Various 2x SLI GTX 480 Benchmarks

H

heatlesssun

Extremely [H]
Joined
Nov 5, 2005
Messages
44,154
Just threw together some various bechmarks, most of them are canned but I played through the begnings of Metro 2033 and AvP. Note that EVERY game is set to maximum in game settings (even the Heaven 2.0 Benchmark) with the sole exception of Metro 2003, using AAA and Advanced Physix is off, however DX 11 DoF as well as Tessellation are on. Also note that the Bokeh filter is NOT available for SLI configurations. Nothing was set through the nVidia Control Panel. Resolution for all games is 1920x1200.

I'll keep this spot updated and I'll entertain requests if I have the game.

Test Rig:

Windows 7 Ultimate x64 RTM fully patched as of 04/20/2010
Intel i7-980x@ 4.13 GHz (25 x 165 BCLK @ 1.365V)
Corsair H50 cooler
ASUS P6X58D Premium BIOS 703
eVGA GeForce GTX 480 2x SLI
HP w2408h 24" 1920x1200 LCD Monitor
Enermax EGX1000EWL Galaxy 1000W
Corsair XMS3 6GB DDR3
2x WD 1002FBYS 1TB SATA 7200 RPM non-RAID
Antec 1200 Case

Metro%202033.jpg


AvP.jpg


Crysis.jpg


dirt2_game%202010-04-19%2018-16-54-84.jpg


JustCause2%202010-04-19%2023-10-41-95.jpg


JustCause2%202010-04-20%2000-35-48-16.jpg


JustCause2%202010-04-19%2023-08-21-22.jpg


unigine%202010-04-20%2002-21-46-79.jpg


HAWX_dx10%202010-04-20%2001-52-30-93%202010.04.20.1007.15.jpg


LostPlanetDX10%202010-04-19%2015-44-56-04.jpg


StreetFighterIV%202010-04-19%2015-34-23-33.jpg


RE5DX10%202010-04-20%2001-46-29-24.jpg
 
Last edited:
I swear the SLI scaling in Metro is funny. I'm running a GTX 470 SLI setup and if I'm indoors in the tunnels performance and scaling are all over the place. EVGA Precision is telling me that the card utilization is between 50 and 90 but if I'm outside its more like 85-90 all the time and the frame rates are always high and stable. I don't get it and was wondering if you saw anything like that in your testing.
 
Everything is at 1920x1200. Indoors in Metro is quite a bit less demanding than outdoors for sure. What I found is that anything beyond AAA KILLED performance with everything else the same.
 
heatlesssun said:
Everything is at 1920x1200. Indoors in Metro is quite a bit less demanding than outdoors for sure. What I found is that anything beyond AAA KILLED performance with everything else the same.
Click to expand...

Metro 2033 seems to be the only game so far that kill all the hardware..

no matter if you have tri-SLI or not.. its unplayable in outdoor...
 
shansoft said:
Metro 2033 seems to be the only game so far that kill all the hardware..

no matter if you have tri-SLI or not.. its unplayable in outdoor...
Click to expand...

While it doesn't perform nearly as well outside I disagree, its actually quite playable at these settings for me. 24 FPS is good enough for this game to be enjoyable ESPECIALLY in the rich outdoors environments, and that's at the low end, it really tends to stay above 30 FPS.
 
heatlesssun said:
While it doesn't perform nearly as well outside I disagree, its actually quite playable at these settings for me. 24 FPS is good enough for this game to be enjoyable ESPECIALLY in the rich outdoors environments, and that's at the low end, it really tends to stay above 30 FPS.
Click to expand...

below 40fps in this game really hard for me to make my mouse move.... feels weird.. unlike Crysis smooth everything out..
 
shansoft said:
below 40fps in this game really hard for me to make my mouse move.... feels weird.. unlike Crysis smooth everything out..
Click to expand...

Actually I think that that the WHOLE game doesn't feel very precise mouse-wise so I turned down my DPI on my Sidewider X8 and it works much better.
 
heatlesssun said:
Actually I think that that the WHOLE game doesn't feel very precise mouse-wise so I turned down my DPI on my Sidewider X8 and it works much better.
Click to expand...

