NVIDIA introduced their next generation GPU architecture based on Fermi in September 2009. The company has now revealed more information about the GF100 (Graphics Fermi 100) chip based on the new architecture. The GF100 chip architecture is very different from previous GPUs from NVIDIA.

The new chip is divided into four main graphics processing clusters (GPC) with each cluster featuring 16 streaming multiprocessors (SM) and each SM having its own geometry setup and processing unit called Polymorph Engine. The SM also has 32 shader cores (CUDA processors) and four SMs in each GPC share a single raster engine unit. All the SM cores have 64KB of L1 cache each and communicate to each other through 48 ROPS that are attached to 768KB of L2 cache. The L2 cache has full read/write access and shares data between the 16 SM units. The cache and ROPS are tied to the main memory bus, which is 384-bit GDDR5 memory interface.

NVIDIA didn’t provide any information about chip size, clock speed, board specifications, memory size, cost and release date. Websites like PC Perspective and Hot Hardware have posted preview of the GF100 chip and they all agree that “When it comes to traditional gaming, as well as upcoming DX11 gaming.  Off the cuff I would guess that in current applications, the GF100 based high-end card will be 15% to 20% faster overall than the HD 5870.  It will not overtake the HD 5970 dual GPU card, but it looks to certainly trounce the single HD 5870.  I think that NVIDIA is also doing tessellation right, and when games come out which will implement it in a more pervasive manner, NVIDIA will take a much larger performance lead.”

Below we have added a video showing Far Cry 2 performance of GF100 when compared to previous generation GT200 GPU.