I am posting this as a companion piece to " Layte's super-duper 480SLI Benchmark-O-rama!"
http://forums.overclockers.co.uk/showthread.php?t=18168653
The purpose of this is to give an idea how an ATI Quadfire 4GIG setup (2 x Sapphire 4gig 5970's) performs compared to a GTX 480 SLI setup. The contention amongst many people is that GTX 480 SLI scales better, has better minimum framerates and possibly better average framerates, compared to an ATI multi gpu setup. So I wanted to find out. Therefore I will compare my results to Layte's.
I tried to keep the benchmark settings the same as Layte's, in the same order, and resolution tested at was 1920x1200. However in benchmarks where Nvidia uses 16xAA, usually only 8xMSAA is available for ATI. I also benchmarked with my system in it's daily setup which has the GPU speed set only slightly higer then a stock 5870 which is considerably slower then lots of 5970 users out there. Catalyst 10.7a drivers with 10.8a application profiles were used as I find they give the best multi gpu performance so far.
System :
i7 980X @ 4.2ghz
- 1.29v vcore,1.34v vtt
Sapphire 5970 4gig x 2 Quadfire
- 900mhz (+50mhz) core, 2400mhz memory (stock)
Catalyst 10.7a with 10.8a application profiles
Results :
*** I will be comparing the results below to the Nvidia results in this thread http://forums.overclockers.co.uk/showthread.php?t=18168653 ***
Just Cause 2
1920x1200,8xAA,16xAF
Dark Tower
Desert Sunrise
Concrete Jungle
Analysis :
Almost nothing in it, comparing Nvidia and ATI.
Crysis Warhead
1920x1200, 8xAA, Enthusiast Settings, DX10, Ambush Flythrough
Analysis :
Nvidia score 29.29 min framerate, 89.97 max, and an average of 69.90.
ATI score 39.50 min framerate, 91.42 max, and an average of 69.11.
So really nothing in it here either, though ATI has a higher minimum framerate.
Crysis
1900x1200, 8XAA, Very High settings, DX10, Assault Harbour
Analysis :
Nvidia score 35.22 min framerate, 89.88 max, and an average of 62.37.
ATI score 37.70 min framerate, 109.14 max, and an average of 64.74.
Again nothing in it here, except ATI's higher max framerate.
Stalker Call of Pripyat
1920x1200, Ultra settings, Enh. DX11
Analysis :
ATI has considerably higher maximum framerates. e.g 426.4fps max in day test for ATI v 295.1 for Nvidia. Nvidia's minimum framerates and average framerates are higher and in a couple of cases considerably higher e.g. 85.6fps min in sunshafts test v 47.9 for ATI.
Streetfighter IV. 1920x1200, 8XAA, all settings maxed out (16xAA option not available ingame for ATI, limited to 8XAA)
Analysis :
Almost nothing in it. ATI less then 15fps faster but limited to 8xAA. Nvidia can pull off C16xAA.
Resident Evil 5
1920x1200, 8X MSAA, DX10. (16xAA option not available ingame for ATI, limited to 8X MSAA)
Analysis :
ATI wins in all areas here especially in Area 1 and 4, most notably in Area 1 with a score of 423.9 v 245 for Nvidia. However in Area 2 and 3 the results are closer. Nvidia also has 16xAA to its credit.
Dirt 2 (game benchmark, not demo benchmark)
1920x1200, DX11, all settings maxed out
Analysis :
ATI has a slightly higher minimum and maximum framerate. Nothing in it really. I did notice the screen judder the odd time which gave the impression the framerate was not what the benchmark reported. But I have seen this personally on ATI and Nvidia systems and I think it's the game engine. After a few laps it seems to iron out when you are actually playing the game.
Devil May Cry 4
1920x1200, max settings, 8x MSAA (16xAA option not available ingame for ATI, limited to 8X MSAA)
Analysis :
Assuming the test Layte ran was with 2 gtx 480's in sli (not sure as that test is in the single card section); then ATI is dominating in this benchmark by over 100fps+ in 3 out of 4 tests. However I suspect this was a single GTX 480 run so the comparison is probably not fair, especially as the GTX 480 can do 16xAA, where as ATI is limited to 8x MSAA
Aliens V Predator
1920x1200, default settings
Analysis :
Nothing to say about this one. With ATI cards this just does not scale. 1 5970 4gig was only a couple of frames slower then 2.
