PHASE and DRAM CURRENT.

Once you exceed DDR3 2800 and are pushing clocks of 4800 and higher you are into a territory that can not be defined on a calculator prior to attempting such setups. The better boards should run those extended speeds with some adjustments.

 

 
FACT 5:

THE BIOS SETUP AND NEW DANGERS WITH HASWELL

AT this point you may have it all under control as a experienced clocker and will figure the BIOS out yourself as well as a clock setup, but most people don't and order to know where you stand with a decision to mod the CPU you will need to know where the temps are in the clock you wish to run. To do that I will outline the basics to primary clock setups with this WARNING:

BE AWARE: Asus has changed how the CPU Voltage is used under the setting: "ADAPTIVE MODE" if you use that and you hit Haswell with full-on Haswell FPU instructions the BIOS is automatically programmed to add an additional voltage to what you defined and calculated as MAXIMUM.

This can be a disaster if you do not know exactly what you are doing so it is highly suggested that ANY clocking other than the simple automated 4.1GHz clocks established automatically by the BIOS that we ONLY USE: CPU VOLTAGE: MANUAL MODE and lock that voltage regardless.

 

 
 
FACT 6: 
 
 
POWER SUPPLY
 
Be aware the PSU must be Haswell certified. Using an older PSU may not be a wise choice unless you know from the manufacture that PSU is rated for Haswell use. I personally went with the Corsair AX860i and have not regretted the purchase. There are other models out there that are not as expensive however for a single GPU system I would make sure its at LEAST 750watt, 850 is fine..  above that is a waste of money for single GPU towers.
 
Be advised we will be clocking with C States enabled in the BIOS, therefore that PSU must be Haswell certified and if you want the advantages that C States provide even with a higher clock don't skip verifying the PSU!
 
 
 
 
 
 
FACT 7:
 
VIDEO CARD
 
SLi or any dual card setup is a waste of money for FSX.. if you use it for other games that is one thing, if you only focus on FSX don't waste your money on a SLi setup.
 
Also, never buy a 'dual-core' video card for FSX as those cards can run FSX worse! Single GPU core cards only!
 
 
This one is a 'no-brainer', because of the clock per cloc ability of Haswell in compare to processors of the past, AND, the advanced motherboard design, any system running 4.1GHz or above for FSX should be the GTX 780, hands down.
 
For those not clocking or who do not intend to clock higher than the automated system allows, the GTX 680 or even a GTX 580 (which is in reality more powerful than the 680) is your lowest card choice but you can go with the GTX 770/780 as well. The GTX 780 is everyone's friend with Haswell and FSX. Unlike the past where the more powerful GPU could bottleneck the unclocked CPU, the GTX 780 and Haswell clocked or unclocked are a good pair and will show improved results over all other video cards.
 
People who do not clock at all and who are budget limited can go with a 570/670 series with reduced results (try to stay with the GTX 770 instead!) however anyone who shoves a 560/660/760GTX or ANY TI class video card or lower in a 4770K Haswell system based on some 'guru' suggestion at a flight sim site, just got duped by a dope.
 
The 500 series video cards are NOT PCIe 3.0 so be very aware that if you decide to use a older card with Haswell you will see limited results.
 
I understand economics and I understand that sometimes sacrifices must be made, but if it is your goal to buy a Haswell system for FSX, don't neuter it with a poor quality video card and if you do, realize the limitations you will deal with. Do not listen to people who tell you that video card is not important or can be a LOWER end card..   they are ignorant and you will suffer the consequences for that purchase!
 
 
The Titan card should be reserved for the deep-pocket user .. the standard GTX 780 can be clocked to Titan speeds/specs and NEARLY perform the same... Super-clocked 780s can be purchased that are FINE for the job, certified by the manufacture and require no dangerous user clock changes. That is what I purchased, the SC model. It effectively runs Titan clock speeds.
 
The only thing a GTX780 CANT DO over the Titan is give you the same high video memory frame buffer support and even then, that is a lot of money for a much larger frame buffer.. 
 
My take: there are a lot of Titan users out there who got the 'shaft' LOL when the 780 was released. Glad I held off on that purchase.
 
The GTX 780 comes with 3GB of video memory. That should handle multi-monitor setups but if you are deep-pocketed and intend to run a huge array of monitors you can bump up to the Titan... I can't say that will make a difference for you or not,.... it is your choice and your wallet!
 
 
 
ATi video cards..     I won't go there. Nvidia is the GO-TO card for FSX.
 
I am not a fan-boy, facts are facts.
 
 
 
 
FACT 8:
 
THE TOWER SELECTED CAN MAKE A HUGE DIFFERENCE
 
Getting air in and out of the tower can make all the difference to a stable clock. If you do not intend to clock, or, you intend to use the low-end automated clocking setup you will not be effected as much by heat but you should still always research a tower for good airflow.
 
Better liquid cooling systems can have external radiator and fan setups that get the heat of the CPU out of the tower. Some liquid systems are self contained inside the tower. None the less, your voltage regulators and the video card as well as mechanical drives and other components are going to radiate heat. The combination with respect to the ambient temperature and humidify of the room will add to the final result.
 
Seasonal changes force people to change their clock speeds unless their systems run in climate controlled environments that maintain a consistent ambient year round.
 
The faster and more efficient the airflow in and out of the tower, regardless of all else, the better. Of course we don't want our towers buried under a desk or sitting next to a heating vent as that will effectively reduce the ability of the system to remain cool.
 
One other factor is tower design with respect to the fan ports. This is exceptionally critical with the rear tower fan that on most good tower designs sits just above the voltage regulators and directly in line with the CPU heatsink.
 
There are towers that have a open rear fan port with a wire grill cover.. these are very high efficiency and not easy to find in tower designs today. Perforated metal stamped REAR fan ports as shown in this image:
 
 
 
LOWER the efficiency of that rear fan by as much as 50% or GREATER. A simple modification can fix that!
 
a. Mount the fan on the outside of the tower with 2 screws as a temporary marking guide
 
b. Using a fine line sharpie or other such marker follow the inside edge of the fan and mark the tower hole. Remove the fan from the back of the tower.
 
c. Using a cutting tool cut that hole out following the outside edge of your marker line. Take it slow and easy and work it all the way around then remove that perforated metal. You can use a Dremel with a cutting wheel or snips although snips will probably leave nasty edges.. just be careful and work that metal out of that hole. You can file the edge when done to take the sharp edges off.
 
d. Purchase a fan grill like this to cover the hole: http://www.frozencpu.com/products/5319/fgc-08/140mm_Fan_Grill_-_Black.html
 
In this example I am using a 140mm fan grill since the rear fan of this example tower is 140mm
 
You end up with a full-flow rear fan port as shown:
 
 
 
 
Larger 230mm fans that typically sit on the SIDE and the TOP of towers are OK with the perforated metal grills since they are far larger and run at lower RPM, however that REAR port is not! and can be modified to allow far greater airflow, and less noise too!
 
 
For clocking and high efficiency:
 
Look for a tower that has front, side, top and rear fans. Do note that most towers come with cheap fans that although work, they are usually not high quality volume pushing design. You may find the fans that come with the tower will do the job you require, however if you are attempting to run high clocks on AIR, you may want to consider modding the rear fan port and upgrading all the tower fans.
 
 
I personally use the HAF 932 http://www.newegg.com/Product/Product.aspx?Item=N82E16811119160 with a modified rear fan port and upgraded fans. I replaced the Cooler Master fans with my own spec: 2 BitFenix Spectre Pro 230mm (top/side fans) and two Noiseblocker NB-BlackSilentPro PK-3 140mm (1700RPM front and rear fans) that is pressure rated.
 
I got rid of the silly LED lights in the fans as I have no use for or wish to look at that bling. LOL
 
That change along with the fan grill mod dropped the tower temps greatly as well as lowered tower air noise when the fans ran under a load.
 
FRONT - SIDE FANS = INTAKE
REAR - TOP FANS = EXHAUST
 
Balance the flow! IN and OUT. Do not intake or exhaust all fans.
 
 
 
DUST FILTERS: I do not use them. My environment has a central air system that includes electronic dust removal. Even with that I must clean my tower once a year which includes removing the video card, taking the cover off the card and cleaning the heatsink and fan. I know some people have issues with dust but be very aware that dust filters can significantly reduce airflow in a clocking tower and limit the clock. If you can do without them and simply perform maintenance a little more often the tower will breathe much easier. You must make that call based on the tower environment and how often cleaning is required.
 
 
 
 
You do not have to go to the extreme of modding a tower, but do be aware you CAN get far better efficiency out of many towers that may be perfect for your needs but simply need some help with the fan and port design.
 
 
 
Your tower should not be selected based on BLING and LIGHTS. It should be selected based on your clocking needs, the hardware you need supported such as drives and its ability to get the air IN and OUT of the unit as efficiently as possible. This can make the difference between a 4.6 and 4.8GHz clock that does not require seasonal clock reductions.
 
There is no such thing as a high performance air HSF clocking tower that is too big with too many fans. That larger tower is a blessing when clocking. You can always control fan speeds. The Cooler Master HAF 932 series is a good base for a reasonable price. There are other choices, be smart about what you choose and take a careful look at their designs.
 
 
 
 
 
 
 AND LAST:

This is all based on ASUS MOTHERBOARD DESIGN. I can not define setups, settings or make any assessments with respect to other motherboards and how they run Haswell.
 
It would not be in your best interest to purchase any 'budget' Asus board if you intend to clock Haswell above 4.2. You DO NOT NEED the 350+ dollar monster of a motherboard, although according to Asus they say their ROG boards are tested for 4800MHz certified (marketing?  who knows!),.. but at the same time the Asus or other Haswell budget boards that targeted at the budget market, may not have the voltage regulation and the filtering to handle clocks and especially higher memory speeds. Any board that sells for less than 180 bucks may be an issue above 4.5.. I can not define which board of that class will work @ 4.5 and above and which one will not.
 
You get what you pay for BUT ONLY IF if you know how to use it.
 
 
DONT BUY A ROG CLASS MOTHERBOARD THINKING IT WILL NET YOU A BETTER FSX, IT WONT AND YOU WONT GET ANY MORE OUT OF SUCH A BOARD UNLESS YOU ARE A HIGHLY EXPERIENCED OVERCLOCKER WHO KNOWS WHAT THEY ARE DOING.
 
 
 
 

The next section I post will cover STORAGE SYSTEMS and how to select and get the most out of them.....
 
 
 
Please be patient as I am spread fairly thin right now with time.  Geek
 
 
 
 
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 13503
Direct Link To This Post Posted: June-29-2013 at 2:21pm
 
 
 
THIS POST IS NOW COMPLETE....
 
 
 
 
 
ABOUT STORAGE SYSTEMS
-AND-
HOW TO GET THE MOST OUT OF STORAGE FOR HIGH PERFORMANCE
 
 
 
 
If you do not understand anything about hard drives or hard drive technology all you need to know is what is in these key statements which target every possible scenario in setting up a system for efficient and effective file delivery.
 
 
Follow this and you will never go wrong when planning and building a new system
 
 
 

 
MECHANICAL HARD DRIVES:
 
- IF YOU MUST PARTITION ALWAYS USE THE FIRST PHYSICAL PARTITION FOR HIGH PERFORMANCE APPLICATION ACCESS, 2ND PARTITION FOR DEAD STORAGE
 
- ALL DRIVES WITH PERFORMANCE APPLICATIONS INSTALLED MUST BE 'CORRECTLY' DEFRAGEMENTED FOR MAXIMUM FILE ACCESS PERFORMANCE
 
- DO NOT PARTITION DEDICATED HIGH ACCESS APPLICATION DRIVES
 
- THE FASTER THE DISK ROTATION SPEED, THE BETTER
 
- THERE IS NO SUCH THING AS TOO MUCH FREE SPACE
 
- NEVER FILL A MECHANICAL DRIVE MORE THAN 65%
 
- THE LARGER THE DISK PLATTER SIZE, THE FASTER
 
- FREE SPACE IS YOUR BEST FRIEND
 

 
 
SATA CLASS SOLID STATE HARD DRIVES:
 
- SSD MUST BE RUN IN AHCI MODE AND PROPER TRIM DRIVERS INSTALLED
 
- SSD MUST BE ALIGNED PROPERLY OR SUFFER PERFORMANCE LOSS
 
- SSD MAY REDUCE LOAD TIME BUT WILL NOT ADD FRAMES TO FSX
 
- SSD REMOVES MECHANICAL LATENCY FROM FILE CALLS
 
- SSD IS NOT 'ALWAYS' BETTER THAN MECHANICAL (it depends on the drives you are comparing)
 
- NEVER FILL A SSD MORE THAN 85%
 
- NEVER DEFRAGMENT A SSD DRIVE  
 
- DO NOT PARTITION SSD
 
 
Although SSD drives significantly reduce maintenance and do not require defragmenting you must still understand that all SSDs are not created equal. Some are no better and can run slower than some high-end mechanical drives. Research your SSD purchase carefully and do remember, you will usually get what you pay for with SSD..
 
 

 
PCIe BASED STORAGE CONTROLLERS (SSD OR HDD)
 
- MODERN PCIE CONTROLLERS WITH ON-CARD DDR MEMORY AND PROCESSOR PROVIDE THE HIGHEST MECHANICAL DRIVE PERFORMANCE (this is not a 150-200 dollar card)
 
- DEDICATED SSD 'ON-CARD' PCIE DRIVES ARE THE HIGHEST PERFORMANCE POSSIBLE
 
- WILL ABSOLUTELY PROVIDE THE HIGHEST PERFORMANCE WITH THE LEAST CPU HIT
 
- MUST BE CAREFULLY INSTALLED SO PCIE X16 IS NOT NEUTERED TO X8
 
- EXPENSE WILL NEVER BE RECOVERED IN PERFORMANCE VALUE IN FSX
 
- EXTREMELY EXPENSIVE AS COMPARED TO TYPICAL SATA
 
- MOTHERBOARD MUST SUPPORT THE TECHNOLOGY
 
 

 
HARD DRIVE TIPS FOR HIGH PERFORMANCE (ALL DRIVE TYPES):
 
- WHEN SSD SPACE IS LIMITED DUE TO ADDON SPACE CONSUMPTION, USE THE LARGEST MECHANICAL DISK PLATTER DRIVE FOR A DEDICATED FSX INSTALL
 
- DO NOT INSTALL LARGE DATABASE ADDONS (REX-GEX-FEX-AS2012, ETC) TO THE FSX DRIVE OR WINDOWS, POINT THEM TO A 'DEAD STORAGE' DRIVE **
 
- WHEN EVER POSSIBLE, USE SATA CLASS SSD -OR- PCIE STORAGE CONTROLLERS FOR THE BEST DISK PERFORMANCE WITH FSX
 
- RUN FSX AND PRIMARY ADDONS THAT REQUIRE INSTALLATION DIRECTLY INTO FSX ON ITS OWN DEDICATED HARD DRIVE
 
- INSTALL LARGE PHOTOSCENERY DATABASES TO THEIR OWN DEDICATED HARD DRIVE(S) AND POINT FSX TO IT
 
- MULTIPLE DISK SYSTEMS ARE FAR BETTER AND ALLOW MORE OPTIONS THAN SINGLE
 
- DON'T FALL FOR THE "HYBRID MECHANICAL DRIVE" OR 'CACHE' MARKETING GAME 
 
- YOUR SATA CABLES MUST BE RATED FOR THE PORT AND DRIVE SPEED
 
- PSU MUST BE RATED WITH A LARGE NUMBER OF DISKS
 
- THERE IS NO SUCH THING AS 'TOO MANY' DISKS
 
- RAID IS NONSENSE AND A WASTE
 
 
 
** SSD cost per gig is expensive and the storage space can be limited. With mechanical drives defragmenting services can take time and the more data on them the longer that maintenance can take. FEX-REX-GEX-AS2012 -- These are all database applications that SEND files to FSX. They do not have to be installed into the Windows partition or FSX! They have very large numbers of texture files which can take up gigs of drive space. When installing the applications into Windows there is no reason to install them to a Windows drive or a FSX drive. Point the installers to a separate storage drive. The applications will work exactly the same when you launch and use them and simply send files to FSX except you are not eating drive space on performance disks to store them!
 
 
 
 
IF YOU HAVE ENOUGH DISK SPACE, PLACE FSX AND ALL THE ADDONS THAT MUST INSTALL 'DIRECTLY INTO' FSX ON ONE HIGH END SATA OR PCIe SSD OR MULTIPLE SSD'S BUT KEEP IT SEPARATE FROM WINDOWS. RESERVE MECHANICAL DRIVES FOR LAST OR WHERE DISK SPACE BECOMES AN ISSUE WITH SSD.
 
 
 
and last
 
WINDOWS SOFTWARE DISK BENCHMARKS AND THE MARKETING YOU SEE FOR DRIVES MEANS VERY LITTLE UNLESS A USER WHO KNOWS WHAT THEY ARE DOING RUNS THE APPLICATION I/O METER FOR DAYS WHILE THEY USE THEIR SYSTEM WHICH THEN TAKES INTO ACCOUNT 100% RANDOM ACCESS USER HABITS OVER A DEFINED PERIOD OF TIME.
 
Only then can REAL WORLD disk performance results be defined for a specific hard drive or storage system.
 
 
 
 
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 13503
Direct Link To This Post Posted: July-10-2013 at 10:39am
 
 THIS POST IS NOW COMPLETE....
 
 
 
BEFORE READING THIS:


I am making it clear now that this is not a first year MIT Engineering class I am addressing and for those who are highly skilled in electronics or have a background in the subject should clearly understand that I will be addressing this topic on a level that most users can comprehend. I will be using terms and examples that are not 'technically accurate' to someone who may be educated in the subject in order to help those who do not have any idea what they are doing develop a far better grasp of the concepts presented which will in turn give them the confidence to establish a stable high performance clock with their system.


In other words, this is a Flight Sim hardware site and not a engineering lab. In that, I will be keeping this subject, the terms and examples VERY simplistic in order to help those who have no concept of electronics and how to get the most out of them.

 

 

Clocking Haswell: The Basic Rules and Introduction


Before I begin posting Haswell clock setups there are some basic rules everyone should understand. There are internet posted outlines about clocking Haswell with Asus motherboards and most of them are based on Asus data that was released by their reps. I have reviewed that information and found it to be accurate is some respects and not so accurate or at the very least 'iffy' in others.


First lets look at the goals. There are three primary elements to a 'truly' high-end clock, not just one this is true for ANY clock not just Haswell...


CPU SPEED
MEMORY SPEED/TIMING
UNCORE or Northbridge SPEED


These are "The Three Kings" to clocking and each one of them is important. If one is out of whack, the rest suffer however Haswell has changed how one of them works.

 

 

CPU SPEED is the FIRST KING -

This has always been the case however if you have ever read one of my discussions in the past about memory speed and timing then you know my take on this. The highest CPU speed is of course what we strive for but with FSX we do not want to take a 'major' hit in memory speed/timing (latency) nor UCLK Frequency (shown in CPUz on the memory tab as Northbridge Frequency). To do so means that 100Mhz higher tick up in CPU speed is being neutered by the system that feeds it data to process.


A CPU running 4800Mhz with low memory speed and very high memory subsystem latency is not performing better than a system running 4500-4700 with exceptionally fast memory communication support! This is something most people do not understand and the 'gamers' out there perpetuate this rumor since games and FSX are TWO DIFFERENT APPLICATIONS in the way they work and use memory. But GAMES still benefit from this regardless of what 'synthetic benchmarks' display.


At the same time we would not want to take a 3-400Mhz hit on CPU speed to allow just a bit faster memory speed or timing. There is a balance between these systems that must be found and maintained.


Ok so now that you are completely confused let me simplify this for those who do not understand the terms I just used..

Think of the CPU as a customer who places a order for a product they intend to use as soon as it arrives. Think of the system memory as a warehouse in which a business stores its goods for processing/shipping. When a order comes in, if the warehouse is slow then it will take time to get that product off the storage rack and up to the loading dock.

Think of the Northbridge (which has changed over the last 5 years) as the "loading dock" where the goods are sorted for shipping out. If this is slow then regardless of how fast the warehouse gets the order to the loading dock, it will wait there to be shipped.

When you order something online you usually like to see it delivered quickly, right? Well, that order has to be processed, sorted and prepared for shipping, then it leaves the loading dock and travels to you. You cant use that product until it arrives. The CPU functions the same way, when it receives the data it can then move to process it.


All of these factors effect the final real world result with a system.. so we strive to make sure all of them are tuned and running efficiently.

 


MEMORY SPEED AND TIMING is the SECOND KING-

As I just pointed out above, how fast the memory works with the CPU defines how fast the data gets to the CPU and is recalled if needed. There are 2 aspects to memory, SPEED which is the bandwidth it runs, and, LATENCY, or memory TIMING which is the real key to making USE of higher bandwidth.


Think of BANDWIDTH as the size of the road and the number of lanes cars travel on.

Think of LATENCY as the STOP LIGHTS along the way and with respect to memory, every lane of traffic must stop at each light.

Now, what happens when you have a lot of cars even on road that supports many lanes and allows a large flow of traffic when they come to a stop light AND those lights are LONG?  Are they getting to their destination faster or slower?

SLOWER! OK, so the next time someone tells you that memory speed and latency DO NOT MATTER regardless of if it is a game or FSX, DONT argue with a fool, don't point them here and don't waste your time with these people...  simply tell them; as Bill Engvall says; HERES YOUR SIGN!


The design of Haswell has introduced memory speeds with stability that were never possible in the past. Eventually we will all be moving into DDR4 which what these developments are all about, memory speeds above 3000MHz. But regardless of that bandwidth, if the memory TIMING is high (or slow) then the number of WAIT STATES that occur along the way increase, effectively slowing that memory down.


This means if you use DDR3 2133 memory at a timing specification of 11-11-11 that you are effectively SLOWER than someone running DDR3 1600 at a timing specification of 6-8-6. Now, I know this question is crossing your mind; How do I figure out what is faster so I do not get ripped off with the parts purchase. There are formulas that will help define the BASE but without importing the speed and efficiency of the Uncore or Northbridge and the speed of the CPU into those equations you cant define it 100% and can get tripped up using the formulas. It requires a great deal more knowledge of engineering than we will be getting into here.


To simplify this I have defined BASE values for you to use when shopping for a Haswell system parts in the 2nd post of this thread. Its simple;


DDR3 2133 C9-11-10    Which you would not want with Haswell since it will run 2400
DDR3 2400 C9-11-11   This is what most users should go for as it is blazing fast
DDR3 2800 C11-14-14  This can be run but with even with Haswell remember that high CPU clocks can make any speed above 2400 difficult to obtain stable.
DDR3 3000 C12-14-14  NOTE: This is not easy to run at high clocks, stick with 2400-2800 unless you KNOW what you are doing with high-end clocks.


The C stands for CAS Latency.,. it is the primary number and the most important, but if those sub values are very high it effectively REDUCES the efficiency of the CAS value. This is where you must know what you are doing or you can be snookered by a memory company into paying hundreds of dollars for faster SPEED memory that is reality NOT FASTER. They do it all time..   example:

DDR3 1600 C6-8-8   this is SLOWER than DDR3 1600 C6-8-6 or C6-7-8


DDR3 1833 C9-9-9  this is FAR SLOWER than DDR3 1600 C7-8-7!

DONT FALL FOR IT. The memory companies rely on the ignorance of users looking at the SPEED value.

Lower timing at the same speed is ALWAYS better so if you can find memory that runs lower CAS and supporting numbers than what I show for the same speed, its faster but look OUT, that cost usually goes THROUGH THE ROOF and you wont ever get an equal performance value for the cost. Memory companies rob us tech junkies and make us pay through the nose for the best of the best. The values I outlined above will provide the goodies for a reasonable price.


Last, always remember that the faster and lower the latency the more difficult it is to maintain HIGH STABILITY at faster CPU speeds. That is why we must find the balance where the memory speed and latency are running its peak performance with the CPU speed in operation. When clocking above 4.5Ghz always remember this:


If you have a choice to take a 100-200MHz hit on CPU speed to go from DDR3 1600 6-8-6, DDR3 1800 or even 2000 to DDR3 2400 9-11-11, DO IT! That means if you can't run 5000MHz @ 2400 C9-11-11 but can run that speed at 4800, DO IT, you are running far smoother and more efficient than the system running 5GHz on 1600-1800 memory.
 
If you are going from 2133 to 2400 then DONT DO IT. If we are talking about increases in memory speed of 300Mhz or less, shoot for the higher CPU speed and trim the memory in the best you can.

 

 
 
 
 
UNCORE or NORTHBRIDGE is the THIRD KING -

In the past Uncore has been defined as 2 x MEMORY SPEED. This is important to the overall latency of the memory subsystem. i7 made this system far more efficient than it was before i7 hit the market.  Before i7 the northbridge was significantly more critical to performance and we used tricks in the BIOS to get the latency lowered at the northbridge so that 'loading dock' was shooting out orders at higher speed. Some of you may remember my lectures with CORE2 processors on the STRAP and use of engineering software to tickle that northbridge to lower the wait-states.

With i7 that changed and was one of the reasons why i7 leaped forward with performance. With Haswell, this has changed AGAIN. What Intel has done is fired all the lazy workers at the docks who spend a lot of time on breaks and completely replaced them with high-efficiency automation.

We no longer must run 2 x memory speed to obtain the latency result! We can now run 1.5x-1.6x-1.7x and 1.8x the memory speed and still be FASTER than i7 of the past. This is critical for TRUELY STABLE high-end CPU clocks of 4.5Ghz and above and the primary item Asus missed in their clocking setup outlines.

