See a trend? Nvidia is moving away from Optimizing specifically for DX9 because DX9 performance is already there. With DX11 titles they make significant gains. Kepler is designed to excel in DX10+ environments, aka current generation technology. Considering Microsoft and Sony are alleged to be announcing new Consoles within the next year and all signs point to DX10/11 support on them, we can assume that even console ports will start optimizing for DX11. So the question then becomes do you want to pay $300 for a card that does amazing in DX11 but takes a hit on DX9 (like 5-10% but still playable), or $400+ for a card that has a slight edge in older DX9 games and falls behind pretty noticeably in DX11 games?
47FPS average is not what I would call playable.
And to answer your question, I want a card that is an upgrade in everything, be it DX9 or 11. Current gen Keplers are already a massive downgrade from Fermi in compute performance, if gaming is their bag, I expect it to excel in every game. Right now, I have around 20x to 30x more games for DX9 than DX11, and there are plenty more console ports on the way...so yeah, it'll take priority over DX11 for me, because DX11 performance still has a long way to go. If Nvidia really decides to call it quits on DX9 performance, I'll switch to AMD, as the LAST thing I want is to have to build yet another separate rig for an older API.
If you buy a 660 to play at 2560 x 1600 or 5760 x 1080 you are an idiot.
Huh? How the hell is gaming on a card with 2GB VRAM at 2560x1600 stupid, in DirectX 9 no less, when people are still claiming GTX 580s with 1.5GB are overkill for 1080p (and when most folks bought 1GB 460s to game at 1080p)?
