Quote Originally Posted by Rai View Post
Somehow with Fermi failing in every other department, minimums are now all the rage with certain users. That's a single data point, you need graphs like HardOCP to check how badly each card is fluctuating. If the 5870 had very low minimums time and again that would drag the average down. Graphs will show the minimum as well as the overall trend in fps which is a lot more useful than a single data point. It also really depends on which review you sight, here the 5870 is ahead.

http://www.techpowerup.com/reviews/N..._Fermi/10.html
I posted a HardOCP link, and this is what the reviewer said:

You can see that we were really pushing these video cards very hard in this game with these framerates, but I have to say that GTX 480 SLI was the most playable, you could really feel the difference. In fact, I would say that 2X AA at 2560x1600 with all Enthusiast settings is playable on GTX 480 SLI here, but not on the Radeon HD 5870 CrossFireX.
Looking at the graph, the 5870 Crossfire set up plunges below the 30 FPS mark an inordinate amount of times.

Quote Originally Posted by jaredpace View Post
Even on Nvidia's own PR slide, the 5870 is faster than the 480 in crysis. Get tired of seeing carfax defend a 480's performance in every thread
Correcting a false assumption and defending the 480s performance aren't the same thing..

Quote Originally Posted by [XC] hipno650 View Post
I loled so hard from that, so true. I will have to say though it's pretty obvious who in this thread owns both a 5870 and a GTX 480 and has used them both when playing crysis. those who own both know that the difference between the 5870 performance and the GTX 480 performance is much much bigger then people think or than the graphs say. I was straight up disappointed with my 5870 in Crysis, my GTX 480 on the other hand is totally amazing and runs like a dream.
Exactly! Gameplay experience with the 480 in Crysis is just much better than the 5870 no doubt about it.

I also never noticed that graph's Crysis warhead performance for the GTX 480, graph must be a fake considering I have tested it myself and thats not the case.
HardOCP typically gives the most accurate results when it comes to benchmarks, as you can see the entire range of results over a specific period of time rather than just a mere snapshot.

I wish more reviewers used their method.