From time to time, NVIDIA release products, which signify a new evolution stage in the GPU architecture . As a good example it is possible to consider G80. According to associate the new GPU architecture development will be GT300. Fundamental differences in the new architecture begin from the presence of directX 11 support, whereas G80 is limited to the support Of directX 10.
GT300 architecture will be based on the new processor elements organization . If in today's GPU , NVIDIA used SIMD mechanism (single instruction multiple data : one command, many data), then in GT300 the MIMD mechanism will be used(multiple instructions multiple data : many instructions, many data).
It is assumed that this will repeatedly increase the computational effectiveness of this GPU. The cluster of arithmetic-logical devices will be dynamic and united. NVIDIA engineers expect to decrease the frequencies and processor energy consumption , having simultaneously obtained higher productivity in comparison with current generation GPU.
With GT300 will be connected a large CUDA update. Furthermore, since the new GPU will be released according to the 40 NM standards , the quantity of transistors will also increase considerably.
As it is expected, NVIDIA GT300 will appear in the fourth quarter present year.