by Anton Shilov
08/26/2009 | 11:47 AM
Chief executive officer of Nvidia Corp. said during his keynote at Hot Chips conference that graphics processing units (GPUs) have excellent prospects for further performance growth. He also indicated that it makes no sense to integrate central processors and graphics chips since discrete processors have higher performance.
According to Mr. Huang, by 2015 graphics processing units will have computing power that is 570 times higher compared to performance of today’s GPUs. Meanwhile, central processing units (CPUs) will be only three times faster than today’s most powerful chips. Considering the fact that modern graphics chips can offer about 1TFLOPs of computing power, then in 2015 they will offer whopping 570TFLOPs.
The prediction of Mr. Huang sharply contradicts with prediction of William Dally, chief scientist of Nvidia, who expects GPUs to have 20TFLOPs performance in 2015.
During question and answer section at the end of the speech, professor David Patterson of U.C. Berkeley asked if Mr. Huang had to do it over, would he still partition the CPU and GPU into separate chips. The answer Nvidia’s chief exec gave was that there were three constituents, the programmers, OEMs/ODMs, and chip designers, and each had differing requirements that make it difficult to bet on integrating new and very rapidly developing architectures into one device. By separating these functions, each can develop at its own pace and also provide the flexibility to address many market opportunities. Of course, Mr. Huang stressed that the GPU is evolving much faster than any other chip architecture.
The head of Nvidia also enthusiastically painted a picture of a world where the massive threading and computing capability of the GPU can provide many orders of magnitude performance increases over just the multi-core CPU alone.