Opinion so far :
From what I've seen Nvidia GTX 480 SLI dooes not have significantly better minimum or average framerates compared to ATI 5870 Quadfire and by extension 5870 Trifire, since it's generally thought that the 4th ATI GPU does very little and Trifire scales better. In some cases like in Crysis, ATI has higher minimum framerates, and matches Nvidia in Warhead. ATI also on a number of occasions has higher maximum framerates. Overall though - assuming that there is very little difference in performance between Nvidia and ATI in minimum and average framerates, then I also assume ATI provides just a smooth a gaming experience since superior minimum framerates was the reason a lot of nvidia users gave as the main factor in delivering a smoother gaming experience compared to ATI.
Finally, one thing I noticed when monitoring my benchmarks with Afterburner / G19 display, was that the GPU load was pretty evenly spread across all 4 ATI GPU's, and temps generally stayed in low to mid 60's for each core. So Quadfire may have got a lot better then people think and may be scaling.
Although only a synthetic benchmark; here are 2 Heaven screenshots I took some time back, comparing a single 5970 4gig against 2 of them in quadfire at default settings. While the results are nothing special as the CPU is running at stock 3.33ghz and the GPU's are also running at stock; it does show scaling is there.
5970 4gig
5970 4gig X 2 Quadfire
I would very much like to see how GTX 480 3 way performs against 5970 4GB Quadfire. I believe that even at 4.2ghz, while 3 way GTX 480 will be CPU limited just like 5970 4GB Quadfire; it will perform better then ATI seeing as GTX 480 SLI is neck and neck with the best ATI has to offer. I am tempted to pick up 3 GTX 480's in the next week or two. ATI is fine, Gsods and Driver issues are not. If I can get the right fan profile and be confident my 800D with 2x140mm sidefans can cool these 480's then I might go for it.
GTA IV and Batman benchmarks are on the way. If you'd like any others, I'd be happy to oblige
http://forums.overclockers.co.uk/showthread.php?t=18168653
The purpose of this is to give an idea how an ATI Quadfire 4GIG setup (2 x Sapphire 4gig 5970's) performs compared to a GTX 480 SLI setup. The contention amongst many people is that GTX 480 SLI scales better, has better minimum framerates and possibly better average framerates, compared to an ATI multi gpu setup. So I wanted to find out. Therefore I will compare my results to Layte's.
I tried to keep the benchmark settings the same as Layte's, in the same order, and resolution tested at was 1920x1200. However in benchmarks where Nvidia uses 16xAA, usually only 8xMSAA is available for ATI. I also benchmarked with my system in it's daily setup which has the GPU speed set only slightly higer then a stock 5870 which is considerably slower then lots of 5970 users out there. Catalyst 10.7a drivers with 10.8a application profiles were used as I find they give the best multi gpu performance so far.
System :
i7 980X @ 4.2ghz
- 1.29v vcore,1.34v vtt
Sapphire 5970 4gig x 2 Quadfire
- 900mhz (+50mhz) core, 2400mhz memory (stock)
Catalyst 10.7a with 10.8a application profiles
Results :
*** I will be comparing the results below to the Nvidia results in this thread http://forums.overclockers.co.uk/showthread.php?t=18168653 ***
Just Cause 2
1920x1200,8xAA,16xAF
Dark Tower
Desert Sunrise
Concrete Jungle
Analysis :
Almost nothing in it, comparing Nvidia and ATI.
Crysis Warhead
1920x1200, 8xAA, Enthusiast Settings, DX10, Ambush Flythrough
Analysis :
Nvidia score 29.29 min framerate, 89.97 max, and an average of 69.90.
ATI score 39.50 min framerate, 91.42 max, and an average of 69.11.
So really nothing in it here either, though ATI has a higher minimum framerate.
Crysis
1900x1200, 8XAA, Very High settings, DX10, Assault Harbour
Analysis :
Nvidia score 35.22 min framerate, 89.88 max, and an average of 62.37.
