SimForums.com Homepage
Forum Home Forum Home > General Discussion Forums > Hardware, Software, and Computer Technology
  New Posts New Posts RSS Feed - 3DMark06
  FAQ FAQ  Forum Search   Register Register  Login Login

3DMark06

 Post Reply Post Reply Page  123 5>
Author
Message
Fly happy View Drop Down
Senior Member
Senior Member


Joined: October-10-2012
Location: Sweden
Points: 871
Post Options Post Options   Thanks (0) Thanks(0)   Quote Fly happy Quote  Post ReplyReply Direct Link To This Post Topic: 3DMark06
    Posted: July-06-2013 at 10:24am
Hi,
 
I have downloaded 3DMark06 but not installed it yet.
It's a 584MB file currently on one of my storage drives, a WD Caviar Green 1TB (5400RPM).
 
I have a SSD for the operating system, and another SSD for FSX.
 
So, does disk speed matter when you run this benchmark?
I will install it on the OS SSD if that is the case, otherwise the WD Caviar Green.

Kind regards
Hans
 
Edit: 3dMark06 instructions: Turn off hyperthreading in BIOS. Before you start 3dMark06 for the first time, start Nvidia Control Panel or Nvidia Inspector if you use that to control the 3DMark06 profile, and set Power Management Mode: Prefer Maximum Performance and Vertical Sync: FORCE OFF. Don't forget to apply the changes. Reboot after each 3dMark06 run. 
Hans

W7/64 Ultimate, FSX Gold, SB-E 3930K @ 4.7, Sabertooth X79, GTX580, 4x2GB G.Skill RipjawsZ @ 2133-9-11-10-28-1T, Corsair H110+Obsidian 900D,Seasonic P-1000, GEX, UTX, UT2, REX, S-Tech LC, NGX.
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-06-2013 at 10:29am
no..  disk speed does not matter as it completely loads the test render before beginning the test
 
that is another reason why we cant benchmark FSX.. FSX loads and accesses data while flying, a benchmark loads with a direct full package of all information to be rendered before running, or, it accesses a file before running the test.. there is no 'on the fly' loading involved during a test.
 
It can be installed anywhere
Back to Top
Fly happy View Drop Down
Senior Member
Senior Member


Joined: October-10-2012
Location: Sweden
Points: 871
Post Options Post Options   Thanks (0) Thanks(0)   Quote Fly happy Quote  Post ReplyReply Direct Link To This Post Posted: July-06-2013 at 11:20am
Great, thanks. That was the answer I was hoping for.
Kind regards, 
Hans

Hans

W7/64 Ultimate, FSX Gold, SB-E 3930K @ 4.7, Sabertooth X79, GTX580, 4x2GB G.Skill RipjawsZ @ 2133-9-11-10-28-1T, Corsair H110+Obsidian 900D,Seasonic P-1000, GEX, UTX, UT2, REX, S-Tech LC, NGX.
Back to Top
Fly happy View Drop Down
Senior Member
Senior Member


Joined: October-10-2012
Location: Sweden
Points: 871
Post Options Post Options   Thanks (0) Thanks(0)   Quote Fly happy Quote  Post ReplyReply Direct Link To This Post Posted: July-07-2013 at 5:24pm
I have a problem with the downloaded file. It does not have the extension .exe so it can not be run. The name is just "3DMark06_v121_installer" and then blank.
 
I downloaded it twice from here with the same result:
Scroll down to: Download 3DMark
 and then click:
3DMark06
 
What do you think, is there a problem on my end or is the file corrupted?
Kind regards,
Hans
 
 
 
 
Hans

W7/64 Ultimate, FSX Gold, SB-E 3930K @ 4.7, Sabertooth X79, GTX580, 4x2GB G.Skill RipjawsZ @ 2133-9-11-10-28-1T, Corsair H110+Obsidian 900D,Seasonic P-1000, GEX, UTX, UT2, REX, S-Tech LC, NGX.
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-07-2013 at 5:50pm
 
 
 
edit the downloaded file name and add the .exe extension run it
 
 
probably something they did on their end to curtail scanners from flagging the file
Back to Top
Fly happy View Drop Down
Senior Member
Senior Member


Joined: October-10-2012
Location: Sweden
Points: 871
Post Options Post Options   Thanks (0) Thanks(0)   Quote Fly happy Quote  Post ReplyReply Direct Link To This Post Posted: July-08-2013 at 3:41pm
That did it. I installed it and then ran the benchmark without changing any settings.
 
Short version:
3DMark Score 29328.0
SM2.0 Score 10849.0
HDR/SM3.0 Score 14329.0
CPU Score 8465.0

This was with i7 3930K CPU at stock speed (max 3800 MHz) and the Corsair Vengeance memory (4x4GB 1866MHz 9-10-9-27-2T) and the GTX580.
 
When I saw the CPU test on screen I thought what is going on here...it slowed down so much I thought the benchmark was going to crash...but I guess this is how it runs?
I will run the benchmark again with overclocked CPU another time.
Kind regards,
 
Hans
Hans

W7/64 Ultimate, FSX Gold, SB-E 3930K @ 4.7, Sabertooth X79, GTX580, 4x2GB G.Skill RipjawsZ @ 2133-9-11-10-28-1T, Corsair H110+Obsidian 900D,Seasonic P-1000, GEX, UTX, UT2, REX, S-Tech LC, NGX.
Back to Top
Fly happy View Drop Down
Senior Member
Senior Member


Joined: October-10-2012
Location: Sweden
Points: 871
Post Options Post Options   Thanks (0) Thanks(0)   Quote Fly happy Quote  Post ReplyReply Direct Link To This Post Posted: July-08-2013 at 4:23pm
I made a 4400 MHz run as well. Couldn't resist. Just had to know, even though is way past bed time here Smile
 
Short version:
3DMark Score 32954.0
SM2.0 Score 12572.0
HDR/SM3.0 Score 15713.0
CPU Score 9517.0
 
I will make another run with the new memory Gskill 4x2GB RipjawsZ 2133 Mhz 9-11-10-28-2T another time.
 
Hans
 
Hans

W7/64 Ultimate, FSX Gold, SB-E 3930K @ 4.7, Sabertooth X79, GTX580, 4x2GB G.Skill RipjawsZ @ 2133-9-11-10-28-1T, Corsair H110+Obsidian 900D,Seasonic P-1000, GEX, UTX, UT2, REX, S-Tech LC, NGX.
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-08-2013 at 5:25pm
Yes
 
CPU test is extremely intensive..  each 1/2 FPS gained in that is a large leap
Back to Top
Ted striker View Drop Down
Senior Member
Senior Member


Joined: January-24-2006
Location: Denver
Points: 170
Post Options Post Options   Thanks (0) Thanks(0)   Quote Ted striker Quote  Post ReplyReply Direct Link To This Post Posted: July-09-2013 at 12:04am
My scores are much lower. I hope its just my 285gtx holding me back and not something else...

3DMark Score 23894
SM2.0 Score 9531
HDR/SM3.0 Score 10130
CPU Score 8259

Ted
3770K @4.5Ghz, Noctua NH-C12P, Asus Z77-V Deluxe, Corsair 2133-9-11-10-27-2T, 780 GTX, Win7-64 on 256gb Intel SSD, FSX, P3Dv3, P3Dv4 on 500gb Intel SSD, PC Power & Cooling 750W, Antec P193 case
Back to Top
dmz0427 View Drop Down
Senior Member
Senior Member


Joined: June-13-2013
Location: New York
Points: 982
Post Options Post Options   Thanks (0) Thanks(0)   Quote dmz0427 Quote  Post ReplyReply Direct Link To This Post Posted: July-09-2013 at 12:16am
Ted
You are trying to compare your 4 core to a 6 core ,, of course your score is going to be lower on the CPU Parts and over all score
as far as your Gpu/Video yes that 285 is holding you way back and holding your sys back on FSX too,
even if you put a 780 in your sys you will not score quiet as high as that six core, unless you have a drink or 2 ,, and clock the living hell out of your cpu    I can say a 780 would help you a good 10-15% not 600-700 dollars worth , but your sys would run smoother and fsx should too as long as you don't dump a mega load of scenery to it
hope this helps
Back to Top
Great Ozzie View Drop Down
New Member
New Member


Joined: February-12-2008
Location: United States
Points: 14
Post Options Post Options   Thanks (0) Thanks(0)   Quote Great Ozzie Quote  Post ReplyReply Direct Link To This Post Posted: July-09-2013 at 3:28am
Originally posted by Ted striker Ted striker wrote:

My scores are much lower. I hope its just my 285gtx holding me back and not something else...


Yeah... I'd say it is based on my results.  We have similar systems (I'm oc'd @ 4.8) using a GTX680.

3DMark Score 36187
SM2.0 Score 14393
HDR/SM3.0 Score 18345
CPU Score 8742
http://www.3dmark.com/3dm06/17318473

Rob

Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-09-2013 at 6:56am
 
 
 
 
I want everyone to remember something here.. when you look at a 3DM06 score you can see where something is holding you back and in the case of Ted, it is in fact his video card..
 
His CPU score is not that bad.. its the shader scores that are low and remember that the CPU does play into those scores just like it would in FSX. His CPU is bottlenecking the video card.. that is how you can see where the card and CPU play hand in hand.
 
But when you are looking at 'Does this make FSX run better" BE VERY CAREFUL about comparing YOUR score to someone elses. The goal is to compare YOUR score to YOUR score after a change in YOUR system, not so much look at others.
 
For a apples to apples compare you are looking for 800-1000pt increment increases on your OWN SYSTEM in which each one you see defines your FSX should run better
 
and I also want to point out that 1000 in the total 3DMark Score is not a 'huge' leap..   3000 and above would be
 
My last run was posted here after adjusting Northbridge clocks and shaving latency off the memory about 1ns  (40.7ns to 39.6ns).. score increased by about 300-500pts
 
 
3DMark Score 40887.0
SM2.0 Score 16275.0
HDR/SM3.0 Score 20811.0
CPU Score 9801.0
 
 
 
that much change with latency shaving ONLY works if you are already screaming at the memory level to begin with. In other words, lets say your memory checks to be running about 52ns and you manage to shave it down to 48ns.. that is ALWAYS good and your 3DM scores will go up but wont jump like mine did. You have to be at the lower end of 40 and reduce with a higher NB frequency before you start seeing the changes I display and since I am at my peak I am also approaching the instability point where I can not shave any more latency without seriously jacking voltages.
 
 
Haswell is also a memory whore.. the faster the memory and the lower the latency, the better the result with Haswell. I bet IB is the same in many respects.
 
 
 
 EDIT: Oppps!  Guys that last run I did and posted was a test run while I was checking for stability in HW NB clocking with Hyper Threading ENABLED so it is NOT VALID with respect to FSX. FSX wont use HT at all or in any way but 3DM will .. just wanted to be sure everyone understands that. Without HT that score would be closer to 40K. You do need to be sure HT is DISABLED in the bios when checking for 3DM-to-FSX compares. You can tell HT is enabled on someone's system with the scores by looking at the PROCESSOR information: Physical / logical processors  1 / 8 <---  8 threads for a quad core = HT ENABLED
 
 The latency assessment is still valid since I was comparing 2 runs with HT ENABLED but at different UCLK frequencies.
 
 
Back to Top
Ted striker View Drop Down
Senior Member
Senior Member


Joined: January-24-2006
Location: Denver
Points: 170
Post Options Post Options   Thanks (0) Thanks(0)   Quote Ted striker Quote  Post ReplyReply Direct Link To This Post Posted: July-09-2013 at 10:16am
Appreciate the feedback guys. Looks like its time for me to get a 780gtx.  I was pretty pleased with my rig coming from a Q6600, but after seeing your scores I can see there is a bit more potential to take advantage of.

Ted
3770K @4.5Ghz, Noctua NH-C12P, Asus Z77-V Deluxe, Corsair 2133-9-11-10-27-2T, 780 GTX, Win7-64 on 256gb Intel SSD, FSX, P3Dv3, P3Dv4 on 500gb Intel SSD, PC Power & Cooling 750W, Antec P193 case
Back to Top
robmw View Drop Down
Senior Member
Senior Member


Joined: April-02-2005
Location: United Kingdom
Points: 303
Post Options Post Options   Thanks (0) Thanks(0)   Quote robmw Quote  Post ReplyReply Direct Link To This Post Posted: July-09-2013 at 1:47pm
Nick,

Just for interest I ran 3DMark06 to compare with your old system a few days ago when you posted your results moving from the 980X to HW but keeping the GTX480. At 4.2GHz I got a score of 29318 using default settings with my combination of 980X / GTX 680 and 6-6-6-20 1600 memory, using the PCIe 2 board. Very similar raw numbers to yours bearing in mind that you were running slightly faster, with faster memory but a supposedly lesser GPU. The raw memory bandwidth of the 680 is actually a few percent less than the 580 I had before but I couldn't benchmark that as it's dead!