Of course the faster that speed the better, but let me give you an example here of just how efficient the Haswell system is with respect to this...

If you are running a CPU speed of 4800 and your memory speed is 2400 in the past we would want a UCLK or Northbridge speed of 4800 or 2x memory speed. With Haswell we can run a Northbridge speed of 4300 and be running FAR faster i7 than of the past.  Further, if the Northbridge frequency had to be dropped to 3900 from 4300, we only lose 1 nanosecond or LESS of latency in the memory system speed and we are still faster than i7 of the past!

That is an impressive GAIN in technology. It opens the door to very HIGH memory speeds with low latency and allows stability at high CPU clocks with little LOSS to the true speed of the system. I will say again, the faster this is, the better, but we no longer have to strive for or over-volt in order to obtain the result. Therefore the clocking outline with Haswell has been modified from guides of the past with respect to this setting in the Asus BIOS which is controlled by: MIN/MAX CPU CACHE RATIO.

 

 
 

THE DANGERS OF CLOCKING HASWELL

I have stated earlier that I can not define every Asus motherboard or especially motherboards that are not made by Asus. Therefore what I will outline during the clocking process is a way for YOU to verify your voltages and ensure they are within SAFE tolerances with respect to Haswell. I will outline the software to use and its version so you have the correct tools at your disposal for success.


With Haswell there have been some design changes with respect to current and voltage control. The design of Haswell itself can invoke automated increases in voltage past what you set in the BIOS regardless of those values being MANUAL or OFFSET.


Further: The use of ADAPTIVE voltage which is a feature in the Asus BIOS which Asus claims will allow far greater power efficiency comes with a very DANGEROUS caveat as it will allow a HUGE jump in voltage even if the adaptive offset is LOW.


In example; you could calculate and enter a adaptive voltage number that would place the total CPU VOLTAGE at 1.22v and if Haswell is hit with certain load criteria it will override that and you could be staring at a CPU VOLTAGE of 1.375 in the blink of an eye. Experienced users know that such a jump in CPU voltage could be deadly to a processor if the heat transfer system cant handle that voltage. You inexperienced users must understand that for each .10 volt increase in CPU voltage the heat goes UP and with Haswell it can rise significantly fast without the IHS modification that was discussed earlier in this thread and that will be outlined later to show everyone how to do that.


Therefore it is my recommendation that unless you are exceptionally trained in electronics and know how to read and define the increase that will occur under certain LOAD conditions that all users stick to CPU CORE VOLTAGE: MANUAL ONLY and not use offsets or adaptive to control Haswell at higher CPU clock. Be aware that even with manual voltage control there will still be automated increases in voltage, just no where near as much.


With that being said I think its time to start outlining a BIOS setup for clocking however you MUST REMEMBER that because of the confirmed issues that exist in the thermal transfer contact point between the CPU die and the cap that covers it THAT WITHOUT A MODIFICATION TEMPS WITH HASWELL WILL BE VERY HIGH AND CPU CLOCKS THAT REQUIRE MORE THAN 1.20 VOLTS ARE GOING TO NEED SUFICIENT COOLING TO ACCOMPLISH THEM SAFELY.


I have already outlined in post number two (above) of this thread what is and will be required for the clock you intend to run. If you did not read that post then go back up NOW and carefully review it. Failure to do so can result in failure to achieve the clock you wish to obtain OR can result in damage to the CPU

When we test Haswell for FULL LOAD temps we will be using software that will invoke higher loads and temperatures than the CPU will normally run and will most likely never see in real world use. In that, if you decide to NOT run those tests and instead decide to run tests that invoke a far LESS load on Haswell in order to SKIRT the temp issues.. that will be a huge MISTAKE as you will never know if your system is really stable or not.

With Haswell it is possible to run, in example, the OCCT:CPU test and completely pass with flying colors even reading temps of 58-63x MAX and still be UNSTABLE. Understand that when we pull something like that it means those stutters, CTDs, strange crashes and performance issues in FSX may be DIRECTLY RELATED TO THE CLOCK WE ESTABLISHED AND NEVER TRUELY TESTED FOR STABILITY.

Therefore I am letting everyone know right now that what I just stated above about running a lesser CPU load test to pass it with a clock can be done, but it WILL NOT define true stability with Haswell and you could be facing more problems than they are worth dealing with had you run the proper full load tests I outline below. That is why it is exceptionally critical that if you do not intend to modify the CPU IHS contact that the RIGHT COOLING SOLUTION AND TOWER FOR AIRFLOW BE PURCHASED.. and even with the IHS mod you should still make sure you are covered.. the lower the quality of those purchases the LOWER the stable clock.

 

 
 
 
AND LAST..   CHEAP MOTHERBOARDS ARE CHEAP FOR A REASON! Its not just the features they do not have that makes those boards cheaper to the user, its the cheaper parts, the cheaper heat transfer components, the cheaper filtering and voltage regulation that makes them CHEAP.   BUY A CHEAP MOTHERBOARD AND A CLOCK / MEMORY SPEED STABILITY ARE A UNKNOWN FACTOR AND EVERYTHING FROM THIS POINT FORWARD GOES OUT THE WINDOW FOR SUCCESS.  
 
 We will be clocking with C States enabled so you do need that Haswell certified PSU
 
 
 
 

That concludes this section of the thread..   next stop, the Asus BIOS and how to check your clock and your cooling solution.

 

 
 

 

Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 13503
Direct Link To This Post Posted: July-14-2013 at 8:24am
 
 
 
 
THIS POST IS NOW COMPLETE....
 
 
 
 
 
 
Haswell and Overclocking 101 -
 

PREPARE FOR COMBAT

 

Although there are many ways to approach clocking a processor, what we will accomplish here is the easiest path to success. This is a Flight Sim site. This tutorial will not go into the engineering of clocking. Its designed to get a user from point A to point B in the most efficient and the safest way possible.

The term 'SAFE' is only relevant to the fact that what you are about to read is based on the technical data and the real world testing I have worked through and decoded for typical users to apply. I can assure you that none of the voltage or other settings I outline when applied to the correct TEMP LIMITS will exceed design specifications for the processor.

That DOES NOT MEAN ANY overclocking is SAFE and that includes the low-end 'simplistic' and 'automated' clocking the motherboard manufacture may provide through the BIOS or though their software. When we exceed manufacture specifications for clock speeds we take on the RISKS that come with that.

In example, you could have a chip that exhibits no sign of defect at all and for all intents and purposes should not have any issues with running WELL above its rated speed, the majority of processors will all pass that test. But there can be processors out there that have weak internal component or that may have passed QC testing but are on the edge of failure at which point clocking and raising voltages can push them over the edge to fail.

The odds are totally in your favor that wont happen if clocking is done right but that is not to say it WONT or CANT happen.

In other words, you clock it..  you bought it and no one here is responsible for the outcome other than the lose nut behind the keyboard.

 

This process is not hard to accomplish. It simply requires the right tools, the right approach for security and the RIGHT amount of patience to test and confirm you are running within temperature and voltage specs.

 
 
SPECIAL ADDED NOTE: User who do not edit their BIOS manually and use Asus Windows or automated BIOS tool to low-end clock their system automatically MUST USE THE TOOLS OULTINED BELOW AND TEST THEIR SYSTEMS FOR STABILITY AND HEAT THE SAME WAY A USER WHO ESTABLISHED A HIGH-END CLOCK WILL TEST, NO EXCEPTIONS. The Asus software may hit that stable low-end clock, or, it MAY NOT. You have to verify that with the correct software before assuming you are SAFE and STABLE. Also, clocking and installing software rules are the SAME for you.  
 
 
 
 

Now that it has been stated, lets move into preparation for the goodies.
 


==================================================

 

 
 
This outline assumes you have or you will do the following:


1. You have made sure you are running the most current BIOS by checking the specific Asus motherboard page and in the support section verified no BIOS update is available. With new motherboards typically BIOS updates are CRITICAL since this is a new platform and new CPU. Verify the BIOS is in fact the latest and if not, update it properly.

If you are new to the EUFI BIOS and how to update it properly I have outlined the process at the end of this post.

 

2. You have set the BIOS: AI Overclock Tuner to XMP, which loads the on-stick memory programming and the memory is in fact running its correct speed, timing and voltage which the XMP profile should load automatically, and then reboot.
 
 
 
3. You have tested each individual stick (one at a time) of your memory correctly using Memtest 5.0 which is required for Haswell memory tests at this time.

http://gex.flight1.net/forumimages/Memtest86_5.00.zip

THREE full test passes per stick (10 individual tests per pass).. verify, swap sticks and repeat until ALL sticks of memory are confirmed as NOT DEFECTIVE.

If you are new to testing memory correctly with i7 and Haswell I have outline the process at the end of this post.

 

4. You DO NOT intend to clock a system and then proceed to install software and set up Flight sim. WE DO NOT CLOCK a system and install Windows or Flight sim or any other large software routine. We must verify our installs are working correctly along with all the hardware working correctly first.
 
 
 
5. You have DISABLED UAC and WINDOWS DEFENDER: Shut these Windows Features down immediately after Windows is installed and then reboot. This should be done before any drivers are installed.
 
Before you install any drivers into a clean Windows installation perform this task and reboot, then begin installing the motherboard drivers.
 
Windows UAC (User Access Control) is useless to protecting a single user computer. It was designed to protect a office network from malware spreading across the network but will not protect the original computer system the malware was executed on, nor will it protect a single user system.

If you use FSX and 3rd party addons by now you are probably already aware that User Access Control can block proper function of addons and even FSX. UAC should be completely shut down for both system performance as well as stop issues with installing and running addons.
 
Once this is done selecting "RUN AS ADMINISTRATOR" to install or run programs is no longer required (except in the case where a 'administrator command prompt' is required)
 
Windows Defender is well known for being a resource hog and blocking the proper installation of addons.

To disable Windows Defender in Windows 7 do the following:

a. Go to Control Panel: In the corner dropdown select VIEW BY and then 'small icons' (or large, your choice) then select Windows Defender from the Control Panel list and open it.
 
 Under the TOOLS menu select OPTIONS and then do the following:
 
b. Select: Real Time Protection and uncheck: Use Real Time Protection
 
c. Select: Administrator and uncheck: Use this program
 
d. Click SAVE and then if any popup appears, click CLOSE.
 
e. Go to Control Panel - Administrator Tools - Services   In the services list scroll down to: Windows Defender and double click it, a box will appear. If the services happens to be running at this time, click STOP and then in the dropdown select: DISABLED and click APPLY. Close the services list.
 
 
To disable UAC in Windows 7 do the following:
 
a. Go to Control Panel - User Accounts - Change User Account Control Settings

You can also click START and in the search box type: UAC then click on; Change User Account Control Settings

b. In the box that appears set the slider to ZERO (Never Notify) and click OK and reboot the computer.
 
Done
 
 
 
Windows 8 users, you have my condolences. I wouldn't touch Windows 8 for a FSX system. Properly shutting down UAC is completely different in Windows 8. You can review this post about that subject: http://www.simforums.com/forums/windows-7-and-windows-8-uac-blocks-gex-interface_topic45785.html
 
 

7. The motherboard drivers have been installed FROM THE DISK THAT CAME WITH THE MOTHERBOARD in the correct order starting with the chipset drivers.
 
It is important to install drivers in the correct order after a operating system is FIRST installed. The CHIPSET driver from the motherboard disk is always installed first, reboot, and followed up by the SATA system, reboot, and then any other drivers with reboots between each driver installation. The chipset drivers define the chipset registry in which the BIOS of the system works hand-in-hand and from there all the hardware drivers installed into the system will register correctly with the motherboard since the BIOS and its chipset have established the correct ID and driver link in Windows.
 
Always REBOOT after a chipset or any driver is installed before installing the next driver.
 
CRITICAL: It is very important that you install your drivers from the original CD or DVD that came with the motherboard BEFORE applying any driver updates or upgrades. You do not need the bloatware software most motherboard disks come with but you do need to install the drivers from that disk using the motherboard manufactures installers. There is a very important reason for this!
 
I have seen this issue with users systems many times over the years. Users will skip the motherboard disk and go round up the updated drivers first. This is a serious mistake even if the board is older! Most users do not understand that the motherboard manufactures can place special patches and execution code into their product driver installers. This code is executed on first install of drivers to the motherboard and is many times not included with updated drivers they post at their website support.

In just one example, the Intel chipset drivers for motherboards do not fully install without executing command line code (same with the Intel website drivers) which the motherboard manufacture programs into their custom installer to execute. In order for Intel chipset drivers to update properly the first chipset driver install must be run with the proper execution commands and then from there they will be fully installed and will update correctly with later patches.

That is just one example, another is the motherboard manufacture may scan for the OS and install a critical patch before installing a driver and that is expected to be installed before running Windows Update. The driver updates posted at the motherboard support site may or may NOT not include that scan or patch as it is only required on first install of the drivers and the manufacture expects the user to run that driver disk. This is why it is important to use the motherboard driver disk first, then after all the drivers have been installed from that disk a user can proceed to install any updated drivers listed at the motherboard manufacture support site.
 
 
 
 
8. You have BACKED UP the entire Windows drive using either the Windows 7 "Create System Image" OR 3rd party image software that is certified to work with Windows 7 and SSD drives for proper alignment in the event of a disaster and a recovery is required. Windows System Restore can NOT be used for this! You must create a backup image and STORE it away from the Windows drive.

If you do not perform that step and anything goes wrong with respect to corruption, you will be completely reinstalling Windows from scratch.

 
 
 

9. If your intent is to run clocking tests before doing a complete system install of all software then be very aware that you must make sure Windows and all the proper hardware drivers have been installed/updated, Windows updates have been installed and the hardware verified as working correctly with no defects, a complete backup the Windows drive as defined in #3 has been established in the event of a disaster or corruption. Once you have completed your clocking tests, have established a stable clock and recorded the settings, the system is to be returned to DEFAULT NORMAL CLOCK USE BIOS setup for completing the install and is NOT to be clocked while going through that process!

 


Now that we have established the RIGHT environment for this, we can proceed to the next step...

 

 
 
TOOLS YOU WILL NEED


WARNING: If you do not follow my instructions below which includes the use of the RIGHT SOFTWARE to monitor and test Haswell, it can result in damage to the CPU or failure to verify stability.

 

 
The versions of the following tools are required for proper Haswell verification and clock testing:
 
 
MONITOR AND VERIFY:

CPUz v1.65 or above: http://www.cpuid.com/softwares/cpu-z.html

HW Monitor v1.23 or above: http://www.cpuid.com/softwares/hwmonitor.html 

The automatic setup installers will install the correct version for your system (x32 or x64)

Both products are free with no cost.

 
NOTE: RealTemp is in the process of being updated with the latest HW Monitor and Haswell information. RT v3.70 can NOT be used at this time with 100% reliability. However the TECH-INFERNO version of RealTemp can be used with specific instructions about setting it up. http://forum.techinferno.com/downloads.php?do=file&id=43

I suggest you not use this software unless you know exactly what you are doing and stick with HW Monitor for now. I will include a defined setup for RT T|I later in this document if you wish to use it for monitoring your system in the tray.

 
 
DO NOT RELY ON THE ASUS MOTHERBOARD SOFTWARE TO DEFINE CPU TEMPS. Some of the Asus motherboard Windows software uses calculated formulas for T-CASE temperatures that fall 10-15c LOWER than the real temperature of the processor at full load. Using that software to define the true temp of Haswell in a full load test can result in permanent damage to the CPU since it will be reporting temperatures FAR below the real temp of each core.

 

 
LOAD TESTING:

AIDA64 Extreme Edition v 3.00.2500 or above: http://www.aida64.com/downloads

We will be using the STRESS FPU selection (all other tests deselected) of the TOOLS > System Stability Test for checking max heat and stability at the cores. This test will heat Haswell far higher than ANY OTHER TEST.
 
NOTE: The trial version allows a 30 day time limit in which to use the product.  The purchase price is 40 dollars. You do not have to purchase the software but you do need this tool in its trial state to verify your clock. The software is required at this time to define Haswell HEAT and THERMAL properties with your cooling solution and the clock you set up. It can also be used for far more however it is the ONLY CERTIFIED Haswell test product at this time for what we will be using it for and is required for checking your processor for stability and heat issues.

 

OCCT v4.40 or above: http://www.ocbase.com/index.php/download

We will be using the OCCT:Linpack test with 64bit and AVX enabled along with 90% of the memory loaded at the same time which stresses the cache and IMC of the processor and double checks memory too.
 
 
OCCT is free with no cost.
 
 
 
Download and install each of the list software products.

 

 

BIOS UPDATES: ASUS - THE UEFI BIOS AND SETUP
 
 

 
There are many users out there who are coming from older system that have never worked with the newer UEFI BIOS systems.

For those of you that have you are already familiar with them already know how they work but for those who do not and have never see one in operation its important to understand a few things about it so you are prepared.

First and foremost, unlike BIOS design of the past a UEFI BIOS can take longer to boot, can boot and appear to hang (but isn't hanging) and then proceed to reboot, sometimes several times on a clean first start with new hardware. Do not let this alarm you as it can be normal. You will need to get used to that change so you do not mistake a 'hang' or lockup with a normal boot delay.
 
 

Here are the primary items that you should understand as you get started...
 

1. Unlike BIOS systems of the past the UEFI BIOS will allow you to navigate the BIOS similar to Windows with the mouse and click points. This is a great feature.
Boot the system. The full screen logo will appear. The system may boot and then reboot once or twice on first run. Enter the BIOS by hitting F2 or the DEL key.
 
There is typically a E-Z BIOS setup and a ADVANCED BIOS setup which are two completely different areas. We will always be working in the ADVANCED BIOS setup. There will be an option to switch to advanced when you enter the basic EZ mode. You can also go to the EXIT menu as it will have the ENTER ADVANCED MODE option.
 
 
BIOS UPDATES - Important
 

2. With a new system motherboard there can be a important BIOS update available that deals with issues that were found just before or just after release of the motherboard to market. Issues and fixes with new motherboards are very important to be installed prior to doing anything else. I usually do this prior to installing Windows on a new system build that way any code changes that may effect chipset driver and other onboard driver installs is covered.
 
 
CHECK YOUR VERSION - Go to the MAIN Menu and record the BIOS information. The top of the list of information will display the BIOS VERSION and date. You will be surfing to the Asus website for the board to confirm the BIOS is the latest or if a update is available now.
 
Go to Asus and locate your board. For an example here is the Asus Sabertooth Z87:  http://www.asus.com/Motherboards/SABERTOOTH_Z87/ 

 
You must locate the correct Asus website page for YOUR MOTHERBOARD.
 
At the top-side of the page for YOUR board, click SUPPORT and then click the DOWNLOAD tab in the middle of the page. In the dropdown that appears select your OS. In my case and with my OS that is; Windows 7 x64. Be sure to select the correct OS and bit version.
 
A list will appear, click: BIOS and check the list. If you see a BIOS file that is numbered higher that your installed version then you need a update. Download the new BIOS file.
 
 
With the UEFI Haswell BIOS the process from the past is similar with some important differences you must understand.
 
a. The BIOS file provided at the Asus support site for the board MAY require a 'rename' process. The instructions that come with the file should outline that and what must be done. Sometimes it is needed, sometimes not. It should be noted.
 
b. Unzip the file (rename if necessary) and place it on a thumb drive that is FAT32 formatted. Boot the system with the thumb drive and then enter the Advanced BIOS: Tools menu and select the Asus EZ Flash utility. Follow the directions to locate the thumb drive in the list and select the BIOS file, then allow the system to update the BIOS. It will take a minute or so to complete.

WARNING: DO NOT do anything while this process takes place and never shut down power. If there are storms in the area and you are prone to blackouts DO NOT update a BIOS at this time and wait for more stable conditions. DO be aware that Asus systems come with automated BIOS recovery options which will be outlined in your motherboard manual. If something does happen you can use the methods outlined to recover.
 
c. Once complete select to let the system reboot. THIS IS CRITICAL: Unlike BIOS systems in the past on reboot the system may appear to hang for a long period before it continues. This period can last a full minute. DO NOT shut down, reset or disable power! Let the BIOS complete this cycle and then you may see a message that "The BIOS is updating" and then reboot again when its finished.
 

It is very important that you allow these boot cycles to take place without interference! Once complete then the system will boot back to the point where the BIOS can be entered again. The BIOS update process is now complete.
 
 
d. Enter the BIOS and select F-10 and then select; LOAD OPTIMIZED DEFAULTS and then save and exit.. system will reboot. You are now fully updated and ready to go.
 

The BIOS has been completely reset. Any settings from the past must be established again.
 
 
 
GETTING STARTED:
 

3. There is typically a E-Z BIOS setup and a ADVANCED BIOS setup which are two completely different areas. We will always be working in the ADVANCED BIOS setup. There will be an option to switch to advanced when you enter the basic EZ mode. You can also go to the EXIT menu as it will have the ENTER ADVANCED MODE option.
 

Once done you should go to the BOOT menu where you will typically find a 2 settings named something on the order of:

BOOT LOGO DISPLAY: DISABLED
and
SETUP MODE: ADVANCED MODE


Set those two as shown. The first will allow the full boot screen data to appear and the 2nd will always boot you directly to advanced mode bypassing the EZ mode. In the same BOOT menu you can elect to change drive boot priorities but for the most part you should not need to change anything else in this menu unless you run controller cards and require some type of special support.
 

Please Note: Wireless keyboards and mice which are unified though a single USB port such as Logitech wireless keyboard/mouse work fine in the BIOS but may malfunction and not work correctly during operations such as Windows installs from the DVD if the following setting in the BIOS under the Advanced Menu > USB Configuration is not set to:
 
EHCI HANDOFF: ENABLED

That can be set to disabled by Asus by default. If you are effected by this issue, it must be enabled. If you are not effected by the issue leave the setting at its default when installing Windows from the DVD or running a Windows repair. THIS IS ONLY NEEDED FOR LOGITECH WIRELESS UNIFIED DEVICES. Individual USB wireless devices typically do not display any such issues however if you see them you can change the setting above.

Further, under the same menu if you use PCIe storage controllers that use a Linux based boot disk for toolbox access, this setting may need be disabled before use if the disk fails to boot to the utilities on it:

Intel xHCI MODE: DISABLED

That must be re-enabled after using such boot utility disks for proper operation in Windows. It only needs to be temporarily changed if you have issues with such tool disk boots. Otherwise leave it on the default setting of: SMART AUTO
 

4. Set up your memory correctly. Under the AI Tweaker Menu set:

AI OVERCLOCK TUNER: XMP


This will load your memory module on-stick chip data into the BIOS and set the speed, timing and voltages to manufacture specification. At this time DO NOT change memory timing and leave it set to manufacture specification. Later you can test for "Command Rate" changes after the memory has passed MEMTEST and the software has been completely installed and verified.

 
5. Set up the SATA system for the drives you use. Under the Advanced menu - SATA Configuration. If you use any SSD drives make sure the SATA MODE SELECTION is set to: AHCI

Hot plug settings are your choice, leave the rest alone unless you know for a fact one of those settings needs to be changed due to a specific drive requirement.
 
If you use nothing but mechanical drives and know for a fact you will never run SSD NOR do you want or need 'HOT PLUG' support, then set the SATA MODE to: IDE
 
 

6. Advanced Menu - System Agent Configuration, set the following:

CPU AUDIO DEVICE: DISABLED    


NB PCIe CONFIGURATION - Verify the video card is running x16.

If your video card is PCIe 3.0 then set: PCIEx16_x LINK SPEED: GEN 3   
DMI LINK ASPM CONTROL: DISABLED
PEG ASPM: DISABLED
 
 

7. Advanced Menu - Memory Configuration - MEMORY REMAP and (if there) MEMORY SCRAMBLER (both): ENABELD
 
 

8. Advanced Menu - Graphics Configuration -
PRIMARY DISPLAY: PCIE
IGPU MULTI-MONITOR: DISABLED
 
 
 
9. Under the Advanced Menu - CPU Configuration
 
Hyper Thread - DISABLED
 
 
FSX make NO USE of Hyper thread and if you want to clock Haswell with ease of stability and lower temps, shut this DOWN

 


10. Advanced Menu - Onboard Device Configuration - 
 
Set these as you need them. I can not define your required settings for this section. If you don't know, leave it at default. 
 

 
For right now and unless you know for a fact you need BIOS settings addressed for special hardware, leave everything else alone. You should be ready to run MEMTEST FIRST and then install Windows.
 
 


i7 SYSTEMS AND MEMORY TESTING - THE TEST METHOD
 


a. Use version 5 of Memtest and it covers all processors of the past. The latest version of Memtest (v5.0) that supports Haswell is linked here: http://gex.flight1.net/forumimages/Memtest86_5.00.zip  
 

b. Burn a disk with mt500.iso, or, create a USB boot stick with the USB installer. You can review the release notes.


NOTE: Motherboards on the market can define different settings for USB stick booting and because of that I can not define the settings in your specific BIOS for USB stick boot. Check the motherboard manual if you wish to use a USB stick over burning a CD.
 

c. Power down and remove all but ONE memory module from the system. Make sure this module is in the first slot the motherboard manual defines for a single stick of memory. Typically this will be slot 'A2' but VERIFY with the motherboard manual. Leave that stick in the system, remove the others and then proceed.

 
d. Boot the computer with Memtest. NOTE: DO not use any F modes when launching Memtest unless you know your motherboard requires those. Simply let the test start automatically OR hit the spacebar once to begin. If you have issues running Memtest, try F-1 or SAFE mode when the program starts, otherwise just let it run.
 
 
e. Memtest runs 10 sub tests per full test pass. You want to run at least 3 FULL test passes on each module which can take from 1-2 hours per stick depending system speed.
 
 
f. When the first module passes or fails, power down, remove the module and mark it for 'pass' or 'fail' and then install the next module in the same slot and rerun the test. Repeat until all modules are confirmed individually.

A FAIL will be defined by RED text and lines in the lower screen of Memtest. If any RED appears, you can stop the test, the memory has failed.
 
 
 
If you find a module(s) fail and the system is not overclocked then you most likely have a defective memory module which needs to be replaced and not used. Defective memory can lead to everything from Windows corruption, driver corruption, software application corruption, poor performance, intermittent crashes, system lockups and CTD's. (crash to desktop).
 
 
 
 
NOTE: Lets assume all sticks fail.. there is a possibility you may have a defective motherboard memory slot. You can use a different slot and retest. Check the motherboard manual about which slots are the first in each channel bank and if it is possible to use another slot for 1 stick. Usually you can use the B2 slot instead of A2, or the motherboard manufacture may define slot A1. The goal here is to define if the issue is a BIOS setting or a memory slot, or, you really have multiple sticks of memory that are defective.. If you find the sticks pass in another slot, the XMP profile was set correctly and the primary slots are failing memory when the sticks are installed in then, then return the motherboard for a replacement.
 
 
 
 
Once you have confirmed the memory is not defective then continue on.
 
 
 
MAKE SURE YOU HAVE GONE THROUGH THE OUTLINE PRESENTED ABOVE AND HAVE INSTALLED, UPDATED AND BACKED UP WINDOWS CORRECTLY

 

This concludes the preparation phase. Once you have completed the above you are ready to begin the process of clocking and testing your new Haswell system.

 
 
You are now: READY FOR COMBAT
 
 
 
 
 
 
 

Next section: Haswell and Overclocking 102 - ENTERING THE COMBAT ZONE

 
 
 
 
 

 

 

Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 13503
Direct Link To This Post Posted: August-12-2013 at 7:57am
 
 THIS POST IS NOW COMPLETE....
 
 
 
 
 
 
Haswell and Overclocking 102 -
 
ENTERING THE COMBAT ZONE

 

 

 

This section assumes you have read every post in this thread and have followed the guidelines established for setting up a proper Haswell clock.

 

 

It is important to note that regardless of CPU and memory load tests or how long they are run, overclocking can still produce errors. There is no such thing as a 'totally stable' high-end overclock when 3D render is added to the load. The ODDS are far better after extensive load tests confirm there is no exceptionally unstable issue present and if a user knows what they are doing they will probably not see any issues however be very aware that regardless of personal experience you can pass every CPU/memory load test with flying colors and STILL not be 100% stable under 3D render graphic load conditions which may require speed or voltage adjustments to trim out.

 

If a user is not well versed in overclocking this can also be true of LOWER end clocks. This is true of any system overclock and not just Haswell.

 

 

 

I am going to address ALL clocks setups from 4.0GHz through 4.8GHz with Haswell. I have developed a routine for this that will be nearly universal for all clocks in this range. In doing so it has simplified the process by quite a bit.

 

 

I will not be addressing clock speeds of over 4.8GHz. There is a reason for that. In order to obtain such a clock you will have to be running a far better cooling solution than the typical Corsair liquid system can provide, and, the processor you received would have to be one of the top in its class for low voltage/high CPU speed. Since both of these factors will be RARE I will be keeping the clock setup outline to a 4.8GHz max speed.

 

 

If you are intend to push the CPU higher, and you actually KNOW what you are doing, you can use the same information I post here for max temp/max voltage and without exceeding those defined limits attempt a 4.9-5.0GHz clock, but be aware as of the date of this posting no one is running a 24/7 safe voltage clock and stable on Haswell @ 5.0 and above on air or liquid. Those speeds are not a realistic goal with Haswell, 4.8 and below is, assuming the processor will allow it.

 

 

 

 

INFORMATION YOU SHOULD KNOW:

 

The first time we clock Haswell and before considering modification to the thermal contact point between the cap of the processor and the processor die (or IHS mod), the first order of business will be to establish where your individual Haswell processor falls for voltage and clock speed and at what temperature. We want to establish that the clock we wish to obtain can be achieved and if not how high the system will go and for those who are considering a IHS modification, to define if you really need or want to modify that processor thermal connection point which I will outline in another part of this thread.  

 

 

Asus has outlined their own approach to clocking Haswell. They define the memory should not be XMP set which runs it at it lowest speed and highest timing, and instead clock the processor as high as possible first and then set the memory to XMP. I disagree with this approach as it sets the user up for a CPU speed their memory and UCLK can not follow without excessive voltage and current. This is unrealistic to what a user would want to achieve and is in my opinion designed for marketing purposes over real world use. Instead, my approach establishes the memory manufactures speed/timing for the base overclock of the CPU which is what a user will run in the real world.

 

 

 

We will be enabling C-States in the system BIOS, changing them and using that system to lower the CPU temp at idle. This goes a long way in extending the life of a processor when it is not under full load and is why the Haswell certified PSU is needed.

 

 

Haswell has two primary voltage/load states in which the top-end voltage/load temperatures will probably not be seen unless a software product that is 'AVX' instruction enabled is run but that does not mean we do anything FOOLISH such as use ADAPTIVE CPU Voltage. In this day and age and with a Flight sim tower you will most likely never hit Haswell with AVX instructions except with certain load test software which we will use to define the top-end temperature and stability of the clock, but that is not to say at some point a software you use may not introduce those instructions, therefore we NEVER use Adaptive CPU voltage when clocking Haswell because regardless of the 'calculated adaptive value' under the right conditions a CPU voltage 1.35 and GREATER will be invoked, instantaneously and without a warning.

 
 
After we have established the top end temp/voltage the normal use RUNNING temp will be BELOW the max load temps under normal use and high speed. In example, I hardly see temps higher than upper 50's with perhaps a spike in the low 60's running FSX @ 4800MHz on a air cooled Haswell that was modified for the cap thermal connection.

 

 

 

 

 

TWO VERY IMPORTANT FACTORS TO CLOCKING HASWELL

 
-AND-

HOW FAR CAN YOU GO

 

 

 

Lets assume you have the best liquid system on the market money can buy....

 

-First and foremost the QUALITY of the processor in the manufacture process that you received. You may find your Haswell will clock 4.8Ghz on a manual locked 1.35v and fully load test stable, or, you may find your Haswell processor may require 1.45v to run 4.8GHz with all load tests and run stable. The same is true for lower end clocks, you may be able to run 4.5GHz @ 1.18v or it may require 1.32v or more!

 

Unlike Sandy bridge there is an exceptionally large window of voltage vs. clock speeds with Haswell. Mostly due to the unreliable HIS thermal contact in the manufacture process and also driven by the far higher clock-per-clock performance increase in the design of the processor over processors of the past where Haswell gets hot, FAST.

 

This means it is impossible to define an 'exact' CPU VOLTAGE starting point for all systems, NOR is possible to estimate where YOUR Haswell will clock and at what voltage, however the method I post below should get you to the point where you know the limit and where your system stands without a lot of hoops and ladders to jump through.

 

 

 

-Second, the quality of the IMC in the processor will define how well UCLK will run in relation to the speed. The memory speed and the number of sticks installed can also influence that because the higher we go in memory speed and the more memory sticks in the system, the less stable that area of the processor can be. As CPU speed rises more voltage is required and that voltage requirement goes UP with the number of sticks and the faster the memory speed.

 

 

I posted that DDR3 2800 is the highest stable recommended speed for memory, however that does not mean you will be able to run high UCLK if your Haswell CPU IMC happens to be a bit finicky. That is why I defined 2x4GB DDR3 2400 9-11-11-31 is a better bet for users who intend to try for speeds greater than 4.5GHz and are not tech very savvy with electronics.  

 

 

 

 

 

QUICK REVIEW BEFORE STARTING:

 

 

Synopsis:

 

Going back to what I posted earlier in this thread, "The Three Kings", with Haswell that is how we will approach clocking the system. The first order of business is to stabilize the memory controller with the memory installed and run the manufacture speed and timing of the memory, and then establish a CPU speed/voltage that is stable below a defined temperature limit. From there we increase the UCLK frequency (or Northbridge frequency as CPUz displays) and establish its upper end, once those two are locked and stable then the process of reducing voltages starts with repeat load tests till the stable point is found at the lowest voltage settings.

 

 

Memory tweaking IF it can be done is always LAST after all software is installed. Memory clocking is a PERK, not a requirement and if you purchased the right memory most users should simply let the memory do its job at the manufacture rated speed with perhaps one minor tweak.

 

 

 

- Decide what CPU speed you intend to try and run for the initial tests: I addressed this in the 2nd post of the thread and explained very clearly what you need in the way of a CPU cooling solution to try and obtain clock goals. I will RECAP here again so you do not attempt anything foolish starting out and without the IHS cap modification.

 

 

*4.0-4.2GHz - The stock Intel heatsink MAY allow this but the system will most definitely run hotter on that heatsink. If you replaced the Intel heatsink with a Corsair H60 or a larger air heatsink, these speeds will work for you at far lower temperature.

 

 

*4.3-4.5GHz - ON AIR you will need a 'super-cooler' class air cooler and ON LIQUID you will need a Corsair H100/H110 or better quality liquid cooler. On the air super-cooler your limit may very well be 4.4GHz and REMEMBER even with liquid cooling if your Haswell is not a good clocking chip you may not even get past 4500! This All COMES DOWN TO: The luck of the draw.

 

 

*4.6-4.8GHz - ON HIGH END LIQUID ONLY WITHOUT THE IHS MOD and be very aware you may not be able to achieve either speed if the CPU requires more voltage than the average Haswell processor.

 

 

In all cases Intel Hyper Threading will be DISABLED. Hyper Threading serves NO PURPOSE with FSX and if you desire the system run with Intel Hyper Threading enabled you will most likely have to accept a lower clock speed/CPU voltage or exceptionally higher temperatures.

 
 
I recommend users who wish to have HT enabled for professional audio/video or engineering applications to setup and test TWO STABLE BIOS clock outlines and save the clocking profiles using a USB thumb drive and the Asus OC tool for profile saves in the BIOS so switching back and forth between the "upper-end" clock and lower clock with hyper threading enabled is easy to do.  

 

 
 

 

Up Next:  Haswell and Overclocking 103 -  THE COMBAT ZONE
 
 
 
 
 
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 13503
Direct Link To This Post Posted: August-29-2013 at 9:09pm
 
 
 
 
 
 THIS POST IS NOW COMPLETE....

 

 

==============================================================

 

 
 
 
 
 
 

 

Haswell and Overclocking 103 -

 

 

THE COMBAT ZONE

 

 

This may 'look' complicated because of how much I have written here, but it really isn't. I can not assume your individual level of experience around clocking so I must address this subject based on the lowest level of user experience to assure all users have complete information at their fingertips as they follow along.

 

Experienced clockers will get the benefit of that too since sometimes they assume too much especially with a new system platform and they may think they can use their old platform as a base to clock a new one.

 

Users who have extensive experience with overclocking are already well aware of what it is I am about to say in opening this subject however those of you who are new to it or have little experience with clocking I do want to make sure that you too understand the following very clearly:

 

 

Remember that heat, voltage and 'hard load' run-time define CPU useful life which for most people is between 3 to 5 years but can extend into 5 to 7.

 

There is a difference between the manufacture estimated life of 10 years of a CPU, and, the actual "useful life" of a CPU and how long we use a system before replacing it.

 

Clocking can shorten the 'manufacture estimated life' of a CPU and can also shorten 'useful life' if done to extreme. If  clocking is done right none of the settings posted below will shorten the primary useful life of the CPU unless a CPU from the factory passed quality control but may be prone to fail prior to expectancy and that can happen as I posted earlier in this thread.

 

Be very aware that even if you purchase extended Intel warranty service that if the top of the processor is altered and the laser etched numbers and codes are drastically faded or removed, Intel will most likely not honor any extended warranty on any processor. So something as simple as polishing the processor cap can void such a warranty. 

 

The cooler and the less voltage applied to any CPU and the lower the system 'hard-load' 'up-time' is over a period of years, the longer the CPU life. If you are pushing a 24/7 system 'up time' clock that is run daily for hours on end under a hard load at very high voltage and/or temps, then that can lower the CPU useful life over time.

 

Be very aware of these facts. There is no way to calculate when a processor may fail or begin to fail. Clocking when done right and with the right approach is effective and will typically work for all those useful years but only to the point of where the CPU quality will allow, and, the user who would push as high as they can go on 'max' specs (or exceed them periodically) and run the system daily for weeks on end under such loads.  

 

 

I will address automated overclocking first. User who do not wish to work the technical side of clocking in the BIOS can opt to use the 'Automatic Overclocking' Asus BIOS clocking features or Asus automated Windows tools. Be very aware the automated BIOS and Asus software clocking users should still test that type of clock for stability and temps and use of this method may not be the most efficient or the coolest.

 

 

 

 

AUTOMATIC OVERCLOCKING:

 

 

Before you decide to go this route it is important to understand the following:

 

Automatic overclocking methods may be the most simplistic but they may NOT BE the most efficient, stable nor run the coolest temperature possible. On the contrary, the automated method may actually run the processor far hotter than setting up a clock manually.

 

The automated methods are set up by a Asus BIOS programmer and software engineer however their approach can be based on 'brute force' for stability which can run the processor far hotter on higher voltages than ever required.

 

I understand some users simply do not wish to deal with a BIOS or setting up a clock and would prefer a more simplistic 'one click' method, however be warned that if your CPU cooling solution can not handle the methods the Asus programmers place into their automated clocking features then automated clocking may not work for you.

 

You can try it first and if you find the temps too high and too aggressive then move to manually clock using the 4.0-4.1 outline presented under the Manual Overclocking section. Manual clocking is not that hard, it simply requires a bit of patience and the ability to follow a outline.

 

 

 

AUTOMATED CLOCKING:

 

First and foremost, make sure you have established the XMP memory setting in the BIOS for your memory modules which will set the speed, timing and voltage for the memory automatically.

 

Your BIOS should be ready for these automated clock speed changes at this point.

 

 

Asus 'BIOS' Automated Clocking:

 

This type of clocking is the most simplistic but it will usually net the lowest speed bump increase. It will not tune fans or other cooling system devices and the time it takes for the process to complete is only a matter of a few seconds. Because of that this method may use higher voltages than the Asus automated clocking software in Windows.

 

Most Asus boards have a feature listed in their BIOS under the ADVANCED MODE >  AI TWEAKER menu named: OC TUNER  scroll down to locate this feature in the list

 

The default setting for normal operation is: [AS IS]

 

By setting this feature to: [RATIO ONLY] the BIOS will automatically request a reboot in order to run through a short series of tests. At the end of the tests the system will reboot and once in Windows you will find that under a load your CPU speed will increase and decrease automatically. It will also drop in voltage like it always has in the past. How high it will go under a load will be based on the BIOS and the board. Typically you will see a boost to 4.1GHz and there is no other change you need to make.

 

Some boards may read stability or temps and set that limit to 4.0, others 4.2. Most should see a boost of 4.1.

 

Regardless of where the BIOS sets your CPU speed, once the cycle has been set and is complete, you will need to verify your cooling solution can deal with the changes as well as confirm stability. Once this change has been applied you can proceed to the section: Haswell and Overclocking 104 - Stability and Load Testing Software and follow the directions.  (if you have already reviewed section 104 you may move directly to section 105) 

 

 

 

Asus Suite - Windows Automated 'Software' Clocking:

 

Do note that some boards do not come with a 'automated' Asus Suite 'clocking' software. The Asus Z87 Sabertooth does not. The only way to automatically clock such boards is to use the BIOS method. If you prefer to use the Asus Windows software method of clocking then your board purchase should be researched to assure it comes with the software to do the job.

 

The Asus boards that do come with automatic clocking software can vary in the type of software. Typically the boards that are more expensive come with the software "Asus Dual Intelligent Processors 4" and the less expensive boards the software "Asus Dual Intelligent Processors 2" The more advanced Dual Intelligent Processors 4 software will typically run fan speed adjustments at the same time the system is being calibrated for clock speed.

 

Be very aware you can not install Asus Suite software that does not come with your board. That software is specific to your motherboard intergraded circuits so trying to install and use Asus software that is not designed for the board will not work and could cause serious issues. Only use the specific software from the motherboard DVD and updates for that software posted at Asus for the board model you own.

 

The motherboard manual will define how to use this software as it is basically a 'one click' test and verify solution and is easy to use. This software is more aggressive than the BIOS performance clock method and can net higher clock speeds, but one must ALWAYS remember that the cooling solution and stability must be confirmed, regardless.

 

It is IMPORTANT TO NOTE that the more extensive software method can overclock a system PAST 4.1-4.3GHz if the testing cycle detects a cooling solution present through temp and fan tests that can deal with the heat the automated testing process invokes. Because the level of potential final clock speed can be higher, it can introduce the hazards of clocking that even those who manually clock may be faced with or even run a system far higher in voltages than manual clocking since most automated clocking tends to use a far more generic and brute force method of clocking than manual clock setups.

 

As with the BIOS clock method, once the software has completed it testing cycle and the clock has been established, you will need to verify your cooling solution can deal with the changes as well as confirm stability. Once this change has been applied you can proceed to the section: Haswell and Overclocking 104 - Stability and Load Testing Software and follow the directions.  (if you have already reviewed section 104 you may move directly to section 105) 

 

 

NOTE: DO NOT TRUST the Asus software to read temps and verify the stability, always take that responsibility into your hands directly with the right monitoring and load test tools for verification. It is not hard to do and the piece of mind that comes with that verification is important.

 

 

 

 

 

 

 

 

 

 

MANUAL OVERCLOCKING:

The Real Meat and Potatoes!

 

 

This is the preferred method to clock any system at any speed and be sure of stability with long term system reliability and use. This method locks frequencies so they do not vary and provide a user full control over the system. The method I have outlined will also take full advantage of voltage and current drops when the system is not under a load, therefore you get all the advantages of power saving and cooling without the potential performance or stability issues other methods of clocking may introduce.

 

 

Before beginning, and as I have outlined in the computer system bible to optimize Windows, regardless of the method of clocking do make sure your Windows power profile is set to high performance first. We do not want Windows changing your CPU core speeds or parking cores. Setting this up in Windows will remove those features and will ensure that stutters or performance related load changes will not create a problem under high load use or force the CPU to respond slow to load changes.

 

We are not interested in 'full' economy here, we want stability and raw high performance in FSX.

 

If you have not done so, please make sure that the system power profile has been set as outlined:

 

Go to Control Panel – Power Options. In the window that appears there is a dropdown that says: Show Additional Plans, expand it. Place a bullet in the power profile: HIGH PERFORMANCE and then click PLAN SETTINGS.

 

For the monitor you can set this to what ever you like, 20min is usually a good value. For the ‘Put Computer to Sleep option, select NEVER. Click the ADVANCED power settings link.

 

In the advanced power setting box go down the list and make sure these are set or change them as outlined:

 

HARD DISK – Turn off hard disk: NEVER

 

SLEEP – Sleep After: NEVER    

These next 2 settings may or may not appear, if they are there, change them as shown

           - Hibernate after: NEVER 

           - Allow hybrid sleep: OFF

 

PCI EXPRESS: Link State Power Management: OFF

 

PROCESSOR POWER MANAGEMENT:

 

MIN Processor State: 100

System cooling Policy: ACTIVE

MAX processor State: 100

 

Click Apply if you made any changes and then OK to close the box. Close the Edit Power Settings window.

 

NOTE: You have just disable Windows CPU CORE PARKING automatically. Please do not apply that tweak if you found it on the internet as it is not needed.

 

 

 

 

You are now in the combat zone, ready for action  ...and here we,      GO!  <--clicky

 

 

 

 

 

 

INITIAL MANUAL  BIOS SETUP FOR CLOCK TESTING - ALL SYSTEMS

 

Some of these settings I outlined in my last installments to this thread. I am repeating some of them here to make sure your BIOS is in fact setup correctly to begin.

 

 

Enter the Asus BIOS in ADVANCED mode and confirm or make the following changes:

 

===========================================================

 

MONITOR MENU -

 

Scroll down to the bottom of the list

 

ANTI SURGE SUPPORT: DISABLED

 

 

============================================================

 

 

ADVANCED MENU -

 

CPU CONFIGURATION:

 

HYPER THREADING: DISABLED

 

NOTE: I wish to make it very clear that Hyper threading will in no way improve Flight Sim performance. If you need Intel HT support for other applications and you also wish to clock at the same time then I FIRMLY suggest that users establish two clocking setups, one for HT DISABLED and one for HT ENABLED.

 

Start with HT DISABLED for max clocking ability and lower temps and once established then rework a new clock speed that passes stability and temp tests with HT enabled. Keep your final BIOS settings written down so you can make quick changes as required.

 

 

Scroll down to the bottom and enter the menu CPU POWER MANAGEMENT:

 

EIST: DISABLED

TURBO MODE: ENABLED

 

CPU C STATES: ENABLED   - This is set to AUTO by default, change it to ENABLED and under CPU C STATES the following will appear, change as outlined:

 

ENHANCED C1: DISABLED

CPU C3 REPORT: ENABLED

CPU C6 REPORT: DISABLED

C6 LATENCY: SHORT

CPU C7 REPORT: DISABLED

C7 LATENCY: LONG

PACKAGE C3 STATE REPORT: C3

 

 

This setup will assure the system will run the CPU at the maximum lowest stable voltage and current when the CPU is idle and under far less load. This will allow very high CPU speeds with stability and not run high voltage and current draw under less demanding conditions. Unlike processors of the past where C States are disabled for clocking, Haswell works very well with this setup.

 

Back out to the main menus

 

 

 

============================================================

 

 

ADVANCED MENU -

 

PCH CONFIGURATION:

 

PCI EXPRESS CONFIGURATION -

 

DMI LINK ASPM CONTROL: DISABLED

 
PEG ASPM: DISABLED
NOTE: YOUR BIOS MAY READ THIS WAY INSTEAD-
 

ASPM SUPPORT: DISABLED

 

 

PCIe SPEED: Leave this on BIOS default   AUTO is fine

 
 
 
 

-----------

 

ADVANCED MENU -

 

SYSTEM AGENT CONFIGURATION:

 

CPU AUDIO DEVICE: DISABLED

 

 

NB PCIE CONFIGURATION:

 

Verify PCIe slot the card is installed into (typically the first for BEST performance) displays: X16  and the LINK SPEED dropdown is set to GEN3 for the GTX 770/780

 

NOTE: All 600 and 700 series video cards are GEN3. Any video card lower than a 600 series is GEN2.

 

 

DMI LINK ASPM CONTROL: DISABLED

 
PEG ASPM: DISABLED
 
 
 
 
 

GRAPHICS CONFIGURATION:

 

PRIMARY DISPLAY: PCIE

IGPU MULTI-MONITOR: DISABLED

 

 

-------------

 

ADVANCED MENU -

 

MEMORY CONFIGURATION:

 

MEMORY REMAP: ENABLED

MEMORY SCRAMBLER: ENABLER  (may not appear in all systems)

 

 

============================================================

 

 

AI TWEAKER MENU -

 

AI OVERCLOCK TUNER: XMP  

 

-You should have already performed this change by now and also have tested your memory in Memtest! If you have for some reason reset your BIOS, make sure this is currently set to XMP and make sure that memory was properly tested before continuing.

 

If you have defective memory and you have already installed Windows and software then you may be looking at a corrupt Windows and software install. This is why I outlined memory testing is the FIRST item we check and verify before anything else with a new system.

 

 

CPU STRAP: 100

 

BCLK FREQUENCY: 100

 

Scroll Down the list....

 

Enter the DIGI+POWER CONTROL MENU

 

LOAD LINE CALIBRATION: 8   

 

CPU POWER PHASE CONTROL: OPTIMIZED

 

 

Exit the DIGI+POWER CONTROL MENU

 

 -------------------

 

Enter the CPU POWER MANAGEMENT MENU - Confirm or change:

 

ENHANCED INTEL SPEEDSTEP: DISABLED

TURBO MODE: ENABLED

 

 

Exit the CPU POWER MANAGEMENT MENU

 

 

Scroll down to the bottom of the list:

 

 
SVID SUPPORT: AUTO  All boards other than ROG
 
NOTE: This setting could also be listed as: SVID CONTROL instead of SUPPORT
 
 

SVID SUPPORT: ENABLED  ROG BOARDS ONLY

 
 
 
 
----------------------------------------------------------

NOTE for ROG users: The SVID setting will enable communication between the CPU and the voltage regulators and will allow HW Monitor to display the watt load readouts as loads are applied and removed. With ROG boards it has been noted that [AUTO] is DISABLED by default.

 

So far ROG boards are the only boards that require this change. All other Asus boards appear to run ENABLED by default (but without the additional override voltage setting) when SVID SUPPORT is set to AUTO. IF set to ENABLED on non-ROG boards the CPU INPUT VOLTAGE will not be available.

 

Professional overclockers who are trying for very high CPU speed to win a internet prize would typically disable SVID. We are not professional clockers looking to win a trophy. We want control and long term stability.

 

------------------------------------------------------------

 

 

 Scroll down to: CPU SPREAD SPECTRUM: DISABLED

 

 

===============================================================

 

 

The base Asus BIOS setup list for manually clocking all systems is now complete. Please hit F-10 save and exit and then go back into your BIOS and RECONFIRM every setting posted above. It has been noted that some boards may not hold a setting. We want to confirm the above list has indeed been saved and the system is booting this list.

 

Once confirmed, then continue...

