No, quite the opposite actually. Instead of relying on application specific optimizations to attain scaling, every board would be working at 100% efficiency at all times regardless of the application, assuming CPU limitations are not present.Originally posted by: ViRGE
Wouldn't that be a tremendous waste of resources though?
That was just one example. You could have two boards @ 4xAA giving 8xAA for the same performance hit as one board @ 4xAA.2x->4x MSAA is not even close to halving the performance of a card in most situations.
For multi-sampling yes, but not for super-sampling; 2x super-sampling cuts the performance in half on a single card while 4x super-sampling reduces it to one quarter. Super-sampling is highly desirable to combat shader aliasing in modern games but is prohibitively expensive.Originally posted by: ViRGE
Sure, it would be for "free" but you're adding an additional card to do something that would normally reduce performance by what, 20%?