Originally posted by Ruined
IQ differences that can only be detected by zooming in on pictures with paint programs or by any other method than looking at the game with your naked eye in action are 100% useless in terms of comparing in-game IQ differences between the card.
Note I said "only" - read the statement again. I'm saying if you can't detect the differences in game with your naked eye, then any IQ differences that can be found with paint programs, etc, are useless, because in the end you still cannot detect them with your naked eye, and cannot see them. What is not valid about that?
The problem is, the different mip levels are
obvious to the naked eye in some circumstances. I don't have any ATi screenshots for comparison, but take a look at the shots taken on a 5900 by Dave Baumann at Beyond3D - Bear in mind this is with the drivers set to use trilinear filtering: