- Apr 12, 2010
- 6,313
- 0
- 26,460
Hi guys. I finally got around to finding, downloading, and running some gaming benchmarks to compare 5850s in single and crossfire modes. These things are amazing!
The benchmarks I ran are FurMark, Resident Evil 5, Dirt 2, STALKER: Call of Pripyat, and Crysis.
Ok, first of all let's start with my CPU, GPU, and RAM settings. You'll notice my 5850s are OCed - I bios flashed them for 875/1200 at stock volts. Keep this in mind and note that this thread is not to compare OC vs stock 5850s. All games had pretty much the maximum detail settings selected at 1920x1080 resolution. The specifics are in the graph titles.
PC Settings:
First up is the FurMark results.
Here we're seeing an increase of up to 1.9 times average FPS! Wow!
EDIT: New benchmark! This benchmark app was a bit disappointing. I only gives the average FPS as well as total frames rendered and average frame time. I saved a log of the individual frames and their fps then copied it to Excel but the results were rather skewed and not accurate, in particular for crossfire. It showed a 10.3 min and 210 max lol, and the average was about 120 so not even close to the average the program spits out at the end. Similar results on the single 5850 side... anyway, so here's the actual average FPS that it gives you at the end.
(sorry for mislabelling which is which, I'll have it fixed tonight)
Next we have Resident Evil 5. Note that they only give you average FPS at the end. There is however a small graph which I failed to screenshot but I can tell you that the crossfire graph was very smooth while the single 5850 was rather up and down. Still, not much difference in averages (1.27).
EDIT: I ran this one again with my CPU at 4ghz, and the FPS went up a fair bit (135 or so), so that explains the poor scaling - CPU bottleneck!
EDIT: Found out that VSync was enabled in my first Dirt 2 bench. Big mistake there, although it's very interesting how drastic it hurts the min FPS. Logic tells me that it should have just run at a smooth 60hz given how high the fps are without vsync. Anyway, here's the new results without vsync:
Average FPS is boosted 1.71 times.
This next one has a lot of info. The STALKER: CoP benchmark runs 4 times in different settings. Very nice increase here, the minimum crossfire FPS are almost the same as the average single GPU.
And finally, the big one. Can it run Crysis? Hell yes.
Note that in game, I play it at 2xAA Very High instead of 4xAA and it becomes silky smooth - and the difference in detail is completely unnoticeable by myself.
Alright so already lots of great results. RE5 is disappointing, I wish it would give a better summary. Just remember that the CF 5850s had pretty much zero dip in FPS while the single did - but I'm guessing this game was getting CPU bottle-necked since they're both quite similar - or maybe the game does it. I do not know.
I have one final graph to show the increase in performance seen. I did not include RE5 or Dirt 2 on it as they both do not have all info (min, max, avg). The style of graph here adds up the increases, so basically you can see that min is affected the least, but both max and average frames are greatly boosted. For specific increments you'll have to look at the number inside each block (ie, 1.84 = 84% better)
As an interesting aside, here is the results of using Vsync in Dirt 2. I find it fascinating how badly it affects FPS.
*Added 12/17/2010*
For reference, here's my Vantage score with CF 5850s and a high OC, just to showcase the power of two 5850s
The benchmarks I ran are FurMark, Resident Evil 5, Dirt 2, STALKER: Call of Pripyat, and Crysis.
Ok, first of all let's start with my CPU, GPU, and RAM settings. You'll notice my 5850s are OCed - I bios flashed them for 875/1200 at stock volts. Keep this in mind and note that this thread is not to compare OC vs stock 5850s. All games had pretty much the maximum detail settings selected at 1920x1080 resolution. The specifics are in the graph titles.
PC Settings:
First up is the FurMark results.
Here we're seeing an increase of up to 1.9 times average FPS! Wow!
EDIT: New benchmark! This benchmark app was a bit disappointing. I only gives the average FPS as well as total frames rendered and average frame time. I saved a log of the individual frames and their fps then copied it to Excel but the results were rather skewed and not accurate, in particular for crossfire. It showed a 10.3 min and 210 max lol, and the average was about 120 so not even close to the average the program spits out at the end. Similar results on the single 5850 side... anyway, so here's the actual average FPS that it gives you at the end.
(sorry for mislabelling which is which, I'll have it fixed tonight)
Next we have Resident Evil 5. Note that they only give you average FPS at the end. There is however a small graph which I failed to screenshot but I can tell you that the crossfire graph was very smooth while the single 5850 was rather up and down. Still, not much difference in averages (1.27).
EDIT: I ran this one again with my CPU at 4ghz, and the FPS went up a fair bit (135 or so), so that explains the poor scaling - CPU bottleneck!
EDIT: Found out that VSync was enabled in my first Dirt 2 bench. Big mistake there, although it's very interesting how drastic it hurts the min FPS. Logic tells me that it should have just run at a smooth 60hz given how high the fps are without vsync. Anyway, here's the new results without vsync:
Average FPS is boosted 1.71 times.
This next one has a lot of info. The STALKER: CoP benchmark runs 4 times in different settings. Very nice increase here, the minimum crossfire FPS are almost the same as the average single GPU.
And finally, the big one. Can it run Crysis? Hell yes.
Note that in game, I play it at 2xAA Very High instead of 4xAA and it becomes silky smooth - and the difference in detail is completely unnoticeable by myself.
Alright so already lots of great results. RE5 is disappointing, I wish it would give a better summary. Just remember that the CF 5850s had pretty much zero dip in FPS while the single did - but I'm guessing this game was getting CPU bottle-necked since they're both quite similar - or maybe the game does it. I do not know.
I have one final graph to show the increase in performance seen. I did not include RE5 or Dirt 2 on it as they both do not have all info (min, max, avg). The style of graph here adds up the increases, so basically you can see that min is affected the least, but both max and average frames are greatly boosted. For specific increments you'll have to look at the number inside each block (ie, 1.84 = 84% better)
As an interesting aside, here is the results of using Vsync in Dirt 2. I find it fascinating how badly it affects FPS.
*Added 12/17/2010*
For reference, here's my Vantage score with CF 5850s and a high OC, just to showcase the power of two 5850s
Facebook
Twitter
Instagram