Shows the point you're making about the 680 though in that a card 2 generations along actually doesn't perform much better than the 480 in the performance terms relevant to FSX, particularly as the 680 I got is one of EVGA super clocked models. It's saving grace for me is that it does seem to handle higher levels of AA with less impact and it runs very cool.

When I upgrade my thermal compound to the CL Pro and change fans I'll post in the cooling thread as that might be of interest to show the relative benefit (or not) that I got but I'm aware I'm getting a bit off-topic here from the HW, 780 discussion.

And I'm sold on the benefits of upgrading at this point, no doubt, just can't do it at the moment! I will make sure that I don't leave it too long though so that parts are still available and the info you share here will still be current, that's my dilemma.
i7 4790K @4.4 GHz, Asus Z97 Deluxe, 8GB 2400 DDR3, EVGA GTX 1080Ti, Samsung SSDs 850 Pro + WD HD, Windows 7 Pro 64.
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-09-2013 at 2:49pm
Its the PCIe bus that holds the cards back but you see now how the raw draw power of the 480/580 is actually better than the 600 series which is why I never bothered, that and the fact that the 780 would actually be a upgrade as long as it was installed into a fully supported PCIe 3.0 system
Back to Top
Fly happy View Drop Down
Senior Member
Senior Member


Joined: October-10-2012
Location: Sweden
Points: 871
Post Options Post Options   Thanks (0) Thanks(0)   Quote Fly happy Quote  Post ReplyReply Direct Link To This Post Posted: July-10-2013 at 4:25am
It was great fun running the 3dMark06 benchmarks.
Now we have a good base to use if upgrading parts.
I found it hard to judge performance changes in FSX so this tool comes in handy.

Unfortunately I donít know how those different scores relate to FSX performance, but at my level maybe I don't really need to know...For instance, if I see an increase in SMR2.0, what can I expect to improve in FSX,
if I see an increase in HDR/SM3.0, what can I expect to improve in FSX and so onÖ 

And the CPU scoreÖI think we really can be fooled with that one if we look at other peoples scores and hardware.I get a descent CPU score with my 6 core SB-E at 4400MHz in 3DMark06 compared to a 4 core Haswell at 4800 MHz.

But for FSX use,  I would throw my CPU out the window for the Haswell since I know how FSX uses raw CPU power.Heck, I think the SB-E  would lose against a high clocked 4 core Ivy too in FSX.
But I donít care right now, I am DONE with hardware upgrades for this year and probably the next too.
With the GTX580 installed and the new Gskill quad kit (to be installed) I am at the point where I should have been when I got that rig in the first place. :-)

Hans
Hans

W7/64 Ultimate, FSX Gold, SB-E 3930K @ 4.7, Sabertooth X79, GTX580, 4x2GB G.Skill RipjawsZ @ 2133-9-11-10-28-1T, Corsair H110+Obsidian 900D,Seasonic P-1000, GEX, UTX, UT2, REX, S-Tech LC, NGX.
Back to Top
jcmmg View Drop Down
Senior Member
Senior Member


Joined: July-08-2010
Location: EGLL/EGPD/SVCS
Points: 120
Post Options Post Options   Thanks (0) Thanks(0)   Quote jcmmg Quote  Post ReplyReply Direct Link To This Post Posted: July-10-2013 at 7:49am
Hi Hans,

Having a similar rig as yours, find here my scores for different clock speeds. It is very interesting how memory and GPU can modify the results. About FSX, I'm very happy with the results, I get 70% ~30FPS the key is to find your PC balance... About Haswell, well my time for upgrade was last year, the only hope is that Haswell-E still be compatible to LGA2011... letís see how Ivy-E goes (It looks like is going to be a BIOS update only)! hehehehe



Test 2 4.7 but HT On: 36545 http://www.3dmark.com/3dm06/17319028

Test 3 4.8 but HT Off: 36692 http://www.3dmark.com/3dm06/17318948

Test 4 4.7 Final (Voltage adjustments) HT Off: 36128 http://www.3dmark.com/3dm06/17319044


Thanks for CLP I manage to establish an stable 4.7 otherwise I can't get further 4.6

Clear skies,

Julio

Note: I need to check RAM speed....

Mobo: Asus x79 RIVE|Intel i7 3930K OC 4.7 GHz|Memory: 16 GB Dominator 2133|EVGA Nvidia 780 Classified|FSX Gold|WIN7 64 Ultimate|Alienware ALX R4 Case
Back to Top
Fly happy View Drop Down
Senior Member
Senior Member


Joined: October-10-2012
Location: Sweden
Points: 871
Post Options Post Options   Thanks (0) Thanks(0)   Quote Fly happy Quote  Post ReplyReply Direct Link To This Post Posted: July-10-2013 at 8:11am
Julio, thanks for sharing. After I installed the H110 last month there is room for higher CPU clock but I have not gotten around to that yet...need to swap the memory with the new Gskill kit first and memcheck kit first but in time...I will try a higher cpu clock for sure and post results here.
Hans

W7/64 Ultimate, FSX Gold, SB-E 3930K @ 4.7, Sabertooth X79, GTX580, 4x2GB G.Skill RipjawsZ @ 2133-9-11-10-28-1T, Corsair H110+Obsidian 900D,Seasonic P-1000, GEX, UTX, UT2, REX, S-Tech LC, NGX.
Back to Top
Fly happy View Drop Down
Senior Member
Senior Member


Joined: October-10-2012
Location: Sweden
Points: 871
Post Options Post Options   Thanks (0) Thanks(0)   Quote Fly happy Quote  Post ReplyReply Direct Link To This Post Posted: July-10-2013 at 8:22am
BTW Julio, were you able to activate PCIE 3.0 on your system for the GTX680?
Hans

W7/64 Ultimate, FSX Gold, SB-E 3930K @ 4.7, Sabertooth X79, GTX580, 4x2GB G.Skill RipjawsZ @ 2133-9-11-10-28-1T, Corsair H110+Obsidian 900D,Seasonic P-1000, GEX, UTX, UT2, REX, S-Tech LC, NGX.
Back to Top
jcmmg View Drop Down
Senior Member
Senior Member


Joined: July-08-2010
Location: EGLL/EGPD/SVCS
Points: 120
Post Options Post Options   Thanks (0) Thanks(0)   Quote jcmmg Quote  Post ReplyReply Direct Link To This Post Posted: July-10-2013 at 4:38pm
Originally posted by Fly happy Fly happy wrote:

BTW Julio, were you able to activate PCIE 3.0 on your system for the GTX680?

