Originally Posted by friend'scatdied
AMD clocked the 7970 series pretty conservatively, presumably to keep it within 20-30% of the 580.
Why would NVIDIA released a mid-range part that is faster than their last-gen flagship and the competition's flagship? It would mess around with pricing of the (expensive-to-produce) old generation and have problematic consequences for competition. It would reduce margins for all players.
Unless of course the 760 Ti is $500-600 and the 780 is even more expensive.
Even nVidia's initial fumble with Fermi, the GTX460 still turned out to be faster than the 285. When they finally fixed it with the 560Ti it was even faster. Why would it be unreasonable to expect that trend to continue with the GTX660/760 to be faster than the GTX580? It seems unlikely that nVidia will fumble with Kepler as badly as they did with Fermi...
You have to keep in mind that GCN is a departure from gaming performance efficiency to focus more on improving HPC / GPGPU performance, and thus there is a loss in gaming efficiency. While you might see AMD being conservative with the 7970's clock rate when you look at how much overclocking headroom it has, when we look at the fact that the 7970 is currently consuming more power under load than the 6970, what we really see is that AMD clocked the 7970 just enough to make sure it beat the 580 by a tangible margin across the board, rather than sticking with previously established load power consumption levels with the 6970 and then potentially tying or losing to the GTX580, even if only in just a benchmark or two.