because they have performace targets, you know 580/G110 is around 550mm sq and cayman is around 380mm sq. AMD use a VLIW architexture NV uses some kind of MIMD desigin. its justs design trade offs. you then have things like bandwidth/memory MTU's and ROP's which make a big difference not just shaders.Because I want to know why and how things are achieved. If Amd has better performance per sq mm, then why do they not have a single gpu card that can match the 580?
so the big differenece yuo can see here is that VLIW relies on a good compiler for performance, MIMD relies on a good scheduler for good performance.