Quote Originally Posted by LordEC911 View Post
I understand that but it also shows that it is not necessarily bandwidth limited.
Where in those sources does it show/say that texture filtering is "bottlenecked by bandwidth?"

Only reason that GTX285 is performing "better" is because it is doing less work and produces a lesser quality of AF.
its obvious that the quality is higher in rv870. i would expect fermi to have higher quality than the older nv algorithms. they have been using the same one since g80. its also BS that review sites bump up AA and res and make it look like a card is more powerful. there is a reason we use adaptive AF rather than .
the larrabee paper said that interpolation done through software is slower but the difference is negligible. calculation is not the bottleneck.

http://forum.beyond3d.com/showthread.php?t=55241