 

 

 

 

 

 

-- THE INITIAL TEST CLOCK SETUP OUTLINE --

 

 

Follow these steps:

 

 

1. Locate the CPU speed range you intend to work with in the outlines posted below. You will want to start at a LOWER POINT and work your way UP.

 

Going too far, too fast can make this process more difficult to accomplish and far more frustrating. It can also present you with unsafe temps should the cooling solution not be able to handle the CPU voltage you must use. The first order of business is to find out where your CPU sits for voltage @ speed in a stable clock, and, at what MAX LOAD temperature .

 
 
 

I will post examples here:

 

- IF you have read what I have stated about cooling and Haswell and you have built with the goal of a CPU speed of 4.6-4.8GHz then START at 4.4 or 4.5 and work up. DO NOT slam a 4.6-4.8GHz clock! You must firmly understand where your system sits for voltage, cooling and stability before attempting any such clock with Haswell.

 

This is true for ANY Haswell clock.

 

 

- IF your cooling solution setup and goal was considered for a 4.4-4.5GHz clock, then START at 4.2-4.3 and work up.

 

 

- Clocks of 4.2-4.3 should start with the 4.0-4.1GHz outline and then try for 4.2GHz and then see if they can go higher.

 

 

YOU could go ahead a SLAM a clock if you want to, but I would highly advise against this! Being impatient is the #1 reason for failures and frustrations with system clocking and with establishing well defined system stability when clocking. It can introduce DAMAGE directly caused by a loose nut behind the keyboard. This nut is well known for being the #1 reason for failure so do keep that nut tight and well adjusted as you go along here.

 

 

 

2. Once you have established a boot into Windows without crashing using a selected outline and testing CPU voltages, proceed to Haswell and Overclocking 104 - Stability and Load Testing Software and follow the directions.  (if you have already reviewed section 104 you may move directly to section 105) presented below and then follow the directions for your CPU speed. Follow the outline process until you have established a safe temp / stable tested clock and now know the voltage range your CPU works with the initial established speed.

 

 

 

3. Once you have established a successful FULLY tested stable clock and confirmed it with a repeat test, if your temps are in line and appear to have room to increase as well as you have defined the system is not exceeding voltage limitations, then you may have room to work a higher clock.

 

If you wish to test higher clock speeds, then using the CPU and other voltages you have established for the lower speed clock, review the STARTING SETUP OUTLINE for the next higher clock speed range.

 

Be very aware that every CPU and system may require voltages that are different than those defined in each CPU speed outline I have defined. Your success at the lower level can define CPU voltage changes to what I outline for the next level up.

 

In other words, if the next level up in my CPU speed outline defines a lower CPU or other voltage than what you have established as 'fully stable' at a lower CPU speed, then LEAVE the current voltage settings as they are and work up from there based on the posted limitation chart I provided. You may need to make your own calls with that as I can only outline a setup so far.   

 

 

 

4. Once you have established your final tested safe and stable clock speed, the process of lowering voltages and retesting can begin.

 

Note that once you have established your high-end clock speed and if you fall within safe temp and voltage limits you can opt to skip the voltage reduction process. Some users prefer to try and trim back if they can, others may not.

 

Remember: the lower the voltage and temps, the longer the useful life if you tend to run your system daily under high loads, high temps and high voltages. Finding those bottom-end voltage values can take time and patience. It is not a requirement do this however the cooler things run on lower voltage with stability, the better.

 

 

 

DO NOTE: That as you progress here you will notice some of the BIOS settings applied will display YELLOW text as they are increased for the higher-end clock speeds and in some cases, RED. This is NORMAL as the motherboard manufacture will always place absolute conservative color coded alerts in a BIOS in which the settings may not be dangerous but do exceed what they must warn about 'legally' so they can not be held accountable for any clocking a user may attempt, low or high. In this outline I will not advise users to apply critically dangerous level voltages and will warn when the voltages I outline have the potential to sit at or offset automatically to the absolute limit. 

 

 

 

=========================================================

 

PATIENCE IS KEY! There is no such thing as 'FAST AND DIRTY' when it comes to proper clocking. If you get impatient and start skipping tests or go off and do it 'your way', unless you know exactly what you are doing, you will most likely suffer for that one way or another be it intermittent glitches, crashes and errors in the system or in FSX, or, lowered life expectancy of the processor.

 

This process can take a day or more to get nailed down and verified. Some users go further and run 2-3 days of load tests, others do a 1hr test and walk away in which they usually always suffer for that decision.

 

I see this all the time in the FSX forums..  CTDs, visual errors, performance problems with a user who followed a fast and dirty clocking outline and walked away from it with minimal testing time, or, no verification and just cant seem to figure out what is wrong.

 

Windows BSOD and lockups are not the only sign of a unstable clock! Please remember that.

 

==========================================================

 

 

BE SURE TO RECORD YOUR VOLTAGE SETTINGS AS YOU PROGRESS SO YOU KNOW WHERE YOU ARE FOR EACH STAGE IN THE CLOCKING PROCESS.

 

 

PLEASE PAY SPECIAL ATTENTION TO THE VOLTAGE VALUES AND THE DECIMAL POINTS IN THEM. BE VERY CAREFUL TO NOT MOVE A DECIMAL POINT WHEN ENTERING VOLTAGES INTO THE BIOS.

 

 

Remember, it is impossible for me to define ANY 'specific' voltage in this outline for your system and CPU. The suggested values for all voltages are a starting point only, calculated based on engineering documentation and tested here on my CPU and system. They could be too LOW or they could be a bit too HIGH on your system. This is what you will be testing for on your own. Start with the suggested values and move forward from there.  

 

 

After entering a new set of voltage values and on reboot of the system you will notice the voltage readings in the BIOS next to your new input values will always 'read' higher than the value you have entered. This is NORMAL and represents the first of a series of increases the board can apply to voltages with Haswell as loads are applied. ROG boards will tend to run voltages higher than other boards. The voltages I have outlined take into account these increases.

 

 

 

 

 

OK.. Its time to ROCK AROUND THE CLOCK <--clicky

 

 

=============================================================

 
 
 

 

 

 

 

STARTING SETUP OUTLINE: 4.0GHz-4.1GHz:

 

 

 

CPU CORE RATIO: PER CORE        make SURE this is set:  PER CORE

 

Set each of the next 4 boxes to: 40 for 4.0GHz or 41 for 4.1GHz

 

Highlight the setting with the mouse, enter the value in the box and hit ENTER on the keyboard after each numerical entry. BIOS settings will not accept unless you enter the value and hit ENTER

 

You should have each core (4 lines) set now to: 40    (or 41 if you prefer)

 

 

MIN CPU CACHE: 39

MAX CPU CACHE: 39  

Note: Sometimes this requires the MAX value be changed first then the MIN

 

We keep the CACHE speed lower for now and after we establish that the CPU speed has tested clean then move to increase the CPU cache speed to either match the CPU speed, or, close as we can obtain without going over the CPU speed.

 

You have now established a constant 4000MHz (or 4100MHz) CPU, 3900MHz NB Frequency

 

 

Scroll down and set: CPU CORE VOLTAGE: MANUAL MODE

 

Select the setting below and input the following value:

 

FOR 4.0GHz: CPU CORE VOLTAGE OVERRIDE:   1.12    <--- type this number and hit ENTER

 

FOR 4.1GHz: CPU CORE VOLTAGE OVERRIDE:   1.13    <--- type this number and hit ENTER

 

NOTE: My starting voltage value range for a 4.0-4.1 clock is:  1.10 - 1.15. I am splitting the difference here to start out at 1.12 - 1.13

 

 

Be VERY careful with these next values and NOTE THE DECIMAL POINTS!

 

 

CPU CACHE VOLTAGE: MANUAL MODE

 

CPU CACHE VOLTAGE OVERRIDE:   1.22

 

CPU SYSTEM AGENT OFFSET MODE SIGN: +

 

CPU SYSTEM AGENT VOLTAGE OFFSET:   .150

 

CPU ANALOG I/O VOLTAGE OFFSET SIGN: +

 

CPU ANALOG I/O VOLTAGE OFFSET:   .050

 

CPU DIGITAL I/O VOLTAGE OFFSET SIGN: +

 

CPU DIGITAL I/O VOLTAGE OFFSET:   .100     

 

CPU INPUT VOLTAGE:   1.82

 

 

---------------------------------------------------

NOTE: ROG BOARDS THAT DEFINE 2 CPU INPUT VOLTAGES:

 

INITIAL CPU INPUT VOLTAGE:   1.82

 

EVENTUAL CPU INPUT VOLTAGE:   1.82

---------------------------------------------------

 

 

Hit F-10 SAVE AND EXIT.  REBOOT back into the BIOS, enter the AI Tweaker menu, scroll down and CONFIRM every setting listed above starting with CPU CORE RATIO: PER CORE  and ending with: CPU INPUT VOLTAGE:  1.82

 

    

Check you input values directly. If there is any setting that is not right or has changed, reset it as defined above in the outline for your clock speed then hit F-10 save and exit, reboot and verify in the AI Tweaker list in the BIOS again. Once all settings are confirmed, then begin the testing process by booting into Windows. One of two things will happen:

 

a. Your system will boot into Windows just fine and remain stable while you load up the test software. If this is the case please proceed to the next section: Overclocking 104 - Stability and Load Testing and follow the directions. 

 

 

b. Your system will lockup or crash booting into Windows. If this happens, (hard reset if required) boot into the BIOS and raise the CPU core voltage .01volts.

 

Example: If you set CPU CORE VOLTAGE OVERRIDE: 1.12, then raise that to 1.13, F-10 save and exit and try booting into Windows again. Continue raising the voltage by .01v until you have a successful boot into Windows and appear stable. Once you have obtained that goal please proceed to the next section: Haswell and Overclocking 104 - Stability and Load Testing Software and follow the directions.  (if you have already reviewed section 104 you may move directly to section 105)

 

 

=============================================================

 

 

 

 

 

 

 

STARTING SETUP OUTLINE: 4.2GHz-4.3GHz:

 

DO NOT attempt this clock using the Intel factory heatsink. This clock requires a far better cooling solution than Intel provides and even some of the more reserved replacement cooling solutions that are better than the Intel heatsink may struggle.

 

 

 

CPU CORE RATIO: PER CORE        make SURE this is set:  PER CORE

 

Set each of the next 4 boxes to: 42 for 4.2GHz or 43 for 4.3GHz

 

Highlight the setting with the mouse, enter the value in the box and hit ENTER on the keyboard after each numerical entry. BIOS settings will not accept unless you enter the value and hit ENTER

 

You should have each core (4 lines) set now to: 42    (or 43 if you prefer)

 

MIN CPU CACHE: 39

MAX CPU CACHE: 39  

Note: Sometimes this setting requires the MAX value be changed first then the MIN

 

We keep the CACHE speed lower for now and after we establish the CPU speed has tested clean then move to increase the CPU cache speed to either match the CPU speed, or, close as we can obtain without going over the CPU speed.

 

You have now established a constant 4200MHz (or 4300MHz) CPU, 3900MHz NB Frequency

 

 

Scroll down and set: CPU CORE VOLTAGE: MANUAL MODE

 

Select the setting below and input the following value:

 

FOR 4.2GHz: CPU CORE VOLTAGE OVERRIDE:   1.16    <--- type this number and hit ENTER

 

FOR 4.3GHz: CPU CORE VOLTAGE OVERRIDE:   1.18    <--- type this number and hit ENTER

 

NOTE: My starting voltage value range for a 4.2-4.3 clock is:  1.14 - 1.20. I am splitting the difference here to start out at 1.16 - 1.18

 

 

Be VERY careful with these next values and NOTE THE DECIMAL POINTS!

 

 

CPU CACHE VOLTAGE: MANUAL MODE

 

CPU CACHE VOLTAGE OVERRIDE:   1.25

 

CPU SYSTEM AGENT OFFSET MODE SIGN: +

 

CPU SYSTEM AGENT VOLTAGE OFFSET:   .200

 

CPU ANALOG I/O VOLTAGE OFFSET SIGN: +

 

CPU ANALOG I/O VOLTAGE OFFSET:   .050

 

CPU DIGITAL I/O VOLTAGE OFFSET SIGN: +

 

CPU DIGITAL I/O VOLTAGE OFFSET:   .100     

 

CPU INPUT VOLTAGE:   1.85

 

 

---------------------------------------------------

NOTE: ROG BOARDS THAT DEFINE 2 CPU INPUT VOLTAGES:

 

INITIAL CPU INPUT VOLTAGE:   1.82

 

EVENTUAL CPU INPUT VOLTAGE:   1.85

---------------------------------------------------

 

 

Hit F-10 SAVE AND EXIT.  REBOOT back into the BIOS, enter the AI Tweaker menu, scroll down and CONFIRM every setting listed above starting with CPU CORE RATIO: PER CORE  and ending with: CPU INPUT VOLTAGE:  1.85

 

 

If there is any setting that is not right or has changed, reset it as defined above, F-10 save and exit, reboot and verify in the AI Tweaker list in the BIOS again. Once all settings are confirmed, then begin the testing process by booting into Windows. One of two things will happen:

 

a. Your system will boot into Windows just fine and remain stable while you load up the test software. If this is the case please proceed to the next section: Haswell and Overclocking 104 - Stability and Load Testing Software and follow the directions.  (if you have already reviewed section 104 you may move directly to section 105)

 

 

b. Your system will lockup or crash booting into Windows. If this happens, (hard reset if required) boot into the BIOS and raise the CPU core voltage .01volts.

 

Example: If you set CPU CORE VOLTAGE OVERRIDE: 1.16, then raise that to 1.17, F-10 save and exit and try booting into Windows again. Continue raising the voltage by .01v until you have a successful boot into Windows and appear stable. Once you have obtained that goal please proceed to the next section: Haswell and Overclocking 104 - Stability and Load Testing Software and follow the directions.  (if you have already reviewed section 104 you may move directly to section 105)

 

 

 

=============================================================

 
 
 
 

STARTING SETUP OUTLINE: 4.4GHz-4.5GHz:

 

THE HEAT IS ON: THIS IS WHERE YOUR TEMPS CAN BECOME A SERIOUS ISSUE REGARDLESS OF THE COOLING SOLUTION

 

If you are running a air super cooler or less than a Corsair H100 liquid system be very aware that your max CPU voltage will most likely be 1.23-1.26 or LESS and that assumes the ambient temp of the room is not warm and the CPU can run stable at those voltages.

 

This is the point where a CPU modification to the die cap will allow higher clocking at far cooler temperatures. If you can achieve the 4.4-4.5 clock fully stable and tested, you can then evaluate your CPU and if you wish to take the extended steps to modify it. 

 

 

 

CPU CORE RATIO: PER CORE        make SURE this is set:  PER CORE

 

Set each of the next 4 boxes to: 44 for 4.4GHz or 45 for 4.5GHz

 

Hit the ENTER button on the keyboard after each numerical entry. BIOS settings will not accept unless you enter the value and hit ENTER

 

You should have each core (4 lines) set now to: 44    (or 45 if you prefer)

 

 

MIN CPU CACHE: 39

MAX CPU CACHE: 39  

Note: Sometimes this requires the MAX value be changed first then the MIN

 

We keep the CACHE speed lower for now and after we establish that the CPU speed has tested clean then move to increase the CPU cache speed to either match the CPU speed, or, close as we can obtain without going over the CPU speed.

 

 

You have now established a constant 4400MHz (or 4500MHz) CPU, 3900MHz NB Frequency

 

 

Scroll down and set: CPU CORE VOLTAGE: MANUAL MODE

 

Select the setting below and input the following value:

 

FOR 4.4GHz: CPU CORE VOLTAGE OVERRIDE:   1.22    <--- type this number and hit ENTER

 

FOR 4.5GHz: CPU CORE VOLTAGE OVERRIDE:   1.26    <--- type this number and hit ENTER

 

NOTE: My starting voltage value range for a 4.4-4.5 clock is:  1.19 - 1.29. I am splitting the difference here to start out at 1.22 - 1.26  

 

IN THIS OUTLINE if you would like to try a lower CPU core voltage in starting out to discover where your processor sits, that is fine! An alternate would be:

 

FOR 4.4GHz: CPU CORE VOLTAGE OVERRIDE:   1.20    <--- type this number and hit ENTER

 

FOR 4.5GHz: CPU CORE VOLTAGE OVERRIDE:   1.24    <--- type this number and hit ENTER

 

DO be aware that the lower voltage starting point may take more trial and error starting out but at the same time may also net a lower successful clock test speed with 3900MHz cache and therefore lower temps with the initial setup so you can evaluate your system for cooling.

 

I will say this: moving to the CACHE phase in Overclocking 105 will require a higher CPU core voltage override, there is absolutely no question about that! How much? ...you will find out as you progress. I can not estimate that here.

 

 

Be VERY careful with these next values and NOTE THE DECIMAL POINTS!

 

 

CPU CACHE VOLTAGE: MANUAL MODE

 

CPU CACHE VOLTAGE OVERRIDE:   1.28

 

CPU SYSTEM AGENT OFFSET MODE SIGN: +

 

CPU SYSTEM AGENT VOLTAGE OFFSET:   .280

 

CPU ANALOG I/O VOLTAGE OFFSET SIGN: +

 

CPU ANALOG I/O VOLTAGE OFFSET:   .050

 

CPU DIGITAL I/O VOLTAGE OFFSET SIGN: +

 

CPU DIGITAL I/O VOLTAGE OFFSET:   .100     

 

CPU INPUT VOLTAGE:   1.88

 

 

---------------------------------------------------

NOTE: ROG BOARDS THAT DEFINE 2 CPU INPUT VOLTAGES:

 

INITIAL CPU INPUT VOLTAGE:   1.82

 

EVENTUAL CPU INPUT VOLTAGE:   1.88

---------------------------------------------------

 

 

Hit F-10 SAVE AND EXIT.  REBOOT back into the BIOS, enter the AI Tweaker menu, scroll down and CONFIRM every setting listed above starting with CPU CORE RATIO: PER CORE  and ending with: CPU INPUT VOLTAGE:  1.88

 

    

Check you input values directly. If there is any setting that is not right or has changed, reset it as defined above in the outline for your clock speed then hit F-10 save and exit, reboot and verify in the AI Tweaker list in the BIOS again. Once all settings are confirmed, then begin the testing process by booting into Windows. One of two things will happen:

 

a. Your system will boot into Windows just fine and remain stable while you load up the test software. If this is the case please proceed to the next section: Haswell and Overclocking 104 - Stability and Load Testing Software and follow the directions.  (if you have already reviewed section 104 you may move directly to section 105)

 

 

b. Your system will lockup or crash booting into Windows. If this happens, (hard reset if required) boot into the BIOS and raise the CPU core voltage .01volts.

 

Example: If you set CPU CORE VOLTAGE OVERRIDE: 1.24, then raise that to 1.25, F-10 save and exit and try booting into Windows again. Continue raising the voltage by .01v until you have a successful boot into Windows and appear stable. Once you have obtained that goal please proceed to the next section: Haswell and Overclocking 104 - Stability and Load Testing Software and follow the directions.  (if you have already reviewed section 104 you may move directly to section 105)

 

 

 

=============================================================

 
 
 
 
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 13503
Direct Link To This Post Posted: August-29-2013 at 9:40pm

 

 
 
 
 
 
STARTING SETUP OUTLINE: 4.6GHz-4.8GHz:

 

This clock requires high-end liquid cooling only without a modification to the CPU IHS contact point. Even high-end liquid can FAIL this clock without the CPU IHS mod. Be aware that exceptionally dangerous temperatures are possible with this clock outline as well as the potential to push maximum voltage specifications.

 

If you have read through or worked with the outlines as presented thus far you may have noticed that with Haswell as we increase CPU speed as well as cache speed with it, the larger the gap between one level of clocking to the next in voltage range requirement enters the equation.

 

In this next outline and assuming the CPU is not a 'dud' for clocking, you are going to be faced with what are known as the 'walls' in clocking where one user may find their CPU must go from 1.28v to 1.42v in order to obtain a single step up in CPU speed with the cache speed following reasonably behind.

 

As with all other outlines we start with the cache speed set to a fixed lower speed and then proceed to increase it once our CPU speed is tested stable and within temp specs.

 

Rest assured, this clock is not easy to obtain and the CPU mod I will outline later opens the door to this level of clocking far better than working with the current Intel design Haswell chips that all appear to have poor connections between the CPU die and the inside of the cap that covers it.

 

Do not be disappointed if your system can not achieve this clock. A 4.5GHz clock with a CPU cache speed of 4300-4400 or 4500 is an exceptionally excellent clock with Haswell and remember at a Haswell CPU speed of 4500 the system is easily running in a comparable 5GHz range with SB processors.

 

 

 

HOW TO FIREWALL HASWELL

 

-or-

 

BRING UP THE HOLY HAND GRENADE! <--clicky

 

 

I would suggest starting at 4600, then progress to 4700 and then finally, try for 4800. Any one of these clocks will be a superior clock setup.

 

 
 
 

CPU CORE RATIO: PER CORE        make SURE this is set:  PER CORE

 

Set each of the next 4 boxes to: 46 for 4.6GHz, 47 for 4.7GHz or 48 for 4.8GHz

 

Highlight the setting with the mouse, enter the value in the box and hit ENTER on the keyboard after each numerical entry. BIOS settings will not accept unless you enter the value and hit ENTER

 

You should have each core (4 lines) set now to: 46    (or 47 - 48 if you prefer)

 

We are going to run a bit higher cache with this clock than others to start off...

 

MIN CPU CACHE: 41

MAX CPU CACHE: 41 

Note: Sometimes this requires the MAX value be changed first then the MIN

 

We keep the CACHE speed lower for now and after we establish that the CPU speed has tested clean then move to increase the CPU cache speed.

 

NOTE: it is nearly impossible to run a CPU cache speed above 4500 with a CPU speed of 4600 and greater without exceeding SAFE voltage LIMITS. Consider yourself LUCKY if you can run a cache speed of 4400 with a CPU speed of 46-4800.

 

You have now established a constant 4600MHz (or 4700-4800MHz) CPU, 4100MHz NB Frequency

 

 

Scroll down and set: CPU CORE VOLTAGE: MANUAL MODE

 

Select the setting below and input the following value:

 

FOR 4.6GHz: CPU CORE VOLTAGE OVERRIDE:   1.34    <--- type this number and hit ENTER

 

FOR 4.7GHz: CPU CORE VOLTAGE OVERRIDE:   1.37    <--- type this number and hit ENTER

 

FOR 4.8GHz: CPU CORE VOLTAGE OVERRIDE:   1.39    <--- type this number and hit ENTER

 

 

NOTE: My starting voltage value range for a 4.6-4.7-4.8 clock is:  1.30 - 1.43. I am splitting the difference here to start out. 1.34 - 1.37 - 1.39

 

 

Be VERY careful with these next values and NOTE THE DECIMAL POINTS!

 

 

CPU CACHE VOLTAGE: MANUAL MODE

 

CPU CACHE VOLTAGE OVERRIDE:   1.30

 

CPU SYSTEM AGENT OFFSET MODE SIGN: +

 

CPU SYSTEM AGENT VOLTAGE OFFSET:   .300

 

CPU ANALOG I/O VOLTAGE OFFSET SIGN: +

 

CPU ANALOG I/O VOLTAGE OFFSET:   .050

 

CPU DIGITAL I/O VOLTAGE OFFSET SIGN: +

 

CPU DIGITAL I/O VOLTAGE OFFSET:   .100     

 

CPU INPUT VOLTAGE:   1.88

 

 

---------------------------------------------------

NOTE: ROG BOARDS THAT DEFINE 2 CPU INPUT VOLTAGES:

 

INITIAL CPU INPUT VOLTAGE:   1.88

 

EVENTUAL CPU INPUT VOLTAGE:   1.88

---------------------------------------------------

 

 

Hit F-10 SAVE AND EXIT.  REBOOT back into the BIOS, enter the AI Tweaker menu, scroll down and CONFIRM every setting listed above starting with CPU CORE RATIO: PER CORE  and ending with: CPU INPUT VOLTAGE:  1.88

 

    

Check you input values directly. If there is any setting that is not right or has changed, reset it as defined above in the outline for your clock speed then hit F-10 save and exit, reboot and verify in the AI Tweaker list in the BIOS again. Once all settings are confirmed, then begin the testing process by booting into Windows. One of two things will happen:

 

a. Your system will boot into Windows just fine and remain stable while you load up the test software. If this is the case please proceed to the next section: Haswell and Overclocking 104 - Stability and Load Testing Software and follow the directions.  (if you have already reviewed section 104 you may move directly to section 105)

 

 

b. Your system will lockup or crash booting into Windows. If this happens, (hard reset if required) boot into the BIOS and raise the CPU core voltage .01volts.

 

Example: If you set CPU CORE VOLTAGE OVERRIDE: 1.34, then raise that to 1.35, F-10 save and exit and try booting into Windows again. Continue raising the voltage by .01v until you have a successful boot into Windows and appear stable. Once you have obtained that goal please proceed to the next section: Haswell and Overclocking 104 - Stability and Load Testing Software and follow the directions.  (if you have already reviewed section 104 you may move directly to section 105)

 

 
 
 
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 13503
Direct Link To This Post Posted: August-29-2013 at 9:42pm
 
 
 
 
THIS POST IS NOW COMPLETE....

 

 

 
 
 
 
 

 

 

 

Haswell and Overclocking 104 -

 

Stability and Load Testing Software

 

 

I decided to change this outline somewhat. I am breaking this next part down into two sections, one that deals with the tools needed to check and verify your clock and the other to outline the process.

 

You seasoned clockers may find this section boring..  OR you may discover something you missed in the past. I encourage everyone to go over this section carefully so you understand what it is you will be doing and how.

 

I will be addressing this subject as I have all others in this thread, with the assumption the user reading the information does not have a great deal of experience with clocking. You must understand what it is you are doing with setup, monitoring and VERIFYING with the tools before you do it and this section will provide that information.

 

Once you have gone through this section and have the software required to test your clock, set it up correctly and understand how to use it, you can then skip this section and go to: Haswell and Overclocking 105 - Clocking It Up - Stability and Load Testing

 

Use this section as a REFERENCE during your clock testing should you forget what the testing applications do and how to set them up or read them.

 

 

 

=========================================================

 

 

 

 

In this next section you will need the testing tools that were outlined earlier in this thread.

 

AIDA64 - http://www.aida64.com/downloads A few notes here about AIDA64: Purchase of this software IS NOT a requirement. If you have been using the software and the trial time period has run out and you can not use the product at this point nor wish to purchase the software, you can opt to skip the use of AIDA64 completely and simply use OCCT v4.4.0 (or above if its updated at a later date). AIDA64 is a good software application to have in ones arsenal for testing as it offers far more than load tests however I do understand economics and the need to keep costs down so if you can not use this product you can replace its use with OCCT. 

                                                             

Primary use in clock testing: We use the FPU stress test from AIDA64 to isolate the CPU cores from all other systems (memory, memory controller, CPU cache) in order to check CPU temperature and how well the system will handle the clock we have established. The OCCT: Linpack test outlined next will provide a similar heat check test but will add in additional stability checks with respect to memory. AIDA64 will allow a user to disable all but the FPU stress test for a fast way to get a look at the temps we are faced with in clocking the system.

 

Further Use: The cache and memory benchmark test provided with AIDA64 is one of my primary tools to define 'true tuned' system performance as well as memory tweak test performance changes. The CPU cache and memory latency test is actually more important to me than the load tests the software provides.

 

Again, there is no obligation to purchase anything. You can use OCCT for load testing however there are no performance tests with OCCT.

 

The latest BETA version of AIDA64 FPU test will actually run the CPU somewhat cooler than the June 3rd 2013 official release version (at the date of this posting). There have been changes to AIDA64 through their beta's in order to update for the new motherboard sensors as well as address issues with older motherboards. You can find a complete list of all release and beta versions posted here: http://www.aida64.com/downloads

 

Each Beta comes with release notes you can review. The beta's are cumulative so each change from earlier beta releases are included in the latest beta release.

 

I run the latest beta version and will update to the latest release version once it is posted at the AIDA64 website.

 

 

 

LAST: Be aware that unlike OCCT, AIDA64 stress testing does NOT come with 'automated shutdown' features if temperature limits are exceeded. If this is a concern, then use OCCT instead of AIDA64 for max temperature testing. AIDA64 is easy to shut down if you see temps exceed limits with a single click but unlike OCCT it will not shut down automatically.

 

We will be using OCCT regardless of if AIDA64 is used for the initial temperature and stability test of the CPU, or not.

 

 

 

 

 

OCCT 4.4.0 - http://www.ocbase.com/index.php/download  This application is freeware and it will be our PRIMARY testing software. There are two tests we will be using.

 

 

1. The OCCT:LINPACK 64bit AVX enabled test - This test is to define max temp with the maximum amount of memory in use. This test will run the CPU hotter than the latest beta's of AIDA64. We use this test to define HEAT more than stability. If you working with changes to voltages other than Vcore (CPU Voltage) you do not need to run this test with every BIOS change. This test is designed to define if you cooling solution can handle the CPU clock speed and CPU voltage you have set in the BIOS.

 

If for any reason your Haswell happens to be hit with AVX code from Windows software applications (rare but could happen) then you must KNOW where your system will run that CPU for both voltage and temperature.

 

So you clearly understand the situation with Haswell: Unlike processors of the past,  anyone who sets up a clock and does not run this test thinking they will never see any issues by using a lower load test to define stability and get away with lower temps in that test has essentially loaded a gun, pulled the hammer back and is hoping no one comes along and bumps the trigger.

 

We use this test because it stresses more than the FPU and far higher than AIDA64 with multiple tests running at once from the AIDA64 stress test application.

 

This test will allow you to see just how efficient your cooling solution really is and will instantly show you how much headroom you have to work in a clock with temps and CPU voltage.

 

 

 

 

2. The OCCT:CPU 64bit test - This test runs the processor far cooler than the OCCT:Linpack test outlined above, however, this test will place a FAR HIGHER load on primary CPU functions without AVX instruction which simulates how we will run our systems in real world with FSX. This test is the PRIMARY stability test for overclocking.

 

PLEASE UNDERSTAND THIS CLEARLY: You can pass the OCCT : Linpack test with flying colors @ 80-88c and FAIL the standard CPU test running far lower temperatures, miserably.

 

 

 

 

 

 

-- MONITORING SOFTWARE REQUIRED --

 

 
 
 
 

Hardware Monitor (HW Monitor) http://www.cpuid.com/softwares/hwmonitor.html 

 

Please note: For those of you who do not understand 'advertisements' on the internet, you will find the link to the latest version on the SIDE bar of that web page under; Download the latest release -  v1.23 (installation, automatically installs 32 or 64-bit version depending on your system) and then click the blue DOWNLOAD NOW button on center of the next page that appears.

 

A direct link to the 1.23 version is here: ftp://ftp.cpuid.com/hwmonitor/hwmonitor_1.23-setup.exe but do keep in mind that the software is updated from time to time so its always best to check the main page for the latest release version.

 

We will be using this software to monitor the critical TEMPS and the VOLTAGES during the overclock tests.

 

 

 

CPUz - http://www.cpuid.com/softwares/cpu-z/versions-history.html This software is used to verify CPU speed, PCIe link width, northbridge (or UCLK) speed, as well as memory speed and memory timing. It can be used to check the installed memory model, SPD as well as the model of motherboard and BIOS revision installed. It should NOT be used to define the CPU voltage under any load test!

 

As pointed out with HW Monitor, the latest product exe installer release link will be located on the side and then when clicked the DOWNLOAD NOW button in the center of the page will start the download.

 

A direct link to the current (as of this date) CPUz 1.66 version is here: ftp://ftp.cpuid.com/cpu-z/cpu-z_1.66-setup-en.exe

 

 

 

Please install all of the software outlined above and we can begin the process of load testing your clock. Each installer will place shortcuts in your START menu programs list and may offer to or automatically place a shortcut on the desktop during the installation. During a clock testing phase I will typically place shortcuts to each of these programs on the desktop for fast access since there can and will be repeated rebooting and crashing between test runs. Once I have established and completed my clock testing I will delete the shortcuts from the desktop.

 

 

 

 

 

GETTING TO KNOW YOUR TESTING AND MONITOR APPLICATIONS

 

 

There is not a lot of 'detail' to go through here. Most of this is strait forward as I am not going to get into every single detail of each application and only deal with the areas that are critical to understand. The only software that requires 'setting up' or changes for testing is AIDA64 and OCCT. Past that the other applications are simply RUN and GO once you understand what it is you are looking at and monitoring/checking.

 

 
 

 

CPUz:

 
Run this first after BIOS changes to verify the changes you made.

 

This application will display information that you should VERIFY each time you make a change to the following in the BIOS:

 

CPU TAB:

CPU Core Speed

CPU Multiplier

BCLK (Bus Speed)

 

MAINBOARD TAB:

Verify PCIe Link Width

Verify BIOS version

 

MEMORY TAB:

CPU Cache Multiplier

DRAM Frequency

DRAM Timing

 

 

When you run CPUz the first time your goal is to verify the items I show below from the CPU tab, MAINBOARD tab and the MEMORY tab.

 

 

CPU Tab

 

 
This example shows I am running a CPU core speed of 4746.57 (or thereabouts), a CPU Multiplier of 47 and a BCLK (Bus Speed) of 101 (displaying 100.99)

 

101 BCLK x 47 CPU Multiplier = 4747MHz

 

This checks as my current clock is set as defined in the BIOS. You will verify your settings here as well

 

There can be slight variations in BCLK which is why the math/settings and displayed actual can be slightly different. When loads are applied these values will max out and maintain.

 

 

 

Mainboard Tab

 

 

 

This example shows I am running BIOS 1405 (the latest for the Z87 Sabertooth) and most important, my PCIe Link Width displays x16 which is CRITICAL. If your system is not running PCIe x16 and instead displays x8 or x4 there is a issue with the PCIe slot the video card is running. It could be a conflict with another card and this MUST be diagnosed and fixed now before continuing. Consult the motherboard manual about the slot you are using for the video card and the ports or resources it shares with.

 

You must be running x16 for single video card performance. Some motherboards will cut the PCIe Link Width back to x8 if more than one card is in the system. This is where the selection of motherboard when using more than one video card, or, a need to use PCIe x16 slots for addon cards can become an issue.

 

 

 

 

Memory Tab

 

 

 

CPU CACHE:

This example shows my CPU Cache (or Northbridge frequency) is 4443.6    That checks since I am using a BCLK of 101 ( displayed as 100.99) and a CPU Cache multiplier of 44 for my current clock.

 

101 BCLK x 44 CPU Cache Multiplier = 4444MHz CPU Cache

 

 

DRAM FREQUENCY:

DRAM Frequency will always be displayed divided by 2 therefore DDR3 2400 would display as DRAM FREQUENCY 1200

 

If you use DDR3 2133 memory you should see a DRAM FREQUENCY of: 2133 / 2 or 1066.5

 

If you use DDR3 2400 memory you should see a DRAM FREQUENCY of: 2400 / 2 or 1200.0

 

If you use DDR3 2666 memory you should see a DRAM FREQUENCY of: 2666 / 2 or 1333.0

 

If you use DDR3 2800 memory you should see a DRAM FREQUENCY of: 2800 / 2 or 1400.0

 

The same holds true for any memory. Simply divide the DDR3 speed by 2. This assumes you are using a BCLK of 100 as all clocks at this point should be using. I will touch on the use of raising BCLK later in the tips section after a clock has been established.

 

 

As you can see above I show 1212.00 which is correct since I happen to be running a BCLK of 101 this boosts my DRAM Frequency slightly. In my examples shown above:

 

101 x 12 DRAM Multiplier (selected by the BIOS as DRAM SPEED 2424 as close to 2400 as I could set) = 1212 x 2 or 2424MHz DRAM FREQUENCY. The value can be slightly off a whole number multiplier.

 

Users right now should remain focused on BCLK 100.

 

 

 

TIMINGS:

Memory Timing shown matches the manufacture ratings for my memory as set by XMP in the BIOS, however my memory is rated for a "Command Rate" of 2T and I have since tested the Command Rate of 1T and found it stable. This is something that comes LATER, not now as right now we want the memory running the manufacture spec for checking and tuning our CPU clock, later we will look at memory tweaking.

 

 

Your job is to VERIFY your CPU speed, CPU cache speed, DRAM Frequency and DRAM Timings match the correct values. Once done you can close CPUz but you will reopen this application after making changes in the BIOS during clocking to make sure everything checks at each clock setting change.

 

 

 

 

 

 

HW MONITOR:

 

 

This will be our PRIMARY voltage and temperature checking software. After running CPUz to verify settings in the BIOS and check frequencies and closing it out, we launch HW Monitor and expand the box and the columns so we can see the readouts from the VOLTAGES section down to the POWERS section as shown below.

 

 

The critical sections we will be focused on are outlined in red

 

 

 

 

 

Do note that this software can also be used to look at other voltages and temps for power supply output as well as motherboard sensors for voltages and fan speeds. It can also be used to check HDD and video adapter sensors.

 

The ones I have outlined above are the items we will focus on during every phase of clock testing to assure under a LOAD the defined voltage and temperature values I will outline are NOT exceeded.

 

This is critical to assure safe long term clock operation. I will explain what to check and show images of those checks in the next section: 105 - Clocking It Up - Stability and Load Testing

 

This application should remain on your desktop at all times and in a location where you can review the readouts in real time during any phase of load testing.

 

 

 

 

 

 

 

AIDA64:

 

 

If you use this software for checking initial CPU temperature and stability, set it up as follows:

 

1. Launch AIDA64. If a beta update is available it will usually inform you of that and link you to it. Given how new Haswell is and the newer motherboards you probably should consider installing the latest beta to ensure your version of AIDA64 is reading the sensors properly.

 

 

2. In the TOOLS MENU, select the SYSTEM STABILITY TEST when the box appears uncheck all the tests except FPU as shown below.

 

 

 

 

 

Once done you can close the box and then close AIDA64, you are ready to use this software. CLOSE the software so the settings are saved. If you run this software and you crash the setting can be lost. Close it now before using it the first time.

 

 

 

 

 

 

 

 

OCCT 4.4.0 (or above):

 

 

 

This will be our primary load testing software. There are 2 sections of this software we will be using, the OCCT: Linpack test for max temp checking, and, the OCCT: CPU Test for full-on stability.

 

Before using OCCT we want to ensure it is set up correctly to automatically SHUT DOWN the load tests should temperatures exceed 92c. To do so follow these directions:

 

 

1. Launch OCCT. When the software launches you will see 2 boxes, a 'donation' box and a 'commercial use' box. Close these boxes and the main GUI will appear.

 

 

2. In the GUI under the tab: OCCT: CPU TAB I have outlined the settings with yellow boxes below. Set up OCCT: CPU as I show here making the changes outlined in the yellow boxes:

 

 

 

 

 

3. Click the OCCT:LINPACK TAB. I have outlined the settings with yellow boxes below. Set up OCCT: LINPACK as I show here making the changes outlined in the yellow boxes:

 

 

 

 

NOTE: Your MEMORY AMOUNT in MB will not be the same as shown above. The 90% setting will automatically select the correct memory amount based on the physical memory you have installed in your system.

 

 
 

4. Click the orange 'gear' icon as shown below. The settings box will appear. In the list that is displayed change the CPU TEMP LIMITS from 85 to 92 as shown below

 

 

 

 
 

 

Make sure there is a check in each box for each core under the: STOP TESTING IF VALUE IS   and    SHOW IN REAL TIME. Once complete close the settings box and then CLOSE OCCT.

 

Once done and you have closed OCCT, you are ready to use this software. CLOSE the software so the settings are saved. If you run this software and you crash the setting can be lost. Close it now before using it the first time.

 

 

 

 

 
 
 

 

-- OPTIONAL --

 

RealTemp Inferno Edition:

 

In the past we have used the software RealTemp (current version is 3.70) to define CPU core temps. This software as of the date of posting has not been completely updated for Haswell. However, the TECH INFERNO version of RealTemp http://forum.techinferno.com/downloads.php?do=file&id=43 has been updated for Ivy Bridge and will work for Haswell. I personally use this software for both load temp verification AND constant real time tray monitoring of Haswell core temps with Windows boot.

 

 

 

1. Unzip this to any location. I personally take the unzipped folder and drop it into C:\Program Files (x86) but it can live anywhere on your system

 

 

2. Inside the unzipped folder right click the file: RealTemp.exe and select 'SEND TO DESKTOP which creates a shortcut on the desktop for easy access.

 

 

3. If you intend to use this software to monitor Haswell in real tile with Windows and in the Windows tray then COPY the shortcut on the desktop to the: START: ALL PROGRAMS > STARTUP folder of the Windows menu. This will launch RealTemp TI when Windows boots.

 

 

4. SETUP RealTemp TI: Launch the application. Click the OPTIONS button. The options will appear, set it up as shown below in the blue boxes and follow the directions.

 

 

 

NOTE: For the color of the font I use yellow as it is easy to read. You can use what ever color you like. I prefer yellow with the font type and size noted in the image above.

 

 

Once done click APPLY and close the settings box. Then close RealTemp TI to save the settings.

 

5. Reopen RealTemp TI. The readout for the MAX temp will now display in the Windows tray but first you need to set up the tray so it will display all the time. Down by the clock in the Windows task bar click the UP arrow button. Click the CUSTOMIZE link. When the box appears locate: RealTemp and then set the dropdown to: SHOW ICON AND NOTIFICATIONS.

 

The real time MAX temp (or hottest core) of your CPU will now always appear in the taskbar and easy to read. RealTemp is also set up so if you hit 91-92c a RED ALERT klaxon will sound and the reading will flash in the tray.

 

 

 
 
 
UNDERSTANDING THE REAL TEMP DISPLAY:
 
This is fairly strait forward. I have outlined what each section of RealTemp TI displays and can be monitored in real time in the image below:

 

 

 
 
 
 
 
 

Your testing tools are now setup and ready. You can refer back to the information presented above at any time should you need reminders or need to reinstall the outlined software.

 

 

Once you have completed this section you can move forward to the next section: Haswell and Overclocking 105 and start the testing process for your clock.

 
 
 
 
 
 
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 13503
Direct Link To This Post Posted: September-14-2013 at 1:02pm
 
 
 
 
THIS POST IS NOW COMPLETE....

 

 

 

 

 

 

 

 

Haswell and Overclocking 105 -

 

Clocking It Up - Stability and Load Testing
 
 
 
 
OK so you have made it this far, congratulations!.. and you are now ready to begin the process of testing your clock.

 

The first goal when overclocking Haswell is to define where your CPU is sable and at what temperature. Unlike processors of the past, there is a very wide range of CPU speed/voltage/temp possibilities with Haswell which are based on the quality of the thermal contact between the CPU die and the inside of the processor cap (or lid/cover) and the manufacture run quality. Between these two factors it is quite impossible to define where any single Haswell processor will run with respect to speed/voltage and temperature. Factors that will define your clock with Haswell are:

 

a. The cooling solution and how well it can transfer heat.

 

 

b. The airflow for volume of air in and out of the tower.

 

 

c. The ambient room temp. The higher the ambient, the hotter the CPU and system.

 

 

d. The attention you placed into building the system with respect to the inside of the tower and the airflow it provides as well as the attention placed into installing the CPU cooler with the thermal compound (too much thermal compound is not good)

 

If you use the thermal compound product I suggest, Coollaboratory Liquid Pro, it is exceptionally difficult to use too much. Unlike other thermal compounds that are thick, CE-Pro is a thinner liquid and although too much is not good and can make a mess, it will not cause issues if a bit too much is applied.

 

 

e. The Haswell processor you received, its quality and the condition of the thermal contact between the processor die and the inside of the cap that covers it.

 

 

Since every person reading this has a different situation with respect to every one of the 5 items listed above, there is no set way to define your exact starting point other than a conservative approach based on the cooling solutions I outlined in this thread, and, a defined set of conservative directions to follow in order to discover exactly where your system and Haswell processor will run safe and stable.

 

 

 

At this point you have selected one of the outlines listed under: Haswell and Overclocking 103 set up your BIOS as outlined and have found a stable manual CPU voltage that got the system booted into Windows and you are now for the first time ready to see just how stable and HOT your processor is going to run.

 

 

 

 

 

=============================================================

 
 
 
 
 
 
WARNINGS EVERYONE SHOULD READ AND UNDERSTAND

 

 

ROG BOARD USERS: You must be very keenly aware that these boards tend to automatically over-volt higher than the typical Asus boards. I will outline what to watch out for however there is no set way to know 'exactly' what each ROG board may or may not push in voltage increases as a AVX load is applied to Haswell. You will need to pay close attention to the BIOS and the hardware monitor readouts as the changes are applied with the limitations I will provide.

 

 

NON ROG BOARD USERS: Must also verify voltages as they are applied and tested but there is usually more leeway in settings with respect to automatic voltage increases with the non-ROG board designs.

 

 

So everyone is on the same page I must post the following:

 

 

UNLESS OTHERWISE NOTED PLEASE DO NOT EXCEED ANY OF THE DEFINED VOLTAGES OR TEMPERATURES I WILL OUTLINE FROM HERE, FORWARD.

 

If you do then you are leaving the 'COMBAT ZONE' and entering the 'DEAD MAN WALKING ZONE' where in one hour, one day, one week, one month or a few months you could very well be looking at a black screen and a CPU failure alarm, OR, instability due to a processor being damaged over time.

 

You may read somewhere else you can exceed the settings I have defined here. That is up to you. I will never approve nor outline any setting I deem risky.

 

This is not a overclocking website, its a Flight Sim site and the goal here is long term safe and stable as you are not in this game to win a internet trophy for being a fool.

 

Overclocks that are too high and not well checked or considered can make FSX run worse because the CPU cooling solution and the system are being driven too hard.

 

The goal is safe, stable and smooth, and not MAXIMUM SPEED on the bleeding edge of stability!

 

 

 

 

 

ABSOLUTE MAXIMUM SPECIFICATIONS

 

The following list outlines the maximum voltage and temperatures for ANY Haswell system regardless of what CPU speed you intend to run. These are my personal limits based on my calculations and taking into account

 

Please copy and print this cheat-sheet and use it as you load test your clock. You will most likely be testing several voltage changes in the process and will need this information so you know were the limits are. It is always best to NOT run a system to the limit on voltages

 

 

 

CPU CORE VOLTAGE OVERRIDE: 1.43v

 

The actual user BIOS setting for MAX CPU CORE VOLTAGE OVERRIDE is: 1.43v

 

Haswell will over-volt to 1.45/1.46 when a 1.43 CPU CORE VOLTAGE is set in the BIOS and a AVX load is applied, be very aware of this. Those on advanced liquid system that run load test temps exceptionally low at this voltage (70c or less @ max AVX test load) should have no issue with this a perhaps a touch higher however anyone maxing out the CPU TEMP (92c) under a full AVX test load and running this voltage is in a risky situation.

 

 

CPU CACHE VOLTAGE OVERRIDE:  1.30-1.32 will be close to 1.35/1.36 read in the BIOS. 1.35 AS READ in the BIOS is considered a hard maximum.

 

I strongly advise users to keep the actual voltage setting they make @ 1.31 or LESS, a user setting of 1.30 MAXIMUM is preferred!

 

 

 

CPU SYSTEM AGENT VOLTAGE OFFSET: VARIES with the model of motherboard and must be verified with HW Monitor after each change using a AVX load test.

 

Reading up to 1.196v in HW Monitor @ AVX Load in REAL TIME LOAD TESTING is the maximum preferred however 1.20v IS THE HARD MAX

 

SPECIAL NOTE: ROG BOARDS AND SYSTEM AGENT may very well over-volt what you set in the BIOS much further than other Asus boards. ROG users must pay very special attention to the readout in HW Monitor under the VOLTAGES section for VCCSA (SYSTEM AGENT) when working with CPU SYSTEM AGENT VOLTAGE OFFSET

 

 

CPU ANALOG I/O VOLTAGE OFFSET:  .200 hard maximum  (Do not exceed)

 

CPU DIGITAL I/O VOLTAGE OFFSET:  .200  hard maximum  (Do not exceed)

 

 

CPU INPUT VOLTAGE: 1.98  Do not run this high with a CPU CACHE voltage of 1.26 and greater, preferred is 1.92-1.88

 

 

---------------------------------------------------

 

NOTE: ROG BOARDS THAT DEFINE 2 CPU INPUT VOLTAGES:

 

INITIAL CPU INPUT VOLTAGE: 1.88

 

EVENTUAL CPU INPUT VOLTAGE: 1.98 Do not run this high with a CPU CACHE voltage of 1.26 and greater, preferred is 1.92-1.88

 

 

 

=============================================================

 
 
 
 

SPECIAL NOTE FOR DDR3 2700-2800+ USERS:

 

You may or MAY NOT be effected by this...

 

NOTE: This change to the clocking outlines provided in section 103 is for the following conditions and criteria ONLY:

 

1. You use or run higher than DDR3 2700

 

 

2. YOU HAVE started out with the outline for clock setup in section 103 but after many failed attempts at the testing sections presented BELOW you are having difficulties establishing a stable clock in any clock test (AIDA64, OCCT:Linpack or OCCT:CPU)

 

 

3. You suspect high CPU CORE VOLTAGE requirements could be due to memory speed and IMC stability running DDR3 speeds of 2700 and above.

 

 

As I posted out earlier in this thread DDR3 2400 would be the preferred top end memory speed for most since this memory speed is not difficult for Haswell to run and is far easier for the typical user to work with. Some users may opt for memory speeds well above 2400 and as such you may find the system could have difficulties with stable clocking which requires higher Vcore or simply wont stabilize. You could pass the Phase 1 temperature test but just can not seem to establish as stable clock in Phase 2 or Phase 3.

 

Those who run DDR3 frequencies above 2700 and are having difficulties obtaining stable results, this could be due to the higher DDR3 memory speed. Although Haswell will run exceptionally high memory speeds those frequencies can also hinder CPU and cache speed limits.

 

In the clocking outlines I provided the settings were established to ensure memory speeds of above DDR3 2000 to 2666 would have a baseline stability point however memory speeds above DDR3 2666 can present issues as CPU speed and CPU cache speed increases.

 

If you are using DDR3 2700 and above, you meet the 3 criteria I outlined in this section and are having difficulties establishing a stable clock, you can try increasing the following values:

 

 

CPU CACHE VOLTAGE OVERRIDE:   1.30   This is the MAX preferred setting for any clock. Although 1.31 can be done, when checking in the BIOS after setting this and rebooting into the BIOS to verify, the absolute maximum voltage reading in the BIOS should NEVER EXCEED 1.35v.  

 

 

CPU SYSTEM AGENT VOLTAGE OFFSET:   .300    ROG users be cautious and verify VCCSA VOLTAGE in HW Monitor does not exceed 1.20v under AVX load

 

CPU ANALOG I/O VOLTAGE OFFSET:   .150

 

CPU DIGITAL I/O VOLTAGE OFFSET:   .200    This is the absolute MAXIMUM voltage for this setting.

