EVGA 8800GTX - A gaming analysis on FEAR and CPU scaling

[dominick32]

EVGA 8800GTX - A gaming analysis on FEAR and CPU scaling

I decided to perform another series of tests on the new EVGA 8800GTX to really try and get in depth and pin point any/if any CPU bottleneck on these 8800 GPU's. The results are interesting to say the least.

System Specs:
E6700 Conroe "Week 27"
Asus P5B-Deluxe Wifi (vcore mod)
OCZ Titanium Alpha VX2 Ram
Vapochill LS "Jin Mod" Cooling

I chose to use the in game FEAR benchmarking utility measuring for average FPS from various CPU clockspeeds. I also categorized an Overclocked GPU as well as a Stock GPU chart to make deciphering any bottlenecks even easier. All tests were performed on maximum/ultra quality and every single piece of eye candy enabled. The resolution is the standard benchmarking/gaming res of 1024 X 768 - 32bit color.

E6700 @ 2.66 GHz, Stock 8800GTX



E6700 @ 2.66 GHz, Overclocked 8800GTX (660/1040)



E6700 @ 2.93 GHz, Stock 8800GTX



E6700 @ 2.93 GHz, Overclocked 8800GTX (660/1040)



E6700 @ 3.30 GHz, Stock 8800GTX



E6700 @ 3.30 GHz, Overclocked 8800GTX (660/1040)



E6700 @ 3.60 GHz, 8800GTX



E6700 @ 3.60 GHz, Overclocked 8800GTX (660/1040)



E6700 @ 4.00 GHz, 8800GTX

[dominick32]

E6700 @ 4.00 GHz, Overclocked 8800GTX (660/1040)



E6700 @ 4.20 GHz, Stock 8800GTX



E6700 @ 4.20 GHz, Overclocked 8800GTX (660/1040)




And here are the charts to compliment the benchmarking results:









After sitting here and looking over these results for the past 40 minutes. It seems to me, that once the CPU clock gets up to around 3.40 GHz, the GPU completely hits a wall and stops scaling with CPU clockspeed at this resolution. It definitely picks up a few frames per second, from 2.66GHz 3.40 GHz every 100 mhz but once you approach 3.40GHz, your FEAR gaming will not scale AT ALL with CPU clockspeed. Again, this is all done at 1024 X 768 resolution. If I turn up the resolution to 1280 X 960, things get very interesting. When high res gaming, it is quite obvious here that the GPU's performance does not scale AT ALL at any CPU clockspeed:

[dominick32]

E6700 @ 2.66GHz, Overclocked 8800GTX, HIGH RESOLUTION BENCHMARK



E6700 @ 3.30 GHz, Overclocked 8800GTX, HIGH RESOLUTION BENCHMARK



E6700 @ 4.20 GHz, Overclocked 8800GTX, HIGH RESOLUTION BENCHMARK




And here are the high resolution charts:





As you can see, at a standard resolution of 1024 X 768, and maximum/ultra settings. An 8800 GPU definitely scales a tiny bit with CPU clockspeed up to about 3.40 GHz on an Intel Core 2 Duo processor. However, under high resolution gaming (for example: max/ultra quality and 1280 X 960) the 8800 GPU absolutely does not scale with CPU power AT ALL. I'm talking zilch, nada, nothing.

Thanks for reading!

Dom

[DTU_XaVier]

Great to know, awesome that you take the time to do this kind of tests
Now I know that my E6600 won't bottleneck if I ever get a GF8 GFX, 'cause let's face it... Who buys these cards to do 1024x768??

Best Regards

[G4h4o8s6T]

Very nice work, and as Xavier said, thanks for taking the time to do these tests.

Now Id really like to see how an 8800 paired with an A64 x2/DC Opteron compares to C2D + 8800 in high res. gaming performance...........like 1680x1050 and up Has anyone seen or done any of these or similar tests??

[DTU_XaVier]

#5 Wouldn't those tests be redundant, as we have already seen here, that with graphic-intensive game-engines, the cards hit a limitation already at relatively low resolutions??

Best Regards

[BlaqMale]

ty sir, would you mind doing the same test for oblivion and or carbon?

[BobyTT]

Great analysis dude! very helpful

[dnottis]

I had done some testing to find the same thing. I didnt narrow it down to 3.4 Ghz - but somewhere over 3.4 I found the CPU to no longer provide any benefit. That was fine with me since I run 3.5 or 3.6 Ghz on the CPU.

[FEEXX]

thx for sharing, sure you had a damn nice spare time at bench!

[gillll]

thank you very much for the benches.

can u add 1280*1024 and 1200*1600 ?