I have tried to turn down setting to get 40 avg fps, its a lot smoother...

this game creeps me out at night, don't want to turn it on right now... :p
 
The only game where i dont see any kind of difference between my 480 and my 285 is crysis and crysis warhead. I swear those games even feel somewhat slower even than my 285. I haven't run the benchmark yet but there is no way in hell that its above 30fps at 1920x1200 with 4x aa at very high. How in the hell are 2 of these things getting an average fps of 58 at those settings?? Every other game I have tried is huge increases in fps but crysis is still sluggish on very high with or without aa at 1920x1200.
 
Actually this benchmark for a single 480 was getting about 35 FPS. If your 480 isn't running Crysis WAY better than your 285 something not right with the Crysis setup. Indeed a single 480 was rivaling my 3 280s in this game.
 
Thanks for posting. I was kind of surprised by the Crysis results, as it's lower than what I get on my 5850's. Same with Metro 2033 actually, although I don't know what/where you tested. I guess Crysis is still a killer.
 
Mr. K6 said:
Thanks for posting. I was kind of surprised by the Crysis results, as it's lower than what I get on my 5850's. Same with Metro 2033 actually, although I don't know what/where you tested. I guess Crysis is still a killer.
Click to expand...

Thanks! That Crysis bechmark tool has giving me quite varying results, this was on the low end but I've seen 10 FPS better on some runs. The new killer is Metro. Are you saying that two 5850s are running Metro better than this or that the results are lower than you expected?
 
heatlesssun said:
Thanks! That Crysis bechmark tool has giving me quite varying results, this was on the low end but I've seen 10 FPS better on some runs. The new killer is Metro. Are you saying that two 5850s are running Metro better than this or that the results are lower than you expected?
Click to expand...
Yeah, I usually get 60-65FPS on the Island demo with 8xAA. I played through Metro at 2560x1600 with Tesselation and Advanced PhysX on, but no DoF because it's bugged as hell at 2560x1600. I also turned off 4xAA not so much because of the framerate drop but because it blurred the hell out of everything. Anyway, I was getting 50-60FPS average at those settings, not sure how it would compare to 1920x1200 with DoF on.
 
Mr. K6 said:
Oh yeah, I demand IQ :p.
Click to expand...

Ok, that makes no sense, at those settings and resoultion you're outrunning EVERYTHING, including 2 and 3 way CF 5870, there's no way you're getting 50-60 FPS at those settings: http://www.hardware.info/nl-NL/articles/amdnampoZGCa/Clash_of_the_Titans_3way_SLI_GTX_480_test/8 And these numbers seem to be in line with my experience on the 480 side. Unless DoF causes that big of hit, I need to check that cause if you're getting 50-60 at 2560x1600 I should be able to easily blow by that at a lower resolution with two 480s.
 
Last edited:
heatlesssun said:
Ok, that makes no sense, at those settings and resoultion you're outrunning EVERYTHING, including 2 and 3 way CF 5870, there's no way you're getting 50-60 FPS at those settings: http://www.hardware.info/nl-NL/articles/amdnampoZGCa/Clash_of_the_Titans_3way_SLI_GTX_480_test/8 And these numbers seem to be in line with my experience on the 480 side. Unless DoF causes that big of hit, I need to check that cause if you're getting 50-60 at 2560x1600 I should be able to easily blow by that at a lower resolution with two 480s.
Click to expand...

Dof causes a HUGE HIT. So yes I would believe that he is getting those FPS with Dof turned off. Depth of Field is a neat feature, but not worth the performance hit at this point in time.
 
heatlesssun said:
Ok, that makes no sense, at those settings and resoultion you're outrunning EVERYTHING, including 2 and 3 way CF 5870, there's no way you're getting 50-60 FPS at those settings: http://www.hardware.info/nl-NL/articles/amdnampoZGCa/Clash_of_the_Titans_3way_SLI_GTX_480_test/8 And these numbers seem to be in line with my experience on the 480 side. Unless DoF causes that big of hit, I need to check that cause if you're getting 50-60 at 2560x1600 I should be able to easily blow by that at a lower resolution with two 480s.
Click to expand...
Well that's what I was saying, it seems off. I think vjcsmoke hit the nail on the head, it must be DoF. The reason I say it's bugged is because when I turn it on at 2560x1600, my FPS goes down to 1-2FPS on the start menu, and stays that way in game too, so something isn't right and I can't test it at my resolution. Reviews showed the GTX480's have a ton of horsepower in Metro 2033, that's why I was wondering what was up (thought maybe it could have been an SLI scaling issue or something).
 