 

Although CPU Analog and Digital can both be set to a MAX of .200 it is preferred that the DIGITAL voltage live .050v higher than the ANALOG hence the difference shown in all of my outlined for A/D voltages.       

 

 

Be very aware that the above changed may or may not be required. As with any clock you must make the call to use these changes based on what you see in load test results. If you make the changes above and see no difference in stability testing, this means none of the above is required or making a difference. You can then return the settings back to the original and rework CPU CORE VOLTAGE instead.

 

It is also possible you simply have a Haswell processor with a finicky CPU IMC and simply does not like DDR3 speeds you are running, OR, the memory itself may be unstable and needs to be replaced. The motherboard itself could also be in the mix and part of the problem. This is why I outlined that DDR3 2400 is a far better solution for those who are not tech savvy. Defining the problem requires far more experience than most users have with computer systems.

 

 

=========================================================

 

 

 

 

 

 

 

THE TESTING PHASES

 

 

Each round of clock testing is performed in 3 phases. The first is the max temperature test, the 2nd is the hard stability test for the CPU speed and the 3rd, CPU cache tuning. As you make changes to your system you will be repeating these tests based on what you changed in the BIOS as you increase your clock.

 

 

a. IF you increase CPU Speed (CPU Ratio), increase the CPU Core Voltage Override (CPU Vcore) by .02v or MORE, or, increase the CPU Analog/Digital voltage offsets, you will repeat Test Phase 1 before moving to Test Phase 2 or Phase 3.

 

 

b. ONLY AFTER YOU HAVE PASSED TEST PHASE 1 IF: you have simply increased CPU Cache multiplier, increased the CPU Cache Voltage Override, increased System Agent but have not changed the values listed above in 'A', then you can SKIP retesting in Phase 1 and go directly to Test Phase 2 or Phase 3 (which ever you are working with).

 

 

The Phase 1 test is for HEAT and MAX LOAD. If you do NOT increase CPU voltage or A/D voltage then your temps should remain just about the same and you do not need to repeat Test Phase 1 once it passes. In clocking we start low and work our way UP, therefore if you do increase the speed and voltages you must retest for max temperature.

 

 

I know some people are going to get the bright idea that they can skip Test 1 and avoid the max temperature test thinking they will never use any AVX enabled software. Can you do this? Yes..   but if you do, remember what I said in Overclocking 104:

 

Unlike processors of the past,  anyone who sets up a clock and does not run this test thinking they will never see any issues by using a lower load test to define stability and get away with lower temps in that test has essentially loaded a gun, pulled the hammer back and is hoping no one comes along and bumps the trigger.

======================================

 

 

With that being said, lets move into clock testing!

 

 

 

 

 

 

 

TEST PHASE 1

 

THE TEST FOR MAX CPU CORE TEMPERATURE

 

 

At this point you have found the stable manual CPU CORE VOLTAGE OVERRIDE that has your system up and running in Windows with the speed profile selected. If you are still not stable booting into Windows you will need continue to raise the CPU core voltage in the BIOS by .01v until you achieve a stable boot.

 

EXAMPLE: If you are using the 4.1GHz outline and started out with a manual CPU CORE VOLTAGE OVERRIDE: 1.12v and are crashing, raise the CPU voltage to 1.13 and reboot. Repeat this process (1.13v, 1.14v, 1.15v) until you have established a stable boot into Windows and can continue below.

 

 

NOTE: It is possible that the system booted right into Windows with no issues and may be starting out with a voltage that is slightly HIGHER than what is required. That is fine! You will be raising the CPU Cache speed at some point and after all is said and done there will be testing sessions that will define the lowest CPU CORE VOLTAGE OVERRIDE. For right now you are looking for MAX TEMP and stability.

 

 

 

2. Launch CPUz and verify your CPU speed, CPU Ratio multiplier, PCIe Link width, Northbridge frequency, memory speed and memory timing as I showed in Haswell and Overclocking 104.

 

If there is anything out of sorts with the readouts in CPUz you must correct those now before proceeding.

 

Once you have verified the system close CPUz.

 

 

 

3. Launch HW Monitor. Using the mouse expand the box and the columns so you can read them as I showed in Haswell and Overclocking 104. Drag HW monitor to the side so it can remain on the screen and you can monitor your temps and voltages.

 

 

 

4. Launch your load test software..  The first test will be for basic stability but mostly for TEMPERATURE. you can use AIDA64 or OCCT:Linpack to accomplish this task as I outlined in Haswell and Overclocking 104.

 

a. If you use AIDA64, start the TOOLS> System Stability Test. The only item that should be checked is: STRESS  FPU

 

b. If you use OCCT, click the OCCT:Linpack tab, make sure the test is setup as I outlined in Haswell and Overclocking 104 and then place a check in the box: AVX Compatible Linpack

 

 

 

You are now ready for your first test run.

 

NOTE: OCCT will automatically STOP if any of the Haswell CPU cores or PACKAGE temps hit and exceeds 92c, AIDA64 will NOT stop. Be very aware of this! We left Intel Thermal Control [ENABLED] in the BIOS so if temps go through the roof the BIOS should protect the CPU and allow you shut the test down manually with no issues.

 

 

5. In AIDA64 click the START button in the lower corner of the test setup box. In OCCT click the green ON button.

 

With AIDA64 the temperatures and MAX load will be instantaneous. With OCCT it will take approximately 15-20 seconds for the test to ramp up to full load.

 

 

 

6. In HW Monitor you will be concerned with 3 primary areas during load testing:

 

 

 
 
a. TEMPERATURES: The PACKAGE MAX temp defines the highest core temp reached. The MAX column TEMP is never to exceed 92c (92c is max, 93c is unacceptable)

 

b. VOLTAGES: The MAX column CPU VCORE is NEVER to exceed 1.46v under a AVX or any other load.

 

c. VOLTAGES: The MAX column VCCSA (System Agent Voltage) is NEVER to exceed 1.20v under AVX or any other load.

 

 

Keep an eye on the temperature for now just to be sure. OCCT should shut the test down automatically with the setup I provided but just to be on the safe-side and until you feel confident in the system, do not leave the system during this test run and keep a eye on the MAX column PACKAGE temp in HW Monitor.

 

If you are using AIDA64 and you see a MAX column PACKAGE temp of 93c CLICK THE STOP button. With OCCT if for some reason OCCT does not shut the test down and you see the same, 93c, click the OFF button manually to stop the test. This would mean your cooling solution can NOT handle the heat. You either need to start at a LOWER CPU voltage/speed outline OR find out what is wrong with your tower for cooling, or, if there is nothing wrong upgrade the cooling solution so it will pass the TEMP tests for the speed you are trying to obtain.

 

I said this before, I will post it here again.. going TOO FAR, TOO FAST is not wise. You should start LOW, find out where your system stands for temperature and then work your way UP.

 

If for some reason you exceed 93c and OCCT did not shut down please check the OCCT settings and verify the temp and checkboxes as I outlined in section 104. There can be slight differences between OCCT temp sensors and HW Monitor but should not be much as OCCT uses the HW Monitor readout.

 

 

You will run this first test (AIDA or OCCT) for a total of 1 hour. At the end of that time you must click the STOP button in AIDA64, OCCT will shut down automatically. You can then note the HIGHEST CPU PACKAGE temp recorded by HW Monitor. This will tell you how much leeway you have in running a higher CPU CORE VOLTAGE OVERRIDE or higher clock speed.

 

As long as your test has run to the 1hr completion and not failed with a error message, the system has not rebooted by itself, no BSOD (blue screens displayed) AND you have confirmed the voltages and temps I outline above in HW Monitor, you are stable and running fine.

 

 

Assuming you PASS temperature checks, during the test if you experience any one of the following:

a. The test stops and posts a FAIL NOTICE which is not related to temperature.

 

b. The system BLUE SCREENS

 

c. The system REBOOTS by itself

 

d. The system LOCKS

 

Then you are running too low of a CPU CORE VOLTAGE OVERRIDE and must reboot (or hard reset) enter the BIOS and raise CPU CORE VOLTAGE OVERRIDE by .01v just like you did when finding the stable Windows boot voltage.

 

REPEAT STEPS 3, 4, 5 and 6 above until you PASS AIDA64: Stress FPU, or, OCCT:Linpack 64bit AVX ENABLED for 1 hour. Once you have passed then you can move to test your system for FULL STABILITY using the OCCT:CPU test

 

 

 

 

 

 

 

 

 

 

 

 

TEST PHASE 2

  

FULL STABILITY TESTING WITH OCCT:CPU

 

 

If you have gotten this far then your system has passed the initial load and temperature test for heat. The next test to be run is OCCT:CPU.

 

This next test will run the CPU far cooler than AIDA64 FPU and OCCT:Linpack but will reveal stability issues the other tests will never show. This next test will define just how stable your system is and as you progress MAY require a few other voltage settings be tweaked, but for the most part and since we are not clocking the CPU Cache up as of yet you should only need to deal with CPU CORE VOLTAGE OVERRIDE just as you did with the initial heat test runs.

 

 

1. Start HW Monitor and set it up on the desktop as displayed in Test Phase 1.

 

 

2. Start OCCT and select the OCCT:CPU tab.

 

 

3. Make sure the OCCT:CPU test is setup as defined in Overclocking 104 with the test set to Automatic 1hr, 64bit enabled with the test type as Large Data Set.

 

 

4. Click the START button in OCCT. Follow the same guidelines for temperature and voltage monitoring in HW Monitor as outlined in Phase 1.

 

 

You are simply repeating the same setup here but running the OCCT:CPU test to confirm stability.

 

 

Just as with the Phase 1 test, the same hold true with the OCCT:CPU test:

 

 

As long as your test has run to the 1hr completion and not failed with a error message, the system has not rebooted by itself, no BSOD (blue screens displayed) AND you have confirmed the voltages and temps I outline above in HW Monitor, you are stable and running fine, during the test if you experience any one of the following:

a. The test stops and posts a FAIL NOTICE which is not related to temperature.

 

b. The system BLUE SCREENS

 

c. The system REBOOTS by itself

 

d. The system LOCKS

 

Then you are running too low of a CPU CORE VOLTAGE OVERRIDE and must reboot (or hard reset) enter the BIOS and raise CPU CORE VOLTAGE OVERRIDE by .01v just like you did when finding the stable Windows boot voltage.

 

In this process if you have raised the CPU Core Voltage Override by more than .02v you must repeat the Phase 1 test once Phase 2 has passed to assure you are not overheating.

 

 

REPEAT STEPS 1 through 4 above until you PASS OCCT:CPU 64bit Large Data Set for 1 hour. Once you have passed both Phase 1 and Phase 2 then you can move to increasing the CPU Cache speed in Phase 3.

 

 

 

 

 

 

 

 

Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 13503
Direct Link To This Post Posted: September-20-2013 at 11:55am
 
 
 
 
 THIS POST IS NOW COMPLETE....
 
 
 
 
 
 
 
 
 
TEST PHASE 3

 

CPU CACHE TUNING

 

The 2nd King

 

 

In phase 1 and 2 you have defined the 1st King of clocking: The CPU Speed and have tested a safe and stable clock setup. In this section we will deal with the 2nd king of clocking, the CPU cache, or, 'northbridge frequency' as read in CPUz under the memory tab. The goal is to increase the CPU cache to a speed that is either equal or close to the CPU speed.

 

 

A few items you should clearly understand before starting this process...

 

1. CPU cache does NOT need to run 1:1 with the CPU speed at speeds above 4300. Although this would be optimal to do, its not required with Haswell.

 

 

2. As CPU speed increases, CPU cache frequency will be far more difficult to run 1:1 with the CPU core speed.  This normal.  

 

 

3. Once a CPU speed of 4500MHz is obtained, running a CPU cache speed of 100-200MHz behind the CPU speed is considered OK and is considered a good clock. But strive for no less than 100MHz if possible.

 

 

4. CPU speeds of 4600-4700-4800MHz will most likely require a CPU cache speed of 200-300-400MHz behind the CPU core speed as it increases. Again, this is OK and normal if required.

 

 

In other words, don't go crazy trying to get that last 100MHz out of the CPU cache. High CPU speed with high CPU cache speed can push the limits of Vcore, CPU cache voltage and System Agent.

 

 

 

======================================================

 

 

 

If you have made it this far then you have established a safe and stable clock using one of the outlines I provided in section 103 and have trim tuned that outline from there. Based on those clocking outlines your CPU cache multiplier is either 39 (3900MHz) or 41 (4100MHz) one of the two.

 

 

Our next goal is to increase the CPU cache speed so it either matches 1:1 for lower end clocks or is set to within a defined range of the CPU speed 300MHz, 400MHz max for very high CPU speeds. We don't want our cache running too far behind the CPU itself NOR do we run the cache faster than the CPU.

 

Given the extensive testing I have done I can pass this information. You can expect the following based on the CPU speed you run:

 

CPU SPEED - CACHE SPEED  (northbridge frequency as displayed in CPUz)

 

4100MHz - 4100MHz

4200MHz - 4200MHz

4300MHz - 4300MHz             This is the cutoff point where 1:1 becomes difficult

4400MHz - 4300-4400MHz   4400MHz is harder to obtain and can be tricky

4500MHz - 4300-4500MHz   4500MHz cache speed may not be easy to obtain

4600MHz - 4300-4500MHz   4500 cache speed is possible, but unlikely

4700MHz - 4300-4400MHz   4500 cache speed is most unlikely

4800MHz - 4300-4400MHz   4500 cache speed is most unlikely

 

Your goal will be to obtain at least the lowest CPU cache speed posted for the CPU speed displayed. 4400MHz CPU cache will most likely be the top-end once above 4500MHz CPU speed.

 

 

 

 

 

 

 

RAISING THE CACHE SPEED:

WHAT YOU SHOULD KNOW FIRST

 

 

 

You can approach this one of two ways, brute force, or, the one/two steps at a time method. The step method will allow you to get accustom to your system and understand where voltage increases are required and at what speeds. The brute force method can obtain a goal faster, or, it can drive you crazy since you are not learning where your processor and system require voltage as you go along and are simply jacking voltages and the cache to try and pass tests, which keep failing. Sometimes you get lucky and sometimes not.

 

In either case you will need to begin the process of raising certain voltages in your BIOS to stabilize the cache as the speed increases. The voltage requirements will be HIGHER as your CPU SPEED increases therefore:

 

IT IS IMPORTANT THAT YOU HAVE FOLLOWED THE OUTLINES I PRESENTED IN SECTION 103 FOR EACH CPU SPEED FIRST, TEST PHASE 1 AND 2 ABOVE AND ESTABLISH A STABLE CPU CLOCK BEFORE RAISING CACHE SPEED IN PHASE 3.

 

 

If you have followed the outlines you have already established voltage setups that have you close to success with the CPU cache. From there you will use trial and error with the OCCT:CPU test to define your final stable CPU cache speed.

 

The voltages that have the most effect on CPU cache stability in a clock are:

 

CPU CORE VOLTAGE OVERRIDE (Vcore)

CPU CACHE VOLTAGE OVERRIDE (Read only in the BIOS)

CPU SYSTEM AGENT VOLTAGE OFFSET  (VCCSA)

 

We will start out working with the last 2 but be very aware that CPU CORE VOLTAGE may need to increase and if it requires more than a .02v increase you will need to rerun the PHASE 1 test to confirm temperatures remain below the danger zone.

 

 

In some rare cases and with higher speed DDR3 memory the CPU ANALOG and DIGITAL voltages can be in the mix here. However those voltages target memory stability more than CPU cache and are rarely needed for cache tuning.

 

Those of you who may be running a very high CPU INPUT VOLTAGE (above 1.92) will need to take into consideration that if you RAISE the CPU CACHE VOLTAGE OVERRIDE to 1.30 that you need to trim the CPU INPUT VOLTAGE back to between 1.88 and 1.92 MAX

 

 

Now that you have a good base in understanding how to approach the CPU cache tuning.. lets get started.

 

 

 

 

 

 

CPU CACHE TUNING: THE OUTLINES TO BEGIN

 

 

Since there are different setups outlined for each CPU speed, and, you may have made tweak changes to those settings I will outline each clock profile starting point below based on the outlines from section 103.

 

 

1. Select the matching CPU SPEED outline you have tested and are currently using from section 103

 

 

2. Make the changes defined under the outline in the BIOS and CONFIRM them with a reboot into the BIOS. Check the voltage readings in the BIOS after the reboot and confirm they do not exceed the maximum defined.

 

 

3. Boot into Windows, launch CPUz and CONFIRM your CPU speed on the MAIN tab, then verify your CPU Cache speed (Listed as Northbridge Frequency) on the MEMORY tab. Once confirmed, close CPUz.

 

 

4. Launch HW Monitor and set it up as outlined under Phase 1. Make sure you can see the readouts and can monitor the voltages as defined for Vcore and VCCSA.

 

 

5. Launch OCCT and select to the OCCT:CPU tab. Make sure it is setup as defined for 64bit and 'Large Data Set'. Once confirmed hit the START button and let the test run.

 

 

NOTE: In this process if you find you must increase CPU CORE VOLTAGE by more than .02v you MUST rerun Test Phase 1 FIRST to verify temperatures are within safe limits before running the OCCT:CPU test to define hard stability.

 

REMEMBER: You can pass Phase 1 (AIDA64 Stress FPU, or, OCCT:Linpack) for stability and FAIL the OCCT:CPU 64bit Large Data Set test.

 

 

 

6. IF you PASS and you are now running the 103 CPU speed profile and have established your CPU cache speed, write down all your BIOS settings for future reference then:

 

a. You can see by the voltages and temps during Phase 1-2-3 that you have ROOM to increase your CPU clock you can now decide to make that move. If you wish to try a higher clock then move to the next to the next level CPU speed profile outlined in section 103 and set it up as defined. You will be starting over to find the next higher step in CPU/cache speeds.

 

 

b. Lock your CPU clock where it is then move to the last King of clocking: Haswell and Overclocking 106: Memory Tweaking

 

 

 

 

 

 

 

CPU CACHE PROFILE OUTLINES

 

 

 

 

STARTING SETUP OUTLINE: 4.0GHz-4.1GHz:

 

Your current settings for this profile are:

 

MIN CPU CACHE: 39

MAX CPU CACHE: 39

CPU CACHE VOLTAGE OVERRIDE:   1.22

CPU SYSTEM AGENT VOLTAGE OFFSET:   .150

 

CPU CORE VOLTAGE OVERRIDE:  You established this in Phase 1 and Phase 2. Leave this voltage alone to begin the process.

 

 

Most likely you will not need to change any voltage for this cache speed however if you FAIL the load test I will outline the voltages to work with here. You will start with the profile established voltages and simply change the CPU Cache Multiplier in the BIOS:

 

 

If you are running a CPU speed of 4000MHz you will set your CPU Cache Multiplier in the BIOS as follows:

 

MIN CPU CACHE: 40

MAX CPU CACHE: 40 

Note: Sometimes this requires the MAX value be changed first then the MIN

 

 

If you are running a CPU speed of 4100MHz you will set your CPU Cache Multiplier as follows:

 

MIN CPU CACHE: 41

MAX CPU CACHE: 41 

Note: Sometimes this requires the MAX value be changed first then the MIN

 

 

 

Run the OCCT:CPU 64bit Large Data Set test:

 

a. If you pass, at this point you can make the decision as to what you wish to do next based on the voltages and temps you are using. You can lock your clock here or decide to move up in CPU speed. If there is room to increase speed, then start with the next CPU speed profile outline listed in section 103. Start there and then start over with PHASE 1.

 

Should you decide to lock your clock where it currently sits stable, proceed to the final phase: Haswell and Overclocking 106: Memory Tweaking

 

 

 

b. If you fail, start by raising the voltages as outlined below:

 

 

CPU CACHE VOLTAGE OVERRIDE:   1.25

CPU SYSTEM AGENT VOLTAGE OFFSET:   .200

 

Reboot and retest. If you pass, go to 'A' above, ... If you fail:

 

CPU CORE VOLTAGE OVERRIDE:  Increase by .01v  Example: If currently 1.13v, increase to 1.14v

 

Reboot and retest. If you pass, go to 'A' above, ... If you fail, raise:

 

CPU CACHE VOLTAGE OVERRIDE:   1.27

CPU SYSTEM AGENT VOLTAGE OFFSET:   .220

 

NOTE: 4000-4100 CPU cache is not that hard to accomplish at lower CPU speed. In theory you should have passed the test by now but as I have said when it comes to Haswell there can be a lot of variations to how the processor runs stable. If you are still failing the OCCT:CPU test you should at this point verify and make sure all your BIOS settings are correct from the first section of the 103 outline, the full BIOS setup. If it all checks out along with the settings you have made above under the AI Tweaker menu, then proceed below and continue.

 

 

Reboot and retest. If you fail continue this process continue by increasing:

 

CPU CACHE VOLTAGE OVERRIDE:   by .01v

CPU SYSTEM AGENT VOLTAGE OFFSET:  by .010v

 

Increase them both at the same time. You will have more room to increase system agent than CPU Cache. Once you have reached 1.30 in CPU cache continue raising system agent in .010v increments.

 

Retest each step up until you reach the manual CPU cache voltage override setting of 1.30v in the BIOS (this will actually read higher in the BIOS once set to 1.30 manually) AND a VCCSA voltage reading of 1.196 in HW Monitor in Windows under a load.

 

Example: If you set Cache Voltage to 1.28v and System Agent Offset to .220 and failed, then raise them again as outlined above noting the limits and check HW Monitor for the full VCCSA voltage in real time load.

 

 

Once you have hit the MAX values for CPU CACHE (setting 1.30v in the BIOS) and SYSTEM AGENT (1.196 in HW monitor) then leave those voltages locked and then start raising:

 

CPU CORE VOLTAGE OVERRIDE:  Increase by .01v  Example: If you raised to 1.14v previous, raise it to 1.15v) Continue this until you pass OCCT:CPU

 

Once you pass then LOWER: CPU CACHE VOLTAGE OVERRIDE: by .01v for each step and retest until you fail, then set it back up one tick to the last successful value.

 

Then LOWER: CPU SYSTEM AGENT VOLTAGE OFFSET:   .010 for each step and retest until you fail, then set it back up one tick to the last successful value.

 

 

Retest each step LOWER and continue to lower them as outlined above until you FAIL, then raise them back up to the last values that passed. This is how we fine tune our voltages.

 

At this point you MUST retest for heat using TEST PHASE 1 if you added more than .02v to the CPU VOLTAGE OVERRIDE. Verify you are still running safe temps with full AVX load.

 

 

Once you have passed all tests at the desired CPU/Cache speed you can make the decision as to what you wish to do next based on the voltages and temps you are using. If there is room to increase speed and you wish to try for a higher CPU speed, then start with the next speed profile outline listed in section 103.

 

IF your current CPU CORE VOLTAGE OVERRIDE or other voltage settings are currently set higher than the next step up in the speed profile outline I posted, then leave them as-is and proceed to change the voltages and multipliers that are lower, only. Start there and then start over with PHASE 1

 

Should you decide to lock your clock where it currently sits stable, proceed to the final phase: Haswell and Overclocking 106: Memory Tweaking

 

 

------------------------------------------------------------------------

 

 

 

 

 

 

 

STARTING SETUP OUTLINE: 4.2GHz-4.3GHz:

 

Your current settings for this profile are:

 

MIN CPU CACHE: 39

MAX CPU CACHE: 39

CPU CACHE VOLTAGE OVERRIDE:   1.25

CPU SYSTEM AGENT VOLTAGE OFFSET:   .200

 

CPU CORE VOLTAGE OVERRIDE:  You established this in Phase 1 and Phase 2. Leave this voltage alone to begin the process.

 

 

You will start with the profile established voltages and simply change the CPU Cache Multiplier in the BIOS:

 

If you are running a CPU speed of 4200MHz you will set your CPU Cache Multiplier in the BIOS as follows:

 

MIN CPU CACHE: 42

MAX CPU CACHE: 42 

Note: Sometimes this requires the MAX value be changed first then the MIN

 

 

If you are running a CPU speed of 4300MHz you will set your CPU Cache Multiplier as follows:

 

MIN CPU CACHE: 43

MAX CPU CACHE: 43 

Note: Sometimes this requires the MAX value be changed first then the MIN

 

 

This is where things get a little trickier. You will start with the profile set voltages as defined above and observe the results. You may very well FAIL the OCCT test, or..   you might pass!

 

 

Run the OCCT:CPU 64bit Large Data Set test:

 

a. If you pass, at this point you can make the decision as to what you wish to do next based on the voltages and temps you are using. You can lock your clock here or decide to move up in CPU speed. If there is room to increase speed, then start with the next CPU speed profile outline listed in section 103. Start there and then start over with PHASE 1.

 

 

Should you decide to lock your clock where it currently sits stable, proceed to the final phase: Haswell and Overclocking 106: Memory Tweaking

 

 

 

b. If you fail, start by raising the voltages as outlined below:

 

CPU CACHE VOLTAGE OVERRIDE:   1.28

CPU SYSTEM AGENT VOLTAGE OFFSET:   .230

 

Reboot and retest. If you pass, go to 'A' above, ... If you fail:

 

CPU CORE VOLTAGE OVERRIDE:  Increase by .01v  Example: If currently 1.18v, increase to 1.19v

 

Reboot and retest. If you pass, go to 'A' above, ... If you fail:

 

CPU CACHE VOLTAGE OVERRIDE:   1.30  (this is max)

CPU SYSTEM AGENT VOLTAGE OFFSET:   .240

 

NOTE: 4200-4300 CPU cache is not as easy as 40-41 accomplish at lower CPU speed but it is also not that hard either! In theory you should have passed the test by now but as I have said when it comes to Haswell there can be a lot of variations to how the processor runs stable. If you are still failing the OCCT:CPU test you should at this point verify and make sure all your BIOS settings are correct from the first section of the 103 outline, the full BIOS setup. If it all checks out along with the settings you have made above under the AI Tweaker menu, then proceed below and continue.