Thanks a lot for your questions as I wasn't aware about this PCIE 3.0 issue!!


Easy: Download->Install (2sec)-> Re-start-> Ready to sim!!!

Now my PC is PCI-E 3.0x 16 @ 16 3.0 enable

And the results on 3DMark06 are:

Score:36421
SM2.0 Score: 13483
HDR/SM3.0 Score: 18068
CPU Score: 10162

Mobo: Asus x79 RIVE|Intel i7 3930K OC 4.7 GHz|Memory: 16 GB Dominator 2133|EVGA Nvidia 780 Classified|FSX Gold|WIN7 64 Ultimate|Alienware ALX R4 Case
Back to Top
Fly happy View Drop Down
Senior Member
Senior Member


Joined: October-10-2012
Location: Sweden
Points: 871
Post Options Post Options   Thanks (0) Thanks(0)   Quote Fly happy Quote  Post ReplyReply Direct Link To This Post Posted: July-11-2013 at 5:09am
Julio, so enabling PCI-E 3.0 increased score from  36128  to 36421. Good to know. 
If you have time, could you run a 4400Mhz benchmark?
It would be interesting to see your score at the same CPU speed as mine.
Hans
Hans

W7/64 Ultimate, FSX Gold, SB-E 3930K @ 4.7, Sabertooth X79, GTX580, 4x2GB G.Skill RipjawsZ @ 2133-9-11-10-28-1T, Corsair H110+Obsidian 900D,Seasonic P-1000, GEX, UTX, UT2, REX, S-Tech LC, NGX.
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-11-2013 at 10:21am
Kind of curious here..   someone with a SBE launch 3DMark06 and then lauch the task manager, right click the 3DM06.exe and select SET AFFINITY and then uncheck the last 2 cores, click OK, close task manager and then run the test again and post the results.
 
You don't have to upload the result.. I was simply curious to see the compare with a SB system that does appear to support PCIe3 with a 680 and 4 cores active for the benchmark and is well clocked,.. and no HT enabled.
 
 
Back to Top
jcmmg View Drop Down
Senior Member
Senior Member


Joined: July-08-2010
Location: EGLL/EGPD/SVCS
Points: 120
Post Options Post Options   Thanks (0) Thanks(0)   Quote jcmmg Quote  Post ReplyReply Direct Link To This Post Posted: July-11-2013 at 5:27pm
Hans,

These are the results of my system @ 44 and PCI-E 3.0

SCORE 33561
SM2.0 Score 12003
HDR/SM3.0 Score 16911
CPU Score 9558


Hope Nick will help to analyse these results.....
Mobo: Asus x79 RIVE|Intel i7 3930K OC 4.7 GHz|Memory: 16 GB Dominator 2133|EVGA Nvidia 780 Classified|FSX Gold|WIN7 64 Ultimate|Alienware ALX R4 Case
Back to Top
jcmmg View Drop Down
Senior Member
Senior Member


Joined: July-08-2010
Location: EGLL/EGPD/SVCS
Points: 120
Post Options Post Options   Thanks (0) Thanks(0)   Quote jcmmg Quote  Post ReplyReply Direct Link To This Post Posted: July-11-2013 at 5:49pm
Originally posted by NickN NickN wrote:

Kind of curious here..   someone with a SBE launch 3DMark06 and then lauch the task manager, right click the 3DM06.exe and select SET AFFINITY and then uncheck the last 2 cores, click OK, close task manager and then run the test again and post the results.
 
You don't have to upload the result.. I was simply curious to see the compare with a SB system that does appear to support PCIe3 with a 680 and 4 cores active for the benchmark and is well clocked,.. and no HT enabled.
 
 

Nick here you have it:

SCORE 34237
SM2.0 Score 13156
HDR/SM3.0 Score 17983
CPU Score 8140


My system @ 47, PCI-E 3.0 and Affinity 0 1 2 3

BTW: for some reason 3DMark06 is not detecting the true RAM speed at 1067...




Mobo: Asus x79 RIVE|Intel i7 3930K OC 4.7 GHz|Memory: 16 GB Dominator 2133|EVGA Nvidia 780 Classified|FSX Gold|WIN7 64 Ultimate|Alienware ALX R4 Case
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-11-2013 at 6:06pm
Ok great, now lets level the field
 
Post your full Nvidia card Core and Memory clock speeds, the clocks you ran the benchmark on. This should be the specification rating for the card when it is in full 3D mode
 
Post your system memory speed and timing
 
Using CPUz, click the Memory tab and tell me what: Northbridge Frequency reads.. 
 
You can get your memory information from that tab too and also include the FSB:DRAM Ratio
 
If you don't have CPUz, get it from here: http://www.cpuid.com/downloads/cpu-z/1.65-setup-en.exe  click the blue DOWNLOAD NOW button
 
 
 don't worry about 3DM detection. .it sometimes cant read squat LOL
 
Back to Top
jcmmg View Drop Down
Senior Member
Senior Member


Joined: July-08-2010
Location: EGLL/EGPD/SVCS
Points: 120
Post Options Post Options   Thanks (0) Thanks(0)   Quote jcmmg Quote  Post ReplyReply Direct Link To This Post Posted: July-11-2013 at 6:27pm







Sorry Nick, but I'm traveling tomorrow, I will catch up on monday!..
Mobo: Asus x79 RIVE|Intel i7 3930K OC 4.7 GHz|Memory: 16 GB Dominator 2133|EVGA Nvidia 780 Classified|FSX Gold|WIN7 64 Ultimate|Alienware ALX R4 Case
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-11-2013 at 7:11pm
Ahh that's right.. you guys run QPI link..   ok this works!
 