[DTU_XaVier]

Everyone asking for performance numbers at higher resolutions... Stop!
The guy's not showing what his card can do, he's showing when it stops having a CPU-bottleneck... He has already provided these results, and used a good amount of time on it I'd say, so stop bugging him... Higher resolutions are irrelevant for this thread

Best Regards

[ineedaname]

yea guys higher resolutions are pointless cuz you'll just end up seeing the same trends for cpu vs gpu

[Adapa]

Great work dude, compliments my results with X2 4400@ 2.67 vs E6600@ 2.67GHz. Simply proves the theory every gamer should know by now, there is no big difference in performance once you game at maxed settings regardless of your cpu.

[DTU_XaVier]

Great work dude, compliments my results with X2 4400@ 2.67 vs E6600@ 2.67GHz. Simply proves the theory every gamer should know by now, there is no big difference in performance once you game at maxed settings regardless of your cpu.

Unless your max-settings are 1024x768 no AA/AF, and your CPU is a Celeron D

Best Regards

[ewitte]

Everyone asking for performance numbers at higher resolutions... Stop!
The guy's not showing what his card can do, he's showing when it stops having a CPU-bottleneck... He has already provided these results, and used a good amount of time on it I'd say, so stop bugging him... Higher resolutions are irrelevant for this thread

Best Regards

I'm just laughing that its labled high resolution. I don't think of 1280x960 as anywhere near high resolution.

[dominick32]

I'm just laughing that its labled high resolution. I don't think of 1280x960 as anywhere near high resolution.

I apologize my benchmarking wasnt up to your standards. Not everyone is lucky enough to have a 24" LCD monitor. At the time of testing I had a 19" LCD monitor with only the capability to provide a 1280 X 960 benchmark considered high res. And as you can see, the PC definitely considered that resolution "high res" as the GPU did not scale with the CPU anymore. That would mean all of the work was being done on the graphics processor in resolutions over 1024 X 768 ....

For your information, I received a BenQ 24" Widescreen last night to be exact. Ran 1600 X 1200 in fear which produced exactly the same results, do you consider that high resolution?

[DTU_XaVier]

I'm just laughing that its labled high resolution. I don't think of 1280x960 as anywhere near high resolution.

It's relatively high, it just seems low, because of the pure performance of these cards.... There will come a day when 8800GTX won't run games at that res, some day 2 years from now

Best Regards

[ewitte]

For your information, I received a BenQ 24" Widescreen last night to be exact. Ran 1600 X 1200 in fear which produced exactly the same results, do you consider that high resolution?

Your benchmarking is fine. It showed what it needed. Yeah usually anything 1600x1200 or more would be considered. Is that the native resolution? I'd think widescreen would be 1920x1200. BTW you should be able to bench higher than your native. It will scroll when you move the mouse but the fps should be about the same as if the monitor supported it officially.

[ewitte]

It's relatively high, it just seems low, because of the pure performance of these cards.... There will come a day when 8800GTX won't run games at that res, some day 2 years from now

Best Regards

Time to upgrade then.

[Neova]

I just read the 3D Mark scaling thread and now this. Great info!

If you can update this thread with some 1600x1200 action and perhaps 1 more game engine (Quake 4/Doom 3 for opengl or HL Source or even UT2K5), it would give everyone a very good idea of how the top game engines performs with respect to CPU or GPU resources.

Looking forward to your future analyses.

[ewitte]

Here are some numbers I just got this morning.

FEAR 1600x1200, 4xAA, 16xAF everything on maximum AND softshadows on

Minimum - 41fps
Average - 81fps

E6700 @ 2.67Ghz
XFX 8800GTX overclocked slightly (600/1000... this card goes to 660/1060).

[Drag]

Can you overclock an video card without additional cooling?
And what evga video card did you use?
the normal one ore the calibre?
http://www.anandtech.com/video/showdoc.aspx?i=2873&p=3

[zarathrustra]

Here is the same test with game settings maxed out at high res (1280x960). Cpu is an Opteron 170 at 3.0GHz, and the XFX 8800gtx is clocked at the same 660/1040.



This is at the same settings, but with 16x AA/AF forced in the drivers.



I'll try push the Opty to 3.1GHz and post another screen shot, but by the looks, a dual core A64 at 3.0GHz is causing a substantial bottleneck.

[zarathrustra]

Here's the same settings at 1280x960, with no aa/af, but at a CPU clock of 3.1GHz on the Opty 170.


Shows the Advantage of c2d over even the highest speed A64 dual cores.

Can anyone say "bottleneck"