 

 

Reboot and retest. If you fail continue this process continue by raising:

 

 

CPU SYSTEM AGENT VOLTAGE OFFSET:   by .020v and RETEST in OCCT until you reach 1.196 in HW Monitor in Windows under a load. Example: If you set .240 the raise this to .260, .280, .300, etc, and check HW Monitor for the full VCCSA voltage in real time load until it peaks at 1.196v

 

 

Once you have hit the MAX values for CPU CACHE (1.30v) and SYSTEM AGENT (1.196 MAX column in HW Monitor) then lock those voltages and start raising:

 

CPU CORE VOLTAGE OVERRIDE:  Increase by .01v  Example: If you raised to 1.19v previous, raise it to 1.20v.  Continue this until you pass OCCT:CPU

 

Once you pass then LOWER: CPU CACHE VOLTAGE OVERRIDE: by .01v for each step and retest until you fail, then set it back up one tick to the last successful value.

 

Then LOWER: CPU SYSTEM AGENT VOLTAGE OFFSET:   .010 for each step and retest until you fail, then set it back up one tick to the last successful value.

 

 

Retest each step LOWER and continue to lower them as outlined above until you FAIL, then raise them back up to the last values that passed. This is how we fine tune our voltages.

 

At this point you MUST retest for heat using TEST PHASE 1 if you added more than .02v to the CPU VOLTAGE OVERRIDE. Verify you are still running safe temps with full AVX load.

 

 

 

Once you have passed all tests at the CPU/Cache speed you can make the decision as to what you wish to do next based on the voltages and temps you are using. If there is room to increase speed and you wish to try for a higher CPU speed, then start with the next speed profile outline listed in section 103.

 

IF your current CPU CORE VOLTAGE OVERRIDE or other voltage settings are currently set higher than the next step up in the speed profile outline I posted, then leave them as-is and proceed to change the voltages and multipliers that are lower, only. Start there and then start over with PHASE 1

 

Should you decide to lock your clock where it currently sits stable, proceed to the final phase: Haswell and Overclocking 106: Memory Tweaking

 

 

 

 

 

------------------------------------------------------------------------

 

 

 

 

 

 

 

STARTING SETUP OUTLINE: 4.4GHz-4.5GHz:

 

 

Your current settings for this profile are:

 

MIN CPU CACHE: 39

MAX CPU CACHE: 39

CPU CACHE VOLTAGE OVERRIDE:   1.28

CPU SYSTEM AGENT VOLTAGE OFFSET:   .280

 

CPU CORE VOLTAGE OVERRIDE:  You established this in Phase 1 and Phase 2. Leave this voltage alone to begin the process

 

 

 

In this clock outline the final voltages will most definitely be higher than in the last 2 profile outlines. CPU cache speeds of 4400 and 4500 are the most difficult to obtain stable. But that also depends on the CPU itself and the CPU speed too.

 

In this scenario instead of going ALL OUT I suggest users start at 4300 cache speed for a 4400-4500MHz CPU speed and THEN raise the CPU cache 1x, and then retest until you reach your goal.

 

THEREFORE TO START OUT

 

If you are running a CPU speed of 4400-4500MHz you will set your CPU Cache Multiplier in the BIOS as follows:

 

MIN CPU CACHE: 43

MAX CPU CACHE: 43

Note: Sometimes this requires the MAX value be changed first then the MIN

 

IF YOU FAIL then follow the directions below to increase voltages and retest until you pass.

 

ONCE YOU PASS then increase the MIN/MAX CPU CACHE and retest until you have obtained your final goal:

 

 

MIN CPU CACHE: 44

MAX CPU CACHE: 44  - and then repeat your tests increasing voltages as required. Once you have established a safe and stable clock with this CPU cache speed you will stop here if your CPU speed is 4400MHz. You can also opt to stop here with a CPU speed of 4500MHz.

 

This next step is not easy to obtain but can be done:

 

MIN CPU CACHE: 45

MAX CPU CACHE: 45  IF YOU MAKE IT HERE consider yourself VERY LUCKY! A 4500-4500 clock is exceptional.

 

REMEMBER: Not all Haswell processors will run the same and you may find once you reach the MAXIMUM VOLTAGES possible or the MAX TEMP with CPU CORE VOLTAGE that you must accept what the system and cooling solution will allow.

 

 

 

Run the OCCT:CPU 64bit Large Data Set test:

 

a. If you pass, at this point you can make the decision as to what you wish to do next based on the voltages and temps you are using. You can lock your clock here or decide to move up in CPU speed. If there is room to increase speed, then start with the next CPU speed profile outline listed in section 103. Start there and then start over with PHASE 1.

 

 

Should you decide to lock your clock where it currently sits stable, proceed to the final phase: Haswell and Overclocking 106: Memory Tweaking

 

 

 

b. If you fail, start by raising the voltages as outlined below:

 

CPU CACHE VOLTAGE OVERRIDE:   1.30 (this is the MAX)

CPU SYSTEM AGENT VOLTAGE OFFSET:   .290

 

Reboot and retest. If you pass, go to 'A' above, ... If you fail:

 

CPU CORE VOLTAGE OVERRIDE:  Increase by .01v  Example: If currently 1.23v, increase to 1.24v

 

Reboot and retest. If you pass, go to 'A' above, ... If you fail then raise:

 

CPU SYSTEM AGENT VOLTAGE OFFSET:   .300

 

Reboot and retest. If you fail continue this process continue by raising:

 

 

CPU SYSTEM AGENT VOLTAGE OFFSET:   by .010v and RETEST in OCCT. Repeat this until you reach 1.196 in HW Monitor in Windows under a load. Example: If you set .300 then raise this in steps to .310, .320, .330 and check HW Monitor for the full VCCSA voltage in real time load until it peaks at 1.196v   Above .300 is going to be getting close to the max VCCSA to keep a sharp eye on that MAX VCCSA voltage.

 

 

Once you have hit the MAX values for CPU CACHE (1.30v) and SYSTEM AGENT (1.196 MAX column in HW Monitor) then lock those voltages and start raising:

 

CPU CORE VOLTAGE OVERRIDE:  Increase by .01v  Example: If you raised to 1.24v previous, raise it to 1.25v Continue this until you pass OCCT:CPU

 

Once you pass then LOWER: CPU CACHE VOLTAGE OVERRIDE: by .01v for each step and retest until you fail, then set it back up one tick to the last successful value.

 

Then LOWER: CPU SYSTEM AGENT VOLTAGE OFFSET:   .010 for each step and retest until you fail, then set it back up one tick to the last successful value.

 

 

Retest each step LOWER individually and continue to lower them as outlined above until you FAIL, then raise them back up to the last values that passed. This is how we fine tune our voltages.

 

 

At this point you MUST retest for heat using TEST PHASE 1 if you added more than .02v to the CPU VOLTAGE OVERRIDE. Verify you are still running safe temps with full AVX load.

 

 

REMEMBER: You may have accept a 100-200MHz cache difference between the CPU speed and the cache speed. This all depends on the CPU and the system cooling solution.

 

 

Once you have passed all tests at the CPU/Cache speed you can make the decision as to what you wish to do next based on the voltages and temps you are using. If there is room to increase speed and you wish to try for a higher CPU speed, then start with the next speed profile outline listed in section 103.

 

IF your current CPU CORE VOLTAGE OVERRIDE or other voltage settings are currently set higher than the next step up in the speed profile outline I posted, then leave them as-is and proceed to change the voltages and multipliers that are lower, only. Start there and then start over with PHASE 1

 

Should you decide to lock your clock where it currently sits stable, proceed to the final phase: Haswell and Overclocking 106: Memory Tweaking

 

 

 

 

 

------------------------------------------------------------------------

 

 

 

 

 

 

STARTING SETUP OUTLINE: 4.6GHz-4.8GHz:

 

This clock speed profile is most likely only going to be used by those with exceptional high-end liquid cooling OR those who have performance the IHS modification (outlined later in this thread) and using high end liquid or very high-end air cooling solutions.

 

This outline is the same as the 4400-4500 outline with the exception of the starting voltages. You will use the same method starting with a CPU cache speed of 4300 and work your way up until you reach the highest CPU cache speed you can obtain safe and stable. You will most likely be working with the absolute maximum voltages possible to obtain your final clock speeds

 

 

Your current settings for this profile are:

 

MIN CPU CACHE: 41

MAX CPU CACHE: 41 

CPU CACHE VOLTAGE OVERRIDE:   1.30v This is already MAX however it can go to 1.31 if all else fails. Remember I consider above 1.30 pushing the CPU very hard and a reboot voltage reading in the BIOS of 1.35v is the absolute maximum.

 

.

CPU SYSTEM AGENT VOLTAGE OFFSET:   .300  For ROG boards this may very well be sitting on the edge of maximum. You must check the VCCSA voltage under a load in HW Monitor to confirm VCCSA MAX is 1.196 or less. 1.20v can be done but that is the absolute maximum voltage.

 

CPU CORE VOLTAGE OVERRIDE:  You established this in Phase 1 and Phase 2. Leave this voltage alone to begin the process

 

 

THEREFORE TO START OUT

 

If you are running a CPU speed of 4600-4800MHz you will set your CPU Cache Multiplier in the BIOS as follows:

 

MIN CPU CACHE: 43

MAX CPU CACHE: 43 

Note: Sometimes this requires the MAX value be changed first then the MIN

 

IF YOU FAIL then follow the directions below to increase voltages and retest until you pass.

 

ONCE YOU PASS then increase the MIN/MAX CPU CACHE and retest until you have obtained your final goal.

 

 

MIN CPU CACHE: 44

MAX CPU CACHE: 44  - and then repeat your tests increasing voltages as required. Once you have established a safe and stable clock with this CPU cache speed you will stop here if your CPU speed is 4700MHz and consider yourself lucky. You can also opt to stop here with a CPU speed of 4800MHz.

 

 

This next step is not easy to obtain AT ALL using safe stop points:

 

MIN CPU CACHE: 45

MAX CPU CACHE: 45  IF YOU MAKE IT HERE consider yourself VERY LUCKY! A 4800-4500 clock is out of this world and considered the holy grail of safe clocking with Haswell.

 

 

Run the OCCT:CPU 64bit Large Data Set test:

 

a. If you pass, at this point you can make the decision as to what you wish to do next based on the voltages and temps you are using. You can lock your clock here or decide to move up in CPU speed. The limit is 4800MHz. If there is room to increase speed, then start with the next CPU speed profile outline listed in section 103. Start there and then start over with PHASE 1.

 

 

Should you decide to lock your clock where it currently sits stable, proceed to the final phase: Haswell and Overclocking 106: Memory Tweaking

 

 

 

b. If you fail, start by raising the voltages as outlined below:

 

CPU SYSTEM AGENT VOLTAGE OFFSET:   .310

 

Reboot and retest. If you pass, go to 'A' above, ... If you fail:

 

CPU CORE VOLTAGE OVERRIDE:  Increase by .01v  Example: If currently 1.34v, increase to 1.35v

 

Reboot and retest. If you pass, go to 'A' above, ... If you fail then raise:

 

CPU SYSTEM AGENT VOLTAGE OFFSET:   .320

 

Reboot and retest. If you fail continue this process continue by raising:

 

 

CPU SYSTEM AGENT VOLTAGE OFFSET:   by .010v and RETEST in OCCT. Repeat this until you reach 1.196 in HW Monitor in Windows under a load. Example: If you set .320 then raise this in steps to .330 and check HW Monitor for the full VCCSA voltage in real time load until it peaks at 1.196v   Above .300 is going to be getting close to the max VCCSA to keep a sharp eye on that MAX VCCSA voltage.

 

 

Once you have hit the MAX values for CPU CACHE (1.30v) and SYSTEM AGENT (1.196-1.20 MAX column in HW Monitor) then lock those voltages and start raising:

 

CPU CORE VOLTAGE OVERRIDE:  Increase by .01v  Example: If you raised to 1.35v previous, raise it to 1.36v Continue this until you pass OCCT:CPU

 

REMEMBER: The MAX CPU CORE OVERRIDE is 1.43v which will most likely read 1.456-1.462 in HW Monitor under a load. I specified 1.462 to be the absolute maximum.

 

 

Once (or IF) you pass then LOWER: CPU CACHE VOLTAGE OVERRIDE: by .01v for each step and retest until you fail, then set it back up one tick to the last successful value.

 

Then LOWER: CPU SYSTEM AGENT VOLTAGE OFFSET:   .010 for each step and retest until you fail, then set it back up one tick to the last successful value.

 

Retest each step LOWER individually and continue to lower them as outlined above until you FAIL, then raise them back up to the last values that passed. This is how we fine tune our voltages.

 

Most likely you will find that you can not lower the CPU Cache and System Agent voltages, but the lower/stable, the better.

 

 

At this point you MUST retest for heat using TEST PHASE 1 if you added more than .02v to the CPU VOLTAGE OVERRIDE. Verify you are still running safe temps with full AVX load.

 

 

REMEMBER: You will have accept a 300-400MHz cache difference between the CPU speed and the cache speed at these CPU clock speeds. This all depends on the CPU and the system cooling solution however in this case I have posted a additional solution should you find your CPU cache running 400-500MHz behind the CPU speed. If you are running 300-400MHz behind the CPU speed you can opt to lock it as that is a great clock. If you are running 400-500 behind and wish to see IF you can trim that in better, then see the section below: HOW TO GET A LITTLE MORE FOR A LITTLE LESS

 

 

 

Once you have passed all tests at the CPU/Cache speed (and you are below 4800) you can make the decision as to what you wish to do next based on the voltages and temps you are using. If there is room to increase speed to the max of 4800 and you wish to try for a higher CPU speed then continue by using PHASE 1, 2 and 3 for the next step up until you reach the highest speed you can run (4700 or 4800)

 

 

Should you decide to lock your clock where it currently sits stable, proceed to the final phase: Haswell and Overclocking 106: Memory Tweaking

 

 

 

Congratulations are in order if you managed a 4800/4500 clock. Even a 4800/4000 clock is considered exceptional with Haswell. BeerParty

 

 

 

 

 

=======================================================

 

 

 

 

 

THIS IS OPTIONAL BASED ON THE CIRCUMSTANCES:

 

HOW TO GET A LITTLE MORE FOR A LITTLE LESS

 

This is optional, not a requirement. Its not going to make your system ignite and fly off the launch pad at Mach 10.. so lets just get that said up-front. This can be done if the CPU is finicky and unstable at a higher CPU speed and is running a wide gap in frequency with the CPU cache.

 

I did this for stability and not so much performance after burning my Haswell testing 5GHz clock speeds and voltages I have warned people NOT to exceed. CPU damage is not always catastrophic, on the contrary it usually comes on in the form of loss of stability at higher clock speeds.

 

The damage forced my Haswell to no longer run stable at 4500/4500 on lower voltages. I will be replacing my Haswell after this past summers testing. I risked that damage to my Haswell in order to define the real world limits and pass them on to you.

 

You will need to make the call on this one but just be aware doing this when you are already running within 300-400Mhz of the CPU speed is not going to make a huge difference. Those of you that may find your CPU cache is running 400-500MHz behind the CPU speed should consider doing this.

 

 

This method can be used for any higher-end clock whereby you wish to trim back on one end for a bit more stability and increase the performance on the other side.. its a GIVE AND TAKE setup and not a "performance boost"

 

We don't want to lose too much CPU speed to accomplish this. In other words, if you find you must drop the CPU speed by more than 150-200MHz in order to raise the CPU cache 50MHz, then do not use this trim method.

 

Some of you may have noticed that I am running a CPU cache speed of 4444MHz and a CPU speed of 4747MHz which places my CPU core and cache speed to within 300MHz of each other. I used a setup that slightly alters BCLK to accomplish that and will describe that in this section.

 

If you happen to be running a CPU speed of above 47-4800 and your cache is running greater than 300MHz behind the CPU speed, once you have established the highest values you can obtain, if you wish to try to tune a better cache speed I have outlined that here. Its not a requirement but when you are pushing CPU speeds of 4700-4800, if the cache is running 400+MHz behind a little trimming on the CPU speed to allow a higher cache speed (within 300MHz) is not a bad move.

 

To accomplish this:

 

1. Enter the BIOS and set the BCLK FREQUENCY: 101

 

NOTE: This will increase the CPU speed, Cache Speed and the Memory Speed.

 

 

2. Check the DRAM FREQUENCY. It will have increased by a small amount. Check the list and make sure you are set to the closest memory speed just ABOVE the manufactures default memory speed.

 

In example, if my default manufacture memory speed is 2400, the next step UP in the list displays 2424. Therefore I will confirm/set my memory speed to 2424MHz in the BIOS since it is closest to 2400.

 

 

 

3. Verify the CPU CORE RATIO is set to PER CORE and then set each of the next 4 boxes to 1x LOWER than your current multiplier.

 

Example: If I am running a CPU multiplier of 48, LOWER this to 47

 

47 x 101 BCLK = 4747MHz CPU speed...   I just dropped my CPU speed 53Mhz

 

 

4. LEAVE both the MIN/MAX CPU CACHE alone. Since we are trying to get the CPU speed and the CPU cache speed within 300MHz of each other, this will INCREASE the CPU cache speed

 

In Example: If I am running a CPU MIN/MAX CACHE of 44 then:

 

44 x 101 BCLK = 4444MHz   I just RAISED my CPU cache by nearly 50Mhz

 

 

4747 - 4444 = 303MHz     GOOD ENOUGH! 

 

 

Of course you must TEST this change with OCCT:CPU 64bit Large Data Set. If the tests fails, use the same PHASE 3 clock testing outline you worked with to obtain the current CPU cache and work with the voltage outlines as defined until the test passes.

 

 

 

 

 

 

=============================================================

 

 

 

 

 

This completes the three primary phases of clocking a Haswell processor. In the next section we will move on to the last king of clocking, the memory latency.

 
 
 
 
 
 
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 13503
Direct Link To This Post Posted: September-20-2013 at 12:28pm
 
 
 
THIS SECTION IS UNDER CONSTRUCTION
 
PHASE 1 AND PHASE 2 OF THIS SECTION ARE NOW COMPLETE. PHASE 3, THE ADVANCED MEMORY TWEAKING SECTION IS STILL UNDER CONSTRUCTION. YOU ARE WELCOME TO USE PHASE 1/2 AT THIS TIME.
 
 
 
THANK YOU
 
 
 
 
 
 
 
 
 
 
 
 
Haswell and Overclocking 106 -
 
Memory Testing and Tweaking: The 3rd King
 
 
 

 

Now that you have established a safe and stable CPU clock we must test our memory to be sure that the clock we have setup is not destabilizing the memory itself. This testing can reveal an unstable IMC which may require a little tweaking of the DRAM Voltage or CPU Analog/Digital voltages. These tests can also reveal weak or defective memory as well so it is a very important last step to setting up any clock.

 

 

If we find the initial tests pass as a confirmation our current clock is secure and stable, then we move into the last king of clocking, memory tweaking.

 

 

For most this is going to be a very simple one or two setting change. I am not going to be getting into high-end memory clocking because that kind of clocking requires far more experience than the typical user has and high-end memory clocking presents a lot of dangers in stability and corruption.

 

 

Instead I will be going over a basic way to trim memory latency, which in turn increases performance by ramping up the communication rate between the CPU cache and physical memory.

 

 

 

 

 

 

 

THE TEST SOFTWARE AND SETUP

 

 

During your system build and if you followed my outlines you have already tested your memory using Memtest in a clean boot environment. Memtest can also be used now in the same way to check your system for stability at the IMC and the memory modules while clocked. In this section I will outline a method to test your memory in Windows which is a more realistic test analog in so much to how we use our systems. The test software is easy to use but does require a bit of setup to use it correctly. I will outline the process below.

 

 

Download this small memory testing application: http://hcidesign.com/memtest

 

I have packaged and linked it here directly if for some reason the site that hosts the application goes down: http://gex.flight1.net/forumimages/TEST_RESULTS/WinMEMTest.zip

 

 

Unlike Memtest, this application works by testing memory chunks per processor core. You will be launching one instance of the application for each physical processor core and then dividing memory amounts to be tested between each physical core. Since it is not possible to test memory reserved or in use by Windows a little math is needed to correctly assign memory amounts to each core and then run the test.

 

This test is about checking real time and real world stability of your memory in a clocked state.

 

Before running this test, clean boot the system and give the system about 3-5 minutes after the reboot to fully load. Once on the desktop proceed as follows:

 

1. Unzip the test application. There are instructions included with the package which you can review. I have provided the steps below.

 

 

2. First, since our Haswell processors are 4 physical cores we must define the memory amount to run on each core. We do want to test as much memory as possible without treading into the memory reserved or in use. In order to simplify this, I have outlined the amounts to use per core based on the installed memory amount.

 
FOR:
 
16GB installed physical memory use: 3328 per core
 
12GB installed physical memory use: 2699 per core
 
8GB installed physical memory use: 1690 per core
 
6GB installed physical memory use: 1178 per core
 
4GB installed physical memory use: 666 per core
 
 
 
These are the values you will use in Test Phase 1 based on the amount of physical memory installed.

 

 

 

 

 

 

TEST PHASE 1:

 

MEMORY VERIFICATION IN A CLOCKED SYSTEM

 

 

1. Open the unzipped directory and launch the first instance of: memtest.exe. A small box will appear, select OK then move the small box to the side. Repeat this 4 times, launch the application again, and place 4 instances of the application on the screen so you can clearly see all 4.

 

We launch one instance for each core of the processor.

 

 

2. Input the amount of memory based on the amount of physical memory you run, which I outlined above, into each box in the top section that says: "All unused RAM". Replace that text with the memory amount defined.

 

 

3. Click the START TESTING button of EACH box. A message box will appear for each instance started. Simply click OK and close each message box.

 

 

4. Allow the test to run. At the bottom of the boxes you will see the COVERAGE readout and the ERRORS readout. Once EACH BOX displays 100% Coverage and there are 0 errors you can close the box. Allow each instance to run until they reach 100% Coverage.

 

 

-ALL TESTS PASS-

 

IF you pass all 4 instances which is most likely the result that you will see at this point if you are running the manufacture speed/timing of the memory, then you are finished for now with Phase 1 of this test. Your memory has checked in the clocked system. Click the STOP TESTING button, close the test boxes and proceed to PHASE 2 below.

 

 

 

-ONE OR MORE TESTS FAIL-

 

IF errors are detected the application will start posting error reports. A failure indicates one of the following three possibilities:

a. Your CPU IMC (memory controller) is not quite stable: A bit of voltage may be required to trim it in.

 

b. Your memory is not stable in the clock you established: This is the most common and can be due to voltage or the memory itself. Memory does not necessarily have to be 'defective' to be 'weak' in a clocked environment. Usually cheaper quality designed memory suffers in that respect.

 

c. RARE but possible: Your memory has started to show signs of manufacture defect since you tested it last.

 

 

NOTE: This information presented below is the same for memory that run manufacture specification for speed/timing, OR, memory that is being clocked up for speed/timing.

 

 

The reason for the failure is most likely 'B' (or 'A') and some minor tweaking may be required to stabilize the memory or the CPU IMC.

 

 

There have been cases where I have seen memory sticks PASS Memtest with flying colors and a week or two later (or even a month later) begin to show signs of manufacture DEFECT. This is NOT completely out of the question at this time and is rare to see happen but it can and does happen occasionally. Before checking for that possibility we will first make a few minor changes in the BIOS and rerun the memory test as outlined above.

 

 

 

 

-THE VOLTAGE CHANGES IF TEST FAILURES APPEAR-

 

The voltages we will work with are:

DRAM VOLTAGE - This was set automatically using the memory XMP profile you loaded when you set up your BIOS.

 

I will discuss this subject here in some detail because there is a lot of misinformation around this voltage posted on the internet.

 

You have room to work with DRAM voltage. How much is defined by THE MEMORY MANUFACTURE and not Intel specifications or motherboard manufactures. Although there is a Intel spec for DRAM voltage it is based on far more than one factor. In fact professional clockers know they can run DRAM voltage to above 1.80v (DO NOT ATTEMPT THAT YOURSELF) as long as the memory design can handle that voltage and proper memory cooling is applied as well as the CPU is properly cooled.

 

The voltage your memory can run safe is defined by its design. Most high performance memory has a XMP set voltage of 1.60-1.65v. Some memory is XMP programmed to run at 1.50-1.55v. That does not mean the memory can not run higher voltage, it simply means the manufacture placed that into its XMP profile for typical operation. HOWEVER, sometimes that limit is placed on the memory modules that run 1.50-1.55 due to the way it is designed and it can not run higher safely or with a reliable guarantee no damage will occur. The on-stick voltage regulator quality can define what it can handle on the input side and in some cases can define that voltage limit.

 

THEREFORE if you question what voltage your memory can and can not run safely, seek the advice of the memory manufacture directly. They will tell you what their modules can and can not run.

 

This is a rule of thumb I have followed to start out and its never presented any issue:

 

IF the memory is rated 1.50-1.55 then a max of 1.60v can be run without asking the memory manufacture about the limitations, past that ASK.

 

IF the memory is rated 1.60-1.65 then a max of 1.70v can be run without asking the memory manufacture about the limitations, past that ASK.

 

 

If you question any of this, refer to the memory manufacture in their support forum for a direct answer to the max voltage their product will run. I have found many times memory rated for 1.55 will in fact run 1.65v with no issues and have verified that with the memory manufacture before increasing memory voltage on lower rated sticks. You must have the manufacture define that limit to be safe.

 

 

 

That being said the next 3 voltages deal with CPU IMC stability with Haswell

 

These two are the primary with memory clocking:

 

CPU ANALOG I/O VOLTAGE OFFSET  max of .200

CPU DIGITAL I/O VOLTAGE OFFSET  max of .200

 

We try to keep Digital .050v higher than Analog but that is not always possible.

 

 

This one is secondary:

 

CPU SYSTEM AGENT VOLTAGE OFFSET max of 1.196-1.20 as read in HW Monitor under a full load.

 

 

Between the DRAM voltage, Analog/Digital and the System Agent we must find out if the stability issue in your memory test failure is related to voltage, or, the memory itself. We will start with the DRAM Voltage.

 

Enter your BIOS. Under the AI Tweaker menu scroll down to the setting DRAM VOLTAGE. This was set by the XMP profile and now you are going to tweak this manually. NOTE: If for any reason you happen to reset XMP in the BIOS this voltage will revert to its XMP defined state.

 

Look at your currently set DRAM VOLTAGE:

 

a. If you are using 1.50-1.55v memory and the memory manufacture has specified that 1.60-1.65v is safe, set your DRAM VOLTAGE to the limit the manufacture has defined and then retest using the PHASE 1 outline above. Typically that will be 1.60-1.65v max.

 

If you have not discussed the limits with the memory manufacture you can follow this outline...

 

b. If you are running 1.50v DRAM VOLTAGE, raise it to 1.55v and retest using the PHASE 1 outline above.

 

 

c. If you are running 1.55v DRAM VOLTAGE, raise it to 1.60v and retest using the PHASE 1 outline above.

 

 

d. If you are running 1.60v DRAM VOLTAGE, raise it to 1.65v and retest using the PHASE 1 outline above.

 

 

e. If you are running 1.65v DRAM VOLTAGE, raise it to 1.67-1.68v and retest using the PHASE 1 outline above.

 

 

ALL TESTS PASSED: Go to TEST PHASE 2 below.

 

 

ONE OR MORE TESTS FAIL:

 

NOTE: If you are using 1.50-1.55v memory and at this point have not discussed the memory voltage limits with the memory manufacture I would do so now. If they specify the memory can run 1.60-1.65 then use that voltage and rerun the Phase 1 test.

 

 

If you have failed the test with the highest safe DRAM voltage possible then leave that DRAM voltage set in the BIOS and the next step is to raise the Analog/Digital voltage offsets. Be aware of the limits I defined. For test purposes we will raise the A/D voltages to their limits in order to quickly define if they are the issue.

 

 

Enter the BIOS and under the AI Tweaker menu scroll down and record the current Analog and Digital voltages that you worked with and that passed the 105 OCCT:CPU tests. These will be your LOWER end voltages.

 

Then set both A/D voltages as defined:

 

CPU ANALOG I/O VOLTAGE OFFSET:  .200

CPU DIGITAL I/O VOLTAGE OFFSET:  .200

 

NOTE: If you fail at these MAX voltage settings then the A/D voltage is now out of the equation. Reset the A/D values back to what you had originally and then proceed to change System Agent as defined below.

 

 

Rerun Test Phase 1 and if you pass repeat lowering voltages to A/D using this outline as defined until you reach your original starting point:

 

CPU ANALOG I/O VOLTAGE OFFSET:  .150

CPU DIGITAL I/O VOLTAGE OFFSET:  .200

 

CPU ANALOG I/O VOLTAGE OFFSET:  .100

CPU DIGITAL I/O VOLTAGE OFFSET:  .150

 

CPU ANALOG I/O VOLTAGE OFFSET:  .050

CPU DIGITAL I/O VOLTAGE OFFSET:  .100

 

 

Once you FAIL reset the values to the next level UP and then retest with Phase 1. If you PASS then lock the voltages there and then go to TEST PHASE 2 below.

 

 

If you have failed at the highest MAX voltage then the next step is to trim up System Agent.

 

NOTE: If you are already running the MAXIMUM safe limit for System Agent (1.196-1.20 as displayed in HW Monitor under a load) you can not raise the offset voltage any higher. This would mean the issue is directly related to either CPU Cache voltage (you can try a CPU Cache Voltage Override of 1.31v and retest), or, your clock is simply not stable with the memory you use, or, your memory has issues be it a defect or the quality of the modules.

 

 

Enter the BIOS and under the AI Tweaker menu scroll down, record your current System Agent voltage and then raise:

 

 

CPU SYSTEM AGENT VOLTAGE OFFSET:   .010v  Example: If your current offset voltage is .280, raise it to .290 and retest Phase1 however FIRST run HW Monitor and check the MAX voltage with a simple 2 minute run of OCCT:Linpack AVX to assure the voltage is not exceeding 1.20v.  The memory test will not invoke the full voltage override. You must use OCCT (Or AIDA64 FPU) to run a quick 2 minute check to confirm.

 

Continue to raise System Agent by .010v until you either PASS or hit the 1.196-1.20 limit in HW Monitor.

 

If you PASS then lock the voltages there and then go to TEST PHASE 2 below.

 

 

If you FAIL and you have reached the VCCSA voltage limit, then:

 

 

 

----- IF you are running the manufacture speed/timing of the memory and have not clocked the memory in any way:

I would STOP all clocking, reset the BIOS back to default, reset the BIOS under he AI Tweaker menu to XMP and run MEMTEST in a clean boot environment ONE stick at a time (repeat the tests you ran before installing Windows) to CONFIRM you do not have a defective memory module that has shown its ugly head. Most likely this is the cause of the issue. It is rare but does happen. If you find a FAIL and a defective stick, replace the memory, rerun the Memtest checks and then you can reset your clock back up and start over at PHASE 1 in this memory section.

 

 

It is however possible you simply have a weak memory module or poorly designed memory modules. If you PASS Memtest in a non-clocked state but can not pass PHASE 1 of this memory outline then you will either need to:

 

a. Contact the memory manufacture and RMA the modules since you are NOT clocking them and they simply wont run a CPU clock, stable.

 

 

b. If you happen to be using cheaper memory, then upgrade the modules to a different manufacture and better grade of product.  

 

c. LOWER your CPU cache clock speed by 100MHz or 1x lower in cache multiplier. If you do this you should start over using the next lowest CPU/Cache speed profile as outlined under sections 103 and/or 105. Lowering the cache means lower voltages and cooler temps.    

 

 

 

 

 

------ IF you are running outside the manufacture speed/timing of the memory and are working a clocked memory setup:

Then you will simply need to reset the memory back to manufacture default specification for speed/timing and the voltage that passed the Phase 1 memory test.

 

Your memory simply will not clock any higher or run lower latency.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

TEST PHASE 2:

 

BASIC MEMORY TWEAKING AND VERIFICATION

 

 
  

TESTING SOFTWARE TO DEFINE MEMORY PERFORMANCE:

Use of AIDA64 Cache and Memory Benchmark will define the latency changes when memory speed is increased or memory timing is lowered.

To access this test in AIDA64, in the top menu click: TOOLS then select: Cache and Memory Benchmark. Once the test box appears, click the Start Benchmark button. Do not touch the system while the test is running.

The test result we are interested in is the MEMORY LATENCY. This value is located at the end of the top line in the AIDA64 Cache and Memory Benchmark readout as shown below as an example from my Haswell system

 

The READ-WRITE-COPY values are important but it is the LATENCY that defines primary memory performance with respect to the memory speeds. The higher the READ-WRITE-COPY and the lower the LATENCY, the faster the memory is communicating with the CPU. However without understanding the engineering behind memory and motherboards one could assume that a READ-WRITE-COPY speed of a few thousand/hundred less with a exceptionally low memory LATENCY would be bad, this can be false and misleading.

Although if extremely low memory speeds are seen that is of course not good and would most likely indicate a problem, it is the LATENCY value that drives performance and therefore you can easily trade some memory speed for lower latency and come out on top.

The CPU must be properly clocked first in order to use the list below and define where the system sits for memory communication performance. The CPU cache latency is also important. Small drops in CPU cache latency represent large speed increases with respect to data being processed.

 
 
To simplify all this, the primary value we are interested in is the memory latency result. With a good CPU clock, a latency value of:

Above 45ns+ (especially 50ns or higher) = Room for improvement

43-45ns = Not bad at all, nice setup

41-43ns = Now we are talking! Excellent performance

38-41ns = Click Here  (usually comes with the higher speed/lower timing memory product and high CPU clocks) This is the area I personally strive to achieve but its not a requirement.

<38ns = Holy Whitewalls Batman! This is not easy to obtain without serious stress and voltage even with modern hardware but it is coming soon.

The values outlined above are not just defined by them memory speed/timing, but by the CPU/CACHE speeds as well. The list provided above is based on a system being clocked between 4.2-4.8GHz+ with fast high quality memory installed. The faster the CPU in relation to the memory speed and timing, the lower the latency. It is a combination of all factors that define the latency, not just memory.

The memory purchase itself puts the finishing touches on the clock. That is why with Haswell running DDR3 2400 9-11-11 memory @1T with a well clocked CPU is better than DDR3 2133 9-10-11 (or lower speed memory)
 
 

If you are running a latency of 41-45ns you have a great setup and don't 'obsess' over lowering it. 41-43ns is an exceptional result. With lower CPU clocks a latency of 44-46 is not bad either. Of course you can always see how low you can run stable by applying memory tweaks.

 
 
A few items you should understand about the test..

a. Running older versions of the benchmark on newer chipsets can display erratic or false results. Make sure the chipset and CPU platforms are supported by AIDA64. i.e.; Running the older version of this test on SB/IB platforms some time ago resulted in excessively low and false memory latency results. (well under 38ns). The latest version of AIDA64 will cover all the older platforms.

b. Quad channel memory platforms can display higher memory latency than triple or dual channel but also typically display a higher memory read/write/copy speed. This is normal since the platform will increase latency for the extended memory channel support and at the same time compensate with higher real word memory speed results. With quad systems latency values of 3-4, sometimes 5ns, higher than shown above are normal for each line.

 

This test can be used at any time to define memory performance and changes that may have occurred with upgrades to memory product as well as clock tuning.

DO NOTE: This PHASE 2 section and the next section PHASE 3 are STRICTLY OPTIONAL. They are NOT a requirement for your clock however every step taken to improve system performance is another step up in system response to Flight Simulator.

 

You can at this time opt to stop where you are and call it a clock. If you have followed my outlines then you have established a clock as well as verified the memory is running stable in the clock at its manufacture rated speed and timing. At this point you can consider yourself about 75-85% stable and I discuss what I mean by that in the section; Haswell and Overclocking 107: Am I Finished and Stable?.

If you decide to proceed, do make sure that you have a secure backup of your Windows drive before performing any memory clocking. Although it is possible to see Windows corruption clocking the CPU, when clocking the memory corruption is far more possible.

 

Typically many systems can obtain a stable tweak of the memory as outlined here in PHASE 2 but it is far more difficult and sometimes impossible to obtain the memory tweaks outlined in PHASE 3.

 

-----------------------------------------------------------------------

 

 

 A SIMPLE MEMORY TWEAK -

In this next phase I am going to outline a very simple memory tweak that just about everyone can use. Success DEPENDS ON THE MEMORY MANUFACTURE QUALITY and not so much the voltages involved. If this can be done it will be obvious very quickly. If not, then you may not be able to obtain this tweak and must simply accept what your memory will allow.

 

This change is done through the setting DRAM COMMAND MODE. Many memory manufactures will set DRAM COMMAND RATE to 2 or AUTO though XMP. Most memory modules are designed to run DRAM COMMAND MODE (AKA: COMMAND RATE) of '2' but the modules are able to run 1 which is faster. Some high performance memory is made to run 1. If DRAM COMMAND MODE is already set to 1 then you are already running this tweak and can stop here.

In Windows you can see the current system DRAM COMMAND RATE in CPUz under the memory tab as shown here:

 
 
 
 
 
 
CHANGE DRAM COMMAND RATE:

1. Enter the BIOS and go to the AI Tweaker menu. Scroll down to: DRAM TIMING CONTROL and enter that menu section.

2. Look for the setting: DRAM COMMAND MODE. It is usually found in the top section at the end of the fist timing list. Record the current value listed in case you need to set it back to the XMP value.  

 
3. Change DRAM COMMAND MODE to: 1 and hit enter

4. Hit F-10 save and exit and reboot into Windows

 
 
-IF YOU CRASH BOOTING INTO WINDOWS:

Noting the LIMITATIONS placed on DRAM VOLTAGE by the manufacture and what I have posted above in PHASE1 under: -THE VOLTAGE CHANGES IF TEST FAILURES APPEAR-

a. Try raising the DRAM VOLTAGE under the AI Tweaker slightly (about.02v)

in example:

If your current DRAM VOLTAGE is set to 1.60, raise it to 1.62

If your current DRAM VOLTAGE is set to 1.65, raise it to 1.67

If you continue to crash booting into Windows, AND, you have room to increase SYSTEM AGENT OFFSET and CPU ANALOG/DIGITAL I/O VOLTAGE OFFSET you can try raising them slightly as defined in the PHASE 1 test under the section: ONE OR MORE TESTS FAIL

IF none of the above allows a stable boot into Windows then your memory simply WILL NOT allow this change. Reset the voltages back to what you tested stable and reset DRAM COMMAND MODE back to the XMP value you wrote down. You are finished and can proceed to the next section.

 

 

-IF YOU BOOT INTO WINDOWS WITHOUT CRASHING:

Once rebooted and assuming the system did not crash rebooting into Windows, run the PHASE 1 test again as outlined above.

 
IF YOU FAIL THE PHASE 1 TEST then follow the outline above: IF YOU CRASH BOOTING INTO WINDOWS until you pass. Make sure you are not exceeding any of the outlined voltages.

IF YOU PASS THE PHASE 1 TEST then you will perform one last OCCT:CPU test run. Set OCCT:CPU to run the 64bit Large Data Set test for 1 hour.

 
If you FAIL OCCT:CPU then follow the IF YOU CRASH BOOTING INTO WINDOWS outline until you pass. If you can not pass OCCT:CPU with safe voltages then your memory simply WILL NOT allow this change. Reset the voltages back to what you tested stable and reset the DRAM COMMAND MODE back to the XMP value you wrote down.

 

IF YOU PASS OCCT:CPU then you have established the 1T memory timing clock tweak and are finished with this section.

 

 

 
 
 
 
 
 

TEST PHASE 3:

 

ADVANCED MEMORY TWEAKING 

 

This section is not for typical users. If you are not experienced with computers and clocking I suggest this section be skipped and move to the next. If you have made it this far and passed all the tests then you have established a great clock as-is.

For those who may wish to tinker further you may continue below. Even experienced users may find after frustrating hours of testing memory speed/timing changes that higher/stable memory performance is simply not possible. This takes a lot of patience to accomplish and may net zero results. Memory clocking can also introduce intermittent stability issues under real world stress loads that can be hard to trim out.

You are welcome to review and/or use this section, however most users should simply move to the next section and accept the current clock that has been obtained.

If you FAILED the PHASE 2 setup and test then you may have little or no success working this section. It is possible you may be able to trim in a memory speed increase, possibly a drop in the tRAS but failure with PHASE 2 typically indicates the memory simply has no room for tuning.   

Depending on the quality of the memory it can be possible to tweak the speed or primary timing values to obtain higher memory performance. Professional clockers will typically exceed manufacture specifications for voltages to obtain such memory clocks. This is not a professional overclocking site and as such I will not be defining dangerous voltage settings but will outline how to go about this type of memory clocking and testing.

Success with this is rare and not the norm and it takes time and patience to see if it can be accomplished, or not.

If you wish to work with testing advanced memory timing and speed changes past the manufacture specification DO make sure you have a full backup of the Windows install in case of disaster. 

 

OUTLINE OF WHAT WE WORK WITH:

Memory is specified in speed and timing. The first number, in example is DDR3 <speed> the values that follow are the timing values:

DDR3 <speed> TIMING VALUES = CAS Latency - RAS to CAS DELAY - RAS Precharge - Cycle Time

Example: DDR3 2400 C9 -11-11-31

When checking to see if it is possible to get more performance out of the memory we purchase the following settings are the focus:

MEMORY SPEED -----> Higher is faster. When working with memory speed we leave the timing values (outlined below) alone and focus on speed changes directly. When working with memory speed if we have changed the DRAM TIMING CONTROL or DRAM COMMAND RATE to 1 as outlined under PHASE 2 above, this is returned to the manufacture set XMP setting, usually 2, before testing for memory speed changes.
 

In CPUz our real time memory speed is displayed under the Memory tab. Remember that the value displayed is 1/2 the DDR3 rating so you must multiply the value shown x2.

 

TIMING VALUES; The FAB-4:

The values shown here can be seen in real time under the memory tab in CPUz

1. CAS Latency (CL)  -----> Usually impossible to change without serious voltage or a serious drop in memory speed. This setting will be left alone as it is exceptionally rare that it can be lowered.
 
This is the hardest number to change and remain stable, usually this must remain the same but it also has the highest impact of performance. 1 value lower at the same speed is a massive jump in performance.


2. RAS to CAS (tRCD) -----> 2nd most difficult to lower


3. RAS Precharge (tRP) -----> 3rd most difficult to lower
 
IF there is any room to work it is usually in these 2 values. Either one will make a difference, the first of the two typically having the most impact but also the most difficult to change and remain stable. Changing these (one or both) is usually where trimming occurs if possible.


 
4. Row Active Time (tRAS) -----> usually this can be lowered but has the least impact on performance. 
 

 
As defined in PHASE 2;
Command Rate (CR)-----> must be tested to see but with good memory this can usually be run @ 1T as was outlined under PHASE 2 but when we are working with advanced clocking this is always set back to the XMP value before attempting such advanced tweak changes. You can obtain much higher performance by lowering the FAB-4 or by increasing the memory speed and leave the Command Rate 2t. Sometime memory modules will allow BOTH a speed increase AND a lower Command Rate. This must all be tested through trial and error.
 

There is one other timing control that can influence performance. Its listed in CPUz under the Memory tab as: Row Refresh Cycle Time (tRFC) This value can have more influence on performance than tRAS.


In all cases, when it comes to memory TIMING lower is faster for any value.

First you must decide what you are going to try and tweak.. speed or timing. The higher the memory SPEED with the XMP manufacture timing, the faster the system memory. The LOWER the memory timing at the manufacture XMP speed, the faster the system memory. Each can provide performance enhancements.

 
Memory SPEED changes can be more difficult to work out simply because of the divider math involved. You can not define specific single digit memory speeds in a BIOS. The divider must work with the CPU speed. In other words, you cant bump memory speed in small values (1-5-10MHz) and must work with the divider math the motherboard allows. In some cases a professional clocker will take a loss in CPU speed to obtain a large increase in memory speed. Unless we are talking about a 235MHz or greater increase in memory speed, reducing CPU speed to get a small bump in memory speed is not worth the real world performance. Do keep that in mind.

Lower memory timing can net better results when memory speed increases Lower memory timing can net better results when memory speed increases are either not possible or only small bumps can be obtained.

Memory timing changes are easier to work with since the memory speed is locked at a established constant and its simply a matter of testing individual incremental lower values using trail and error.

 
 
As stated above, the last 3 primary timing values are the ones most used to find out if such tweaking is possible. CAS is usually impossible to change unless the memory speed is significantly lowered, which we don't want.
 
 
 
 
IN PROGRESS........  
 
 
 
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 13503
Direct Link To This Post Posted: October-12-2013 at 9:42am
 
 
 
 
 
THIS POST IS NOW COMPLETE....
 
 
 
 
 
 
 

Haswell and Overclocking 107:

Am I Finished and Stable?

 
 
 

If you have gotten this far then of course congratulations are in order! The question now is: Is your system stable?

 
Maybe!  Possibly NOT!
 

In the FSX bible I specifically warned that installing Windows and loading the initial drivers and software in a clocked state is UNWISE and should never be done. I also warned that installing large amounts of software into a new highly clocked system can present issues down the road you cant see now but end up pulling your hair out trying to find a problem later.

There is a reason for my assessments and recommendations. With a new system and a new clock setup you are probably not 100% stable and simply don't realize it. People follow a guide, run stability tests and ASSUME too much as well as take very BAD advice on the internet. That goes for the guide I have outlined here.

I stated earlier in the thread: THERE IS NO SUCH THING AS 100% STABLE when we are talking high-end clocking. We can establish a setup that has a 'HIGH PROBABILITY' for being very stable but the fact remains that the system is overclocked, factors change over time, and, as such we are running outside of any engineering specifications that define lab tested stability.

 
THEREFORE: Until such time as we have:

A: Load tested a system for days (usually 2 or more) without stopping the load test and have passed.

And/Or

B: Worked with the system for weeks, usually a month or more, and have found NO crashes or stability issues related to being overclocked

NEVER ASSUME THE SYSTEM STABILITY LEVEL.

Keeping consistently updated image backups during the initial run time with a new system and clock is a very good way to avoid disasters.

Once 'B' has been accomplished, then the 'assumption' of system stability becomes relevant, not until. It is at that point where we can then do things such as uninstall/install large amounts of data, perform system maintenance and other tasks while the system is clocked, but even then under no circumstances do we ever install Windows and initial drivers/software in a clocked state.

 
Further...

Lets assume we have passed many weeks of use with no issues, is everything stable?  Well, environmental factors can and do CHANGE during the course of time. Dust in the system, increases in ambient temperatures during the year, hardware can be stressed and weakened whereby a module or a chip could start to display signs of stability issues after being clocked for many months.

 
So I will repeat this again:

THERE IS NO SUCH THING AS 100% STABILITY WHEN WE CLOCK A SYSTEM. PLEASE REMEMBER THAT. I hope I have made this point very clear.

 
 
Over time even if we do not see issues the system should be cleaned as well as seasonal load test checks should be performed to confirm ambient environment temps have not altered the result. Even with stable temperatures periodic load test should be performed to confirm our hardware remains stable and break-in or burn-in periods have not altered the end result.

 

That being said, with a new system and a new clock setup it is critical to adhere to the warnings about stability and system use. You may very well find that even after every temp/load test has PASSED that the addition of PCIe 3D render loads MAY display minor stability issues... case in point: http://www.simforums.com/forums/bsod-when-flyinghow-bad-is-that_topic47467.html

 

 

REAL WORLD USE SYSTEM CLOCK RELATED CRASHES:

With a new system and new clock setup established if you find during render or other real world use that stability issues exist, (BSOD, CTD, lockups that you have confirmed are clock related) then small increases to Vcore (CPU VOLTAGE) or System Agent (VCCSA) may need to be applied. This is the final step to establishing a stable clock setup. Add a minor increase to Vcore and/or System Agent offset and then fly it again.

Many may never see any issues at all, most wont, but please if you see no issues right away DO NOT ASSUME stability. You have a test factor of 'time' that must pass before that assumption becomes valid.
 
 
OPTIONAL: If you wish to run extended time OCCT:CPU load tests on your system (a day or more without stopping) you can opt to do that. Simply set OCCT:CPU to run in 'INFINITE" mode instead of automatically timed for 1hr. Running such extended tests can provide a higher probability in system stability. I will sometimes run a 24hr load test, not always, but sometimes do that myself. If failures are seen then the same process that was outlined in the thread above for each clock speed outline is used to trim in long term stability. Usually a simple bump in Vcore or VCCSA does the trick.

Even with success in extended CPU/memory load tests you must remember that the addition of PCIe loads can still display minor lapses in stability which is why it is very important to not assume system stability until the system has been used for 3D render over several weeks without any signs of clock issues.

 

 

So please do remember that until your system has been tested in combat over a reasonable period of time you should avoid performing very large installations and major maintenance while clocked. This is why I outlined that all software be installed and that all maintenance is performed prior to beginning a clocking session and system use. During the initial test period if maintenance needs to be performed use your 'save BIOS' profile system to switch back and forth. Once you get past the initial testing period then you may work with the system as you normally would but in the clocked state.

 
 
OPTIONAL WHEN STARTING OUT FOR THE FIRST TIME:

You can of course opt to setup your clock before installing software after Windows/drivers/hardware has been fully installed and then record your BIOS clock settings and return the system to normal operation in order to complete the installation of software and perform cleanup maintenance. 

 
So you do have options when installing and setting up a new system. Since most people tend to be a bit overzealous with wanting to get their system running and flying as quickly as possible I outlined that the system be installed with all software first, then clock, verify with load testing, and then test your results under flight conditions and make changes as needed. Either method is fine, the important factor is we don't want any file corruption in the mix and we want to establish the system has indeed passed and established a 'true' high probability for stability.
 
 

 

Once you have established a stable clocked system over a proper test period of time you can consider yourself a graduate of the Haswell school of clocking. You have proudly earned your Bachelor of Science in overclocking but don't get cocky! LOL .... A masters and a PhD requires far more technical knowledge and experience to obtain.

 

 

 

 

 

The next section of this thread: Haswell and Overclocking: Modifying The Processor For Higher Clock Speeds will deal with the Haswell IHS modification. You can obtain a higher clock by modifying the far less than adequate thermal connection between the inside of the CPU cap and the CPU chip itself. Such modifications come with risks. Although not hard to accomplish for those with a steady hand and a few simple tools, you must also be willing to accept the risks that come with that process.  

 

 

 

Back to Top
 Post Reply Post Reply
  Share Topic   

Forum Jump Forum Permissions View Drop Down

Forum Software by Web Wiz Forums® version 10.14
Copyright ©2001-2012 Web Wiz Ltd.

This page was generated in 0.266 seconds.