
26-08-2008
|
 | Member | | Join Date: Nov 2005
Posts: 2,412
| |
| NVIDIA envisions GPU in 2013
Although NVIDIA refuses to announce any new product at the nVision 08, the firm discusses the chameleon still its technologies, present and future. Thus, Tony Tamasi, a vice-chairpersons of NVIDIA in charge of development, has undertaken a review of developments over the past decades in terms of GPU. Recall that we moved from a model based chip sets, a model partially programmable chip (DirectX 8.0 and Pixel Shaders) to finally reach chips far more flexible and widely programmable with a premium unified architecture. 
Tony Tamasi opportunity to establish a chronology of inflection points in the field of architecture before 3D compare the developments in terms of features and performance of these various architectures. All this in order to determine a projection of performances we have a right to expect a GPU in 2008 and again in 2013. The projection of a performance GPU in 2008, based on previous models lies elsewhere in the 10% performance currently offered by the GeForce GTX 200. As the news to wait until 2013 in our graphics chips, NVIDIA will risk a brief roadmap referring to the care of the native language C + + for programming the chip, the establishment of a virtual pipeline, complete support pointers or a partition active in terms of workloads of the chip. NVIDIA of course insists on the fact that in 2013, the GPU will always be a dedicated chip, but mixed with the tasks of graphics rendering and tasks of generic calculation. The GPU in 2013 according to NVIDIA: some of the possible developments
As for the performance, the GPU display of 2013, according to NVIDIA, a computing power of 20 Teraflops while its memory bandwidth exceeds the terabyte / second. Meet us in five years to see if these forecasts prove correct ... or not! |