The rumors have always pointed to NI being a mix of the original NI front-end with the shader core of the previous generations. Nothing has been wrong.
First, you can already see that G80 and GT200 are pretty damn identical, so it's hardly 3 different architectures. Fermi is of course new, but here's the kicker to your pointless argument:
Until you see what ATI did on the front end, which won't necessarily be known until the NDA is lifted, how can you say its any more of a reconfiguration than Nvidia? In your own diagram, one would just say GT200 is a GT200 is an expanded G80/G92 (in a similar style to 5800 from 4800). And Fermi had a large change in the shader/arithmetic part of the architecture, and not as much around it - one could say R6xx -> R7xx was the opposite, which was identical shader/arithmetic, major changes on the other stuff (such as TMU/RBEs)
So if you're going to play the "not much of an architecture game" card, one could just as easily point to other aspects of Nvidia and say the exact same thing, which was my point, but one you obviously missedBut that's the thing about unified shader architecture, you'll rarely see every aspect overhauled as we saw with G7x->G80 and R5xx->R600




. Fermi is of course new, but here's the kicker to your pointless argument:
Reply With Quote

Bookmarks