Originally posted by jbirney
Ok here is one from a user named Dio who just happens to write drivers for the R9700:
Notice the part where he says A lot is due to the adaptive algorithm which sounds a lot like a speed optimization to me.
Adaptive algorithm doesn't necessarily have to have anything to do with the aniso degree selection method. Every aniso implementation must have a degree selection method, and therefore must be adaptive in some form.
What we don't know is just how adaptive ATI's method is. For example, it seems possible that ATI does indeed disable anisotropic for certain texture stages in order to improve performance. An example of a situation where this might be beneficial would be with a lightmap. Quite simply, applying aniso to a lightmap would result in very little image quality gain. So, ATI may disable anisotropic, for instance, when one texture is modulated with the base texture. This may be another way to have an adaptive algorithm.
As another way of stating it, every single anisotropic algorithm in existence will select different anisotropic degrees in a method very similar to the MIP map selection. That makes every anisotropic algorithm adaptive. But, it is possible that some algorithms are still more adaptive by disabling anisotropic entirely in certain situations (Or, at least, reducing the degree of aniso applied in certain scenarios...).
But the z-rotation issue is certainly not part of ATI's adaptive algorithm, if it is indeed more adaptive than nVidia's (which is possible, but not altogether likely).