Mr. K6 said:
Well that's what I was saying, it seems off. I think vjcsmoke hit the nail on the head, it must be DoF. The reason I say it's bugged is because when I turn it on at 2560x1600, my FPS goes down to 1-2FPS on the start menu, and stays that way in game too, so something isn't right and I can't test it at my resolution. Reviews showed the GTX480's have a ton of horsepower in Metro 2033, that's why I was wondering what was up (thought maybe it could have been an SLI scaling issue or something).
Click to expand...

http://hardocp.com/article/2010/03/26/nvidia_fermi_gtx_470_480_sli_review/3

This IS with DoF off. Something else is going on.
 
great benches. so in next gen maybe probably the gen after that, we can finally run crysis in all glory with min above 60fps. it really taken a long time to catch up to crysis.
 
my average fps on that demo with the same settings in crysis as the op was 17.3fps with a single 480 and an i7 @ 3.6. Not sure how he is getting those results...

edit: ok so wtf just for giggles i took off my 480 overclock from 750 back to stock and my average fps with the same settings as the op went to 31.5. umm wtf? furmark runs fine with my oc settings and gives me no artifacts. this is odd...

edit #2: turns out it wasn't the oc causing the slowdown. It was having the nvidia 3d stereoscopic setting in the control panel enabled. Granted it wasn't active/turned on in the game or in the benchmark but just having the option checked to give you the ability to ctrl t to enable it or disable it on the fly causes crysis to think it is running in stereoscopic 3d even when it's not, which accounts for the drastic loss in fps. I wonder if other games are also suffering from this nasty little driver flaw. Anyone with the nvidia vision glasses who has the driver option checked in the control panel and who isn't even running the game in 3d will have nasty fps. Nvidia's got a big bug in this one to fix.
 
Last edited:
Why don't you guys turn on DoF and see how that works? I'll try turning it off and see what I get.
 
heatlesssun said:
Why don't you guys turn on DoF and see how that works? I'll try turning it off and see what I get.
Click to expand...
I'd be interested in seeing what kind of boost you get. I could throw up numbers too, but I have no idea what you're using for a benchmark. I saw one site bench the "Chase" level, which makes sense, since it's kind of an on rails run through (literally), and as long as you don't move the mouse should be consistent.
 
Just got my 2nd 480 last nite..Will be running benchies soon!!..:)
 
coldstorm said:
could you unengine with normall tessilation try to see where that would fit into these numbers.
Click to expand...

unigine%202010-04-22%2001-14-14-87.jpg


Thanks for that link. I have all these games and their test system is similar to mine, a 6GB i7-980x system at only 75mhz faster than mine so I could compare to the Saphire 4GB 5970. Nice card, TweakTown gizzed over it, but its in the same price ballbark as 2 480s and while less power and cooler running than 2x 480 SLI, they are is doing very nicely against it.
 
Last edited:
wow. 480 sli really killing heaven. so other than that and maybe metro 2033 we have to wait on benches from that they about performing even on some games. not bad for 1000 investment either way. the benefit to that 5970 is less slot 3 vs 4 ( 5 is you keep them one slot apart) . less power and less heat. other than that that about even. from most benches. also these benches do show the physx does kill fps by 1/2 in some case 2/3 or more in other. so it best to have separate card for physx if you wish to run physx or just turn it off.
 
The thing is when you get into higher settings and newer DX 11 games, Fermi really does pull ahead by a lot more oftern than not. The only DX 11 title that's in the 5000s favor is BC2.
 
heatlesssun said:
The thing is when you get into higher settings and newer DX 11 games, Fermi really does pull ahead by a lot more oftern than not. The only DX 11 title that's in the 5000s favor is BC2.
Click to expand...

well they didn't bench metro 2033 but i would give the win to nvidia based on previous benches. but I would still like to see numbers to see how close or far it really is
 
You must log in or register to reply here.
Back
Top