Gtx 580 > 7970?
I apologize in advance if my question is a bit naive.
According to a few video reviews and benchmarks I've see on YouTube (TimeToLiveCustoms, Trubritar, Motherboardsorg), they're claiming the new AMD 7970 out performs the GTX 580--not by a large margin--but enough to suggest it is the better card of the two.
Why is it that your list of High End Video Cards shows the GTX 580 (3gb?) out performing the 7970?
Passmark seems to be pretty unreliable at the top of the charts. For example 580 > 590? And Even 570 > 590? Makes no sense. 580 > 570 > 6990 isn't realistic either.
Different benchmarks (and games) perform differently on different video cards.
For example on Crysis 2 the 580 has a significant lead over the 7970.
Thanks for the answer, I hope mgthack's comments didn't sway the tone of my question (meant no disrespect to your site).
Originally Posted by passmark
I will definitely keep this in mind on my next purchase, I've used the information on your site for three builds now.
My intention wasn't to trash this benchmark because I feel it's fairly reliable for mid to low range cards. However I think it's very important to point out how this benchmark would differ from real world performance - ie, actual gaming.
As mentioned by the admin, Crysis and Crysis 2 tend to favor GeForces, but round up the benchmarks over a variety of games, and the 7970 will be faster than the 580. I believe a high quality 3d benchmark should take into account enough different application profiles such that the score at least approaches real world results. I think it's kind of a cop-out to say, because Game X performs better on the 580, we're not going to do anything about it.
Just comparing to another well-known synthetic benchmark, 3dmark, I think it's worth mentioning that those scores give a better representation of real world performance than G3D. Understandably the reasons are not 100% under PassMark's control, especially when looking at dual GPU/SLI boards.
I'll stop writing because I know this topic has been covered before. The other threads are informative enough.
Last edited by mtghack; 02-19-2012 at 01:41 AM.
We want to be able to test a broad range of cards, on a broad range of operating systems.
We want people to be able to measure how much better their performance will be if they upgrade, for example, from Windows XP with GeForce 2 to Window 7 with a GTX 580.
The current 3dmark11 doesn't run on XP and doesn't run on maybe 80% of video cards that people actually have. (And I mean people in the broad sense, not just gamers).
I would also believe that if you look at slightly older games (DirectX9 era) then PerformanceTest would in fact be reasonable reflective of gaming performance.
In the next major release of PerformanceTest will be including a DirectX 11 test. This should favor the AMD 7970 slightly as its DX11 performance is better than its DX9 & 10 performance. We'll also be having another go at SLI / Crossfire support.