this post shows why Nvidia has large dies.
from G80-GT200 you gain 33% larger SIMD array (I think) The node shrink only help so much in design
and from GT200 to GF100 you gain an even larger SIMD array.
They have only kept adding to the SIMD arrays and have not taken anything out to shrink it's design size.
ATI here did this
3870-4870 did away with Ring bus and added more UDV stuff
4870-5870 did away with crossbar in there and add more powerfull tesseltion unit
5870- 6870 did away with 5:1 simd and went to 4:1 and add even more powerful tessellation unit lastly dual Rasterzisted.
(this excludes any AA and AF changes)





Reply With Quote
Bookmarks