I think a lot of people here underestimate GTX580's performance (based on TPU's benchmarks of old games), and haven't carefully compared GTX580 vs. HD6970 in the same games AMD uses in their slidedeck. Also, let's add that GTX580 overclocks another
21%.
Review 1
Review 2
Xbitlabs did an excellent Overclocked vs. Overclocked review in
all the games where AMD specifically shows HD7970 dominating the HD6970.
So what happens when a review actually includes those same DX11 games that AMD uses in their slidedeck to show HD7970 dominating HD6970 (Shogun 2, Crysis 2, Unigine, STALKER: COP, Lost Planet 2)? Here we go.
GTX580 @ 935mhz is
57% faster than HD6970 @ 940mhz at 1920x1080 AA modes
GTX580 @ 935mhz is
33% faster than HD6970 @ 940mhz at 2560x1600 AA modes
I hope someone does an overclocked GTX580 vs. stock HD7970 vs. overclocked HD7970. GTX580 gains a huge amount of performance over HD6970 when overclocked in DX11 games.
I can see HD7970 being better than an overclocked GTX580 for 2560x1600 or higher due to more VRAM and higher texture filtering. But at 1920x1080/1200, an overclocked GTX580 plows an overclocked HD6970 in the very same DX11 games AMD is showing HD7970 in good light.
If
this slide is correct, then HD7970 would be easily 70-80% faster than an HD6970, considering it's showing HD7970 beating GTX580 by 40-60% already. But if this is true, then HD7970 would be almost as fast as a single HD6990.