14 DAY TRIAL // The Fermi architecture from NVIDIA has been the subject of both admiration and severe scrutiny, mostly because the chips it spawned were, while the fastest in terms of performance, the slowest to actually get to the market, not to mention the hottest and most power hungry. Now, however, with the TSMC 40nm supply steadily getting better, the Santa Clara GPU maker seems to be progressing more quickly in its efforts to bring forth mainstream and entry-level GPUs. In fact, it appears that the upcoming GF104 is not the only chip that will make it to the market this summer.
The GF104, less power hungry than the GF100, albeit weaker, will first show up at the core of the GeForce GTX 460 and will also supposedly power the company's first DirectX 11-capable dual-GPU card. The other two processors that should show up during the coming months are the GF106 and GF108.
The GF108 will cater to the needs of the entry-level market and will reportedly have a 128-bit memory interface and support for GDDR3 and GDDR5 VRAM. Basically, this part will follow the same principle as the GF100, but will have far fewer shaders and a slower clock, among other things.
There is no clear word on the TDP (thermal design power), though it should be lower than that of the GF106-powered boards, speculated at 75W. The GF106 itself should enable replacements for G92-based adapters (GTS250 for instance).
Fudzilla states that, based on what its sources provided, it can be assumed that mainstream and entry-level Fermi products will be out before summer ends, “ready for back to school” as it were. All that remains is to see whether the cards arrive in a timely fashion, or if history repeats itself and delaying tactics become mandatory once again. In the end, it probably depends most on the supply of 40nm chips.