Quote Originally Posted by saaya View Post
pfff i dont buy that... its just as much based on Gt200 as rv870 is based on rv770, and its not a new architecture... its pretty obvious that if there was a clear goal with gf100, it was obviously gpgpu... they went for massive shader core power, which, they CAN use for tesselation... but i really dont think that this was their original idea behind all that shader power...
Well, I guess we can agree to disagree then. Granted, the Shaders, TMUs and ROPs are still called the same thing but look at the difference:

GF100:




GT200:




Some of the main differences of the GF100 (there are MANY more)

- Large dedicated (and unified) cache structure

- Fixed function stages are now integrated into each cluster (or SM as they were called on the GT200) through the Polymorph engine

- Texture units now have direct access to the cache structure

- Scalability through unified ROP + L2 + memory controller structure

- Dedicated L1 Load / Store cache

- Incorporates a raster engine into each GPC which groups the three pipeline stages (Edge Setup, Z-Cull and the Rasterizer)

- GDDR5 memory controller

- DX11-necessary components such as a dedicated tessellator, etc.