Nvidia Corp., according to a media report, plans to refresh nearly the whole lineup of its graphics processors by back-to-school season. The new line will reportedly include GF104, GF106 and GF108 graphics processing units, which support DirectX 11 along with other innovations, and will cover a lot of market segments.
Nvidia plans to release its performance-mainstream code-named GF106 chip in July and mainstream GF108 graphics chip in August, 2010, according to a news-story from DigiTimes web-site. The GF104 graphics processing unit for high-end/performance mainstream is projected to arrive in June, 2010. The timeline of the releases points to Nvidia’s plan to almost completely refresh its product lineup with Fermi architecture-based products by the back-to-school season.
Specifications of the new graphics processors are naturally not known. However, one can speculate about them, keeping in mind specifications of the flagship GF100 graphics processor and the peculiarities of the Fermi architecture.
The GF100 (NV60, G300) graphics processor sports four graphics processing clusters (GPCs) in total, each of which sports four streaming multiprocessors (each SM has 32 stream processors and 4 texture units) and four PolyMorph engines. The chip also has six 64-bit memory controllers with six raster operation (ROP) partions and 768KB universal L2 cache (128KB per ROP).
As it can be observed, Fermi architecture is pretty flexible, so, Nvidia can cut down the number of GPCs or reduce the amount of SMs per GPC, while also balancing the number of memory controllers, ROPs and the size of L2 cache. Quite naturally, the lowest-end Fermi product will the least optimized for GPU computing whereas mainstream GF106 and GF104 are likely to inherit many GPU computing peculiarities of the original GF100.
Nvidia did not comment on the news-story.