ATI score 37.70 min framerate, 109.14 max, and an average of 64.74.
Again nothing in it here, except ATI's higher max framerate.
Stalker Call of Pripyat
1920x1200, Ultra settings, Enh. DX11
Analysis :
ATI has considerably higher maximum framerates. e.g 426.4fps max in day test for ATI v 295.1 for Nvidia. Nvidia's minimum framerates and average framerates are higher and in a couple of cases considerably higher e.g. 85.6fps min in sunshafts test v 47.9 for ATI.
Streetfighter IV. 1920x1200, 8XAA, all settings maxed out (16xAA option not available ingame for ATI, limited to 8XAA)
Analysis :
Almost nothing in it. ATI less then 15fps faster but limited to 8xAA. Nvidia can pull off C16xAA.
Resident Evil 5
1920x1200, 8X MSAA, DX10. (16xAA option not available ingame for ATI, limited to 8X MSAA)
Analysis :
ATI wins in all areas here especially in Area 1 and 4, most notably in Area 1 with a score of 423.9 v 245 for Nvidia. However in Area 2 and 3 the results are closer. Nvidia also has 16xAA to its credit.
Dirt 2 (game benchmark, not demo benchmark)
1920x1200, DX11, all settings maxed out
Analysis :
ATI has a slightly higher minimum and maximum framerate. Nothing in it really. I did notice the screen judder the odd time which gave the impression the framerate was not what the benchmark reported. But I have seen this personally on ATI and Nvidia systems and I think it's the game engine. After a few laps it seems to iron out when you are actually playing the game.
Devil May Cry 4
1920x1200, max settings, 8x MSAA (16xAA option not available ingame for ATI, limited to 8X MSAA)
Analysis :
Assuming the test Layte ran was with 2 gtx 480's in sli (not sure as that test is in the single card section); then ATI is dominating in this benchmark by over 100fps+ in 3 out of 4 tests. However I suspect this was a single GTX 480 run so the comparison is probably not fair, especially as the GTX 480 can do 16xAA, where as ATI is limited to 8x MSAA
Aliens V Predator
1920x1200, default settings
Analysis :
Nothing to say about this one. With ATI cards this just does not scale. 1 5970 4gig was only a couple of frames slower then 2.
Opinion so far :
From what I've seen Nvidia GTX 480 SLI dooes not have significantly better minimum or average framerates compared to ATI 5870 Quadfire and by extension 5870 Trifire, since it's generally thought that the 4th ATI GPU does very little and Trifire scales better. In some cases like in Crysis, ATI has higher minimum framerates, and matches Nvidia in Warhead. ATI also on a number of occasions has higher maximum framerates. Overall though - assuming that there is very little difference in performance between Nvidia and ATI in minimum and average framerates, then I also assume ATI provides just a smooth a gaming experience since superior minimum framerates was the reason a lot of nvidia users gave as the main factor in delivering a smoother gaming experience compared to ATI.
Finally, one thing I noticed when monitoring my benchmarks with Afterburner / G19 display, was that the GPU load was pretty evenly spread across all 4 ATI GPU's, and temps generally stayed in low to mid 60's for each core. So Quadfire may have got a lot better then people think and may be scaling.
Although only a synthetic benchmark; here are 2 Heaven screenshots I took some time back, comparing a single 5970 4gig against 2 of them in quadfire at default settings. While the results are nothing special as the CPU is running at stock 3.33ghz and the GPU's are also running at stock; it does show scaling is there.
5970 4gig
5970 4gig X 2 Quadfire
I would very much like to see how GTX 480 3 way performs against 5970 4GB Quadfire. I believe that even at 4.2ghz, while 3 way GTX 480 will be CPU limited just like 5970 4GB Quadfire; it will perform better then ATI seeing as GTX 480 SLI is neck and neck with the best ATI has to offer. I am tempted to pick up 3 GTX 480's in the next week or two. ATI is fine, Gsods and Driver issues are not. If I can get the right fan profile and be confident my 800D with 2x140mm sidefans can cool these 480's then I might go for it.
GTA IV and Batman benchmarks are on the way. If you'd like any others, I'd be happy to oblige
Last edited: