DirectX 10 Shootout: Nvidia vs. ATI | Tom's Hardware
Well, the article itself demonstrated how poorly most DX10 cards play the current line-up of DX10 games (except the 8800 GTX and 8800 Ultra, of course), but the last 2 pages are complete BS, IMO. I mean, I know Nvidia is by far winning the battle w/ the GTX and Ultra and I own an 8800 GTS 640 MB myself, but this article just screams bias. At the end, they add up all the FPS scores (from 1024 all the way up to 1920) and then divide by the price to give a price/performance ratio. First of all, who spends $250+ to play games @ 1024x768 w/ no AA or AF ? No one because most people who are running these cards are using a 20" or larger screen, so why bother using this score as if it matters ? Better yet, 60% of the tests are at 1280x1024 or lower which, again, is completely unrealistic for cards of this caliber. If you actually add up the 2 highest resolutions they used (1600x1200 w/ AA+AF and 1920x1200 w/ AA+AF), the scores go a little more like this:
ATI HD2900 XT 512 MB - 299.6 fps (25.0 avg)
Nvidia 8800 GTS 640 MB - 320.2 fps (26.7 avg)
Nvidia 8800 GTS 320 MB - 284.6 fps (23.7 avg)
So, at REALISTIC resolutions, the 8800 GTS 640 MB is still a better choice overall, but it doesn't smoke the HD2900 XT by a whopping 20% as TH would have you believe, but a much more believable 7%. Also, the 320 MB, which TH ranked 12% higher, scores 5% lower than the HD2900 XT. Mind you, this is pretty good given the 8800 GTS 320 MB is well over $100 cheaper.
The second bias is that the HD2900 XT 1 GB and the HD2900 Pro were not included, plus throw in a bunch of overclocked Nvidia cards, and you've got a sea of green all over the place.