 
 
Back to Top
jcmmg View Drop Down
Senior Member
Senior Member


Joined: July-08-2010
Location: EGLL/EGPD/SVCS
Points: 120
Post Options Post Options   Thanks (0) Thanks(0)   Quote jcmmg Quote  Post ReplyReply Direct Link To This Post Posted: July-12-2013 at 7:02am
Hi all,

After analyzing my data product of my OC, I can't get around about what it looks like the 3930K cannot have a full performance of PCI-E 3.0, even with Nvidea ".exe", it is a strong processor but I can't reach SM2.0 Score to the same level as 3770K at similar clocks and similar GPU....

That is the difference between PCI-E 3.0 support and PCI-E 3.0 enable :S

What a disappointment....

Clear skies,

Jcmmg

Find here a summary of the results on this thread....


Mobo: Asus x79 RIVE|Intel i7 3930K OC 4.7 GHz|Memory: 16 GB Dominator 2133|EVGA Nvidia 780 Classified|FSX Gold|WIN7 64 Ultimate|Alienware ALX R4 Case
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-12-2013 at 7:15am
That is why I waited for Haswell and the motherboards that fully support and have all the 'kinks' worked out for PCIe 3.0
 
Going from a PCIe 2.0 980x/480 (or 580) system in which the processor is 32nm technology to a SB with no true PCIe 3.0 support made no sense, even with a 680 card. IB was my first choice but the 780 was not available, so I waited for Haswell and the 780 to make the switch
 
 
actually glad I did. The only downfall with the proc is the Intel IHS thermal issue.. if that is fixed nothing can beat it till the next round
 
 
EDIT: I intend to take your setup and neuter mine to the same settings and look at the compare. That's why I asked for your specs. That could answer a lot of questions about the differences and where they are.
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-12-2013 at 7:29am

I will be underclocking the 780 using estimated values to try and bring it closer to a 680 but there are factors I can not change such as the video card memory bandwidth which is far more important than people claim it is..   even Phil Taylor said that, and believe me he knows as he was one of the primary developers of DirectX and practically wrote the book on it.

Back to Top
jcmmg View Drop Down
Senior Member
Senior Member


Joined: July-08-2010
Location: EGLL/EGPD/SVCS
Points: 120
Post Options Post Options   Thanks (0) Thanks(0)   Quote jcmmg Quote  Post ReplyReply Direct Link To This Post Posted: July-12-2013 at 7:34am
Originally posted by NickN NickN wrote:

 
 
EDIT: I intend to take your setup and neuter mine to the same settings and look at the compare. That's why I asked for your specs. That could answer a lot of questions about the differences and where they are.


Looking forward to it, if you need more information, just let me know... 


Mobo: Asus x79 RIVE|Intel i7 3930K OC 4.7 GHz|Memory: 16 GB Dominator 2133|EVGA Nvidia 780 Classified|FSX Gold|WIN7 64 Ultimate|Alienware ALX R4 Case
Back to Top
Fly happy View Drop Down
Senior Member
Senior Member


Joined: October-10-2012
Location: Sweden
Points: 871
Post Options Post Options   Thanks (0) Thanks(0)   Quote Fly happy Quote  Post ReplyReply Direct Link To This Post Posted: July-12-2013 at 7:37am
Julio,  that summary is awesome! Thank you for taking the time to gather all data and present it like that. And Nick and everyone else that contributed to this, thank you. Most informative! 
Hans

W7/64 Ultimate, FSX Gold, SB-E 3930K @ 4.7, Sabertooth X79, GTX580, 4x2GB G.Skill RipjawsZ @ 2133-9-11-10-28-1T, Corsair H110+Obsidian 900D,Seasonic P-1000, GEX, UTX, UT2, REX, S-Tech LC, NGX.
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-13-2013 at 9:32am

 


Before I run this test, I wanted to post my absolute best "stable" benchmark values


Define "stable": This is where I have certified that my system will in fact full-load FPU stress test for a minumum of 1 hour with no issues of heat or instability, and pass.

It is important to note that its easy to fudge scores based on a user overclocking the snot out of their system to the point where it could not pass a SIMPLE load test for 20 minutes without blue-screens but will still run 3DMark. That is why you should never look at scores at Futuremark in their database and compare them to yours. That database is full of users who use hacked drivers, clocked systems that are not stable, overload video cards and processors until the are damaged or nearly damaged just to post the highest score possible.


Ok with that out of the way, these two test results represent my final 'best' clock possible with all the memory, northbridge, latency tweaks applied. The first is with HT enabled which is invalid for FSX compare. The second is with HT disabled. Both scores are certified and recorded at Futuremark in the database

 
 
CPU/SYSTEM -
Haswell 4770K CPU Speed: 4800MHz
Memory: 2400MHz @ 9-11-11-31  1T Command Rate
Northbridge (UCLK): 4500   Note: previous runs were @ NB (UCLK) 4400  (I managed to dial this in stable)
Memory Latency: 39.0-39.5 (varies test to test)
FSB | DRAM Ratio: 1:9
STRAP 100
BCLK: 100MHz
GTX 780 GPU: SC model
PState-P0 Core Speed: 1124MHz
PState-P0 Memory Speed: 3110MHz   
Nvidia Driver Version: 320.18 
Thermalright Archon SB-E x2 Air Cooler 
Asus Sabertooth Z87 Motherboard
OCZ Revo3 x2 480GB PCIe SSD 
 
 
SCORES
---------------------------------------------------------

HT ENABLED -
3DMark Score: 41138.0
SM2.0 Score: 16365.0
HDR/SM3.0 Score: 20953.0
CPU Score: 9858.0

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

HT DISABLED -
3DMark Score: 40252.0
SM2.0 Score:  16429.0
HDR/SM3.0 Score:   20867.0
CPU Score:  9076.0

 

 
I can easily put another 1000 or so points on both those scores by running unstable clocks but this isn't a race, its a true look at real potential.
 
 
As soon as I can I will setup to run a compare using as equal system setting values as possible with the SB-E @ 4700
 
 
I am really spread thin right now and cant seem to catch up daily at this point.. I think I took on too many things at once..   as usual LOL
 
 
 
 
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-15-2013 at 5:29pm
Ok here it is..
 
 
 
 
CPU/SYSTEM SETUP -
Haswell 4770K CPU Speed: 4700MHz      Match to SB-E
Memory: 2133MHz @ 9-11-10-28  2T Command Rate -  Match to SB-E memory speed/timing
Northbridge (UCLK): 4400  -  This is as close as I could match the SB-E @ approximately 4370MHz UCLK
Memory Latency: 43.0ns - 43.3ns  (varies from test to test)
FSB | DRAM Ratio: 1:8
STRAP 100
BCLK: 100MHz
GTX 780 GPU: SC model - UNDERCLOCKED
PState-P0 Core Speed: 875MHz  MAX      This is 314MHz slower than the 680GTX in compare and I think a very fair spread as the core clock drives shader clock and other specs
PState-P0 Memory Speed: 3004 MHz   MAX    Matches the 680
Nvidia Driver Version: 320.18
Thermalright Archon SB-E x2 Air Cooler
Asus Sabertooth Z87 Motherboard
OCZ Revo3 x2 480GB PCIe SSD
 
HT ENABLED -
3DMark Score: 39582.0
SM2.0 Score: 15691.0
HDR/SM3.0 Score: 20163.0
CPU Score: 9527.0
 
HT DISABLED -                                           SB-E (equal spec) 4 cores HT DISABLED-
3DMark Score: 38928.0                                            3DMark SCORE 34237
SM2.0 Score: 15997.0                                              SM2.0 Score 13156
HDR/SM3.0 Score: 20117.0                                     HDR/SM3.0 Score 17983
CPU Score: 8747.0                                                    CPU Score 8140
 
 
 
I am going to run more tests here including dropping video card memory speed and then look at where the CPU speed must come down to in order to see the two somewhat similar
 
It would appear that PCIe 3.0 reg deal may just be a placebo they put out there to trim the gripes about it working with SB-E LOL
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-16-2013 at 7:53am
 
 
 
NEW RUN -
I can not neuter the video card down anymore these are the lowest video core/memory settings I can obtain

Haswell CPU @ 4300MHz
Northbridge (UCLK): 4300  -  I can not run a faster CPU cache and must at the very least match the CPU speed
Memory Latency: 43.6ns - 43.9ns  (varies from test to test)
GTX 780 GPU: SC model - UNDERCLOCKED - Power Limit set to 75%
PState-P0 Core Speed: 855MHz           334MHz slower than 680GTX in compare
PState-P0 Memory Speed: 2500MHz   500MHz slower than 680GTX in compare
All other settings the same

                                                 Over/Under SB-E @ 4 Cores 4700MHz
3DMark Score: 35074.0                +837
SM2.0 Score: 13593.0                  +437
HDR/SM3.0 Score: 18537.0           +590
CPU Score: 8166.0                       +26   this is close


Ok so I think we have a close match. Its not going to be perfect since I can not alter the video card in certain specifications but I think this is a very fair representation.

 
Unfortunately the only way to answer the question of PCIe 3.0 in a SB-E system would be to obtain a 680GTX and run it under the same conditions. That would also show if the 780 is neutered back too far or not far enough but I don't intend to buy a 680 to find out.
 
 
There is no question that this examination has shown that a Haswell @ 4300 is at the very least running the same performance level as a SB @ 4700MHz although without a 680 to confirm a few things it is possible it could be running somewhat better than that. But this real-world test does back up formulas that defined the similar result in theory.
 
 
Next, since we have now established this base which suggests Haswell runs 400MHz faster in performace than SB, lets start bringing a few items back up, starting with memory speed/timing.
 
 
Haswell CPU @ 4300MHz
Memory: 2400MHz @ 9-11-10-28 1T Command Rate -    267MHz increase on equal timing other than Command Rate which was lowered to 1T
Memory Latency: 39.8ns - 39.9ns  a drop from 43.9 (lower is better)
FSB | DRAM Ratio: 1:9 (automatically switched due to memory speed change)
All other settings remain the same

Haswell Compare-                
                                     Over/Under 2133 @ 9-11-10-28  2T
3DMark Score: 35888.0            +814
SM2.0 Score: 14508.0              +915
HDR/SM3.0 Score: 18581.0       +44
CPU Score: 8211.0                  +45

There is an a good increase, but this test also shows why I have stated if you are not able to increase the memory speed more than 300-400MHz on the same CAS latency take the faster CPU speed over the memory speed if system stability wont allow both. Sill this is a nice increase at 814pts worth and this would show a small improvement in FSX, but is it not going to show the same gain one would see going from DDR3 2000 (or lower) to 2400
 

Next, lets run 5 tests adding 100MHz to the CPU speed each (4300 to 4800) and look at the incremental increase leaving the video card neutered and every other setting the same.

Haswell 4400MHz
                                        Increase over 4300MHz
3DMark Score: 36668.0             +636
SM2.0 Score: 15035.0               +365
HDR/SM3.0 Score: 18787.0        +198
CPU Score: 8372.0                   +161
 
 
We see a overall score increase that does not beat the memory speed/timing increase, but the CPU speed shows a nice bump. In a case where this would be the highest CPU speed I could obtain due to temps it would be in my benefit to run a faster memory product with a 400MHz higher memory speed, i.e.; DDR3 2800 @ 11-14-14  1T. Based on the memory speed bump I could expect to see this score increase by as much as 800-1000pts. But since I am not locked @ 4400 and can run higher CPU speed, the 2400 memory I have is still working just ducky for me.
 

Haswell 4500MHz
                                       Increase over 4400MHz
3DMark Score: 37625.0           +957
SM2.0 Score: 15397.0             +362
HDR/SM3.0 Score: 19415.0      +628
CPU Score: 8517.0                 +145
 
If I add this with the last score then I am doing just fine @ 2400 DRAM speed
 
 

Haswell 4600MHz
                                        Increase over 4500MHz
3DMark Score: 38389.0            +764
SM2.0 Score: 15758.0              +361
HDR/SM3.0 Score: 19792.0       +377
CPU Score: 8669.0                   +152
 
 
Still shows a positive increase however there is signs of tapering off. Although a better score the numbers are not has high as the last run.
 
 
Haswell 4700MHz
                                         Increase over 4600MHz
3DMark Score: 39006.0            +617
SM2.0 Score: 15873.0              +115
HDR/SM3.0 Score: 20120.0       +328
CPU Score: 8897.0                   +228
 

We appeared to have a fairly linear CPU increase up to this point but now the CPU performance has appeared to jump in compare to the prior tests. This suggests that as Haswell increases in CPU speed the latency at the CPU cache starts dropping faster and therefore the efficency of the processor begins to increase.
 
Checking the Cache and Memory benchmark here and comparing it to the run @ 4300 under the same memory speed and timing I find a drop from 39.9 to 39.0ns with the CPU cache latency dropping as well. This suggest we are not looking at a linear increase where we can say that Haswell is 400MHz faster than SB at any CPU speed, but as Haswell speed goes up with the higher memory speed and lower timing the equivelent in SB speed begins to rise as well. In that, we could plot the change and estimate the speed compares with decent accuracy
 
 

Haswell 4800MHz
                                        Increase/Decrease over 4700MHz
3DMark Score: 38877.0          -129      well, this is a first..   what happened?
SM2.0 Score: 15035.0            -802
HDR/SM3.0 Score: 20440.0     +320
CPU Score: 9155.0                 +258   
 
Total score and SM2 dropped significantly, but the CPU score again increased dramatically. This confirms what we see in the 4700MHz test that as the CPU speed goes up the efficiency has increased. This means we can roughly extrapolate that for every 500MHz increase in CPU speed for Haswell is equal to an additional 100Mhz of SB speed, in other words @ 4800MHz Haswell = SB @ 5300MHz instead of 5200.
 
This test also demonstrates how a CPU can begin to bottleneck a video card that is slow. The 780 is not a slow card but I have cut the core speed and memory speed as well as the power level down dramatically on that card and that appears to be showing here now. I noticed the scores dropping as CPU speed increased past 4500.
 
 
I have been running conformation tests against all these results to be sure the results I posted above were accurate and there were no 'flukes' involved so I ran the above test again and the result was:
 

REPEAT 4800MHz               Increase/Decrease over 4700MHz
3DMark Score: 38889.0         -117     
SM2.0 Score: 15056.0           -817
HDR/SM3.0 Score: 20410.0    +290
CPU Score: 9173.0               +276
 
This 3DMark06 score (38889) may LOOK great, but the reality is the performance is actually DROPPING with a higher CPU speed!
 
There is absolutely NO QUESTION here.. this is the same as someone who sticks a TI video card into a system and clocks the CPU to 4500+ the user has effectively bottlenecked that system and their performance drops. They may run a 3D test like this and see a high overall score and be fooled into believing that cheap card purchase was right!
 
This is why I have warned people for years about "too much CPU", clocking, and cheap video cards. The 780 is a beast but right now its had its abilities seriously lowered by clocking it down as low as I have, and, this also shows that if I was running a slower video card I would NOT want Haswell running above a certain speed, as the test results above shows this 780 cut down had the best overall 3DM06 score increase @ 4.5GHz  (+957pts). A cheap card would probably start to choke at 4.1 and higher.
 

Last, there is one more setting I must apply.. my CPU Cache is still running 43x or Northbridge 4300. I will increase that to 4500 and rerun the test.
 
Haswell 4800MHz @ 4500 UCLK
                                      Increase/Decrease over 4800MHz  4300 UCLK
3DMark Score: 38549.0          -340  
SM2.0 Score: 14866.0            -190
HDR/SM3.0 Score: 20273.0     -137
CPU Score: 9106.0                 -67      
 
We are definitely going the wrong way here!!  LOL   Increasing NB efficency drops performance this much?  hogwash!  Only if there is a serious bottleneck now in play. I am actually showing the overall performance at a lower CPU clock.
 
The CPU cache test clearly showed by raising the CPU cache multiplier the latency at the CPU cache slightly dropped meaning the result should have been either the same or somewhat better, therefore this test result can only mean one thing.. the video card at its current performance level is completely ruining the performance of this system now.
 
 
With the card running its original clock speed with NB @ 4500 and rerunning the test:
 
Haswell 4800MHz @ 4500 UCLK
GTX 780 GPU: SC model
PState-P0 Core Speed: 1124MHz
PState-P0 Memory Speed: 3110MHz   
                    
3DMark Score: 40302.0
SM2.0 Score: 16418.0
HDR/SM3.0 Score: 20826.0
CPU Score: 9153.0
 
Which makes total sense. The 780 is the perfect card for Haswell running just about any clock speed but especially clock speeds above 4100
 
 
There is no surprise here for me in these results. I have running these tests for years with different cards and different CPU speeds.
 
 
 
So to sum this up,
 
  • A 4 core Haswell is running far higher performance curve in compare to a 4 core SB at equal CPU speed

 

  • PCIe 3.0 may or may not be working, or, may not be working at 100% efficiency on the SB-E system even with the registry hack. The test is inconclusive simply because I do not have a 680 to test however what is shown here does suggest a high probability that it is either not working or can not work at the same level of performance.

 

  • Memory speed/timing without question increases performance.

 

 
 
 
 
It took me nearly all day yesterday to run these test and re-run them several times to conform the results. This is all the time I can put into the thread and need to move on. I hope the information has been helpful.
 
 
 
Back to Top
jcmmg View Drop Down
Senior Member
Senior Member


Joined: July-08-2010
Location: EGLL/EGPD/SVCS
Points: 120
Post Options Post Options   Thanks (0) Thanks(0)   Quote jcmmg Quote  Post ReplyReply Direct Link To This Post Posted: July-16-2013 at 8:17am
Excellent work Nick!

I will try to built some graphs out of those results, if you allows me, so you can use them as well. (weekend)

But there are a couple of things that you haven't mentioned. Well in my particular case, I wonder about the difference between 4 and 6 cores, that although a 6 core processor SB-B manage to score 10162 (CPU), is somehow pushed down and I don't know were the bottleneck is? I've speculated in my previous post about the lack of PCI-E 3.0, but since I don't have a GTX 780 that is inconclusive as well!

Can you explain a bit more about QPI versus FSB/DRAM?

And if you can show how a W7 optimization can affect these scores? if is not too much hassle...

Mobo: Asus x79 RIVE|Intel i7 3930K OC 4.7 GHz|Memory: 16 GB Dominator 2133|EVGA Nvidia 780 Classified|FSX Gold|WIN7 64 Ultimate|Alienware ALX R4 Case
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-16-2013 at 8:59am

Take your HT disable scores with 6 cores active and compare them to the results

 
but do be aware, in FSX the extra cores do not go to the primary thread, they are used for terrain and autogen batch calls. The PCIe 3.0 bus is also helping with those AG calls. This was something that was pointed out by Phil Taylor when he discussed the memory bandwidth on video cards and how the higher bandwidth cards had an advantage. He pointed out that the PCIe bus @ 2.0 could be choked by FSX assuming certain criteria was met and was one of the reasons they change how autogen was being called as well as did away with alpha-fade.
 
Of course today they could bring alpha-fade back and it would probably be a welcome addition since that stops autogen from 'popping' into and out of the scene, but the fact of the matter is with so many people and developers sucking up the CPU and GPU cycles with addons, if they did bring that back it would add another performance drop to the sim even as it is now with the better hardware.
 
 
Only those who are willing to be reasonable around scenery addon complexity with scenery settings would appreciate that return. Those who are already pushing their systems over the edge wouldn't like it because it would mean they have to give up a slider tick or two over their precious scenery they just cant seem to get enough of.
 
 
QPI and FSB are different based on design. QPI as compared to FSB in the past has the ability to move more data faster with less power consumption using a different design in link between the system and the CPU cores.
 
What is seen with Haswell is a major increase in the efficiency around this area, therefore running a UCLK of 2x memory speed is no longer required as a matter of fact you can run a 1.6x 1.7x or 1.8x with very little loss in overall system performance.
 
 
W7 optimizing...  you mean Windows 7 optimizing?     All you are doing on that end is getting rid of background activity that may add to overhead while using a high demand application. At the same time and I have noted, that Windows 7 is FAR better at dealing with services and footprint than Windows XP. When add in the fact that the SATA system and in particular the SATA system of any mechanical drive is still one of the highest CPU demanding background activities as well as add in the mechanical latency of file calls, the combination of optimizing Windows and getting the storage system in line goes a long way to freeing up CPU time.
 
To a CPU a nanosecond is a long time.
 
 
 
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-16-2013 at 9:24am
I think I know where you are going with W7 optimizing..   if you are thinking you can apply 3DM results to that,... DONT LOL
 
 
Now with Windows XP you might to a certain extent due to how inefficient it was with services and other items, but even that wont run any real compare.
 
 
You are getting into Apples vs. Oranges with that.
 
 
FSX polls the hard drive constantly for scenery, it has inputs that change, it has environment changes taking place ,the priority render system changes the result. I can go down a list but you get the idea. 3DM loads into memory and run with none of those influences.
 
I have already said that using 3DM as a base to look at performance changes with FSX can be done and there is no question that for each 3000pts gained there is a very good increase in FSX performance, but there is no way to place that into a perfect graph and say that everything about FSX will follow that value.
 
Its a general observation and one that has proven to be true over and over again, for years. Therefore it has merit but there is no direct and perfect compare between the two.
 
Results can guide but can not predict or define that every aspect of FSX will improve on a linear scale. FSX is impossible to benchmark therefore we cant use an external benchmark for anything more than a compare between the overall result in 3DM06 and the overall result in FSX when a score displays an increase.
 
 
In fact the 1000pt increases require a very keen eye to pick them up. Some do not have those powers of observation and the frame counter in FSX can and is many times useless to define anything. One must be able to use observation in which they can see everything from changes in smoothness and distance clarity to the speed and efficiency of how well scenery appears and runs. In that, if you are seeing a overall FPS increase, 4-5-6FPS, or even if its only 2-3 frames, or perhaps no change in FPS,..  you can then say you have observed a positive change regardless.
 
 
 
 
 
 
Back to Top
NickN View Drop Down
Certified Professional
Certified Professional


Joined: November-21-2007
Points: 18019
Post Options Post Options   Thanks (0) Thanks(0)   Quote NickN Quote  Post ReplyReply Direct Link To This Post Posted: July-16-2013 at 12:01pm
I probably should have posted this with the information above
 
 
This shows where the performance comes from and how small of a change at the latency makes a major difference
 
 
a. First, you will notice the memory READ SPEED @ 2133 (blue box) was actually faster than it is @ 2400. This can be misleading. Remember this similar to how memory companies snooker users into spending money on higher speed memory with higher latency and they end up with something worse. The reality is the slower memory read is faster due to the latency drop between the two compares.
 
 
b. Look at the memory latency (top line, red box). To you 5ns is nothing, to a computer? its a LONG wait!
 
 
c. Look at what happens as very small changes in L1, L2 and L3 cache latency (red box below the memory latency) drop. The read/write and copy speeds of the caches dramatically increase. This is where we have increased the efficiency of the CPU and as CPU speed rises the processors clock-per-clock efficiency rises. When you have the memory and NB tuned in tandem with this and the CPU speed topped out, it all comes together..  
 
 
 "THE THREE KINGS"
 
CPU Speed
Memory Speed and Timing
NB or UCLK Frequency
 
 
 
Back to Top
Ted striker View Drop Down
Senior Member
Senior Member


Joined: January-24-2006
Location: Denver
Points: 170
Post Options Post Options   Thanks (0) Thanks(0)   Quote Ted striker Quote  Post ReplyReply Direct Link To This Post Posted: July-16-2013 at 3:19pm
Originally posted by NickN NickN wrote:


 

Anyone who sticks a lower grade video card in a Haswell system and clocks it up is a fool.

 



" You talking to me?.....Are you talking to me?" (think Robert De Niro in Taxi) LOL

Now that you convinced me to get a 780 GTX, the one I want is sold out everywhere. Actually you convinced me a few weeks ago, but this analysis illustrates the reason why with test results. Thanks for posting it. By the way, what's causing this shortage of 780 GTXs? Did the manufacturers underestimate demand or are they holding back supply to keep prices up?

Ted
3770K @4.5Ghz, Noctua NH-C12P, Asus Z77-V Deluxe, Corsair 2133-9-11-10-27-2T, 780 GTX, Win7-64 on 256gb Intel SSD, FSX, P3Dv3, P3Dv4 on 500gb Intel SSD, PC Power & Cooling 750W, Antec P193 case
Back to Top
 Post Reply Post Reply Page  123 5>
  Share Topic   

Forum Jump Forum Permissions View Drop Down

Forum Software by Web Wiz Forums® version 12.01
Copyright ©2001-2018 Web Wiz Ltd.

This page was generated in 0.156 seconds.