by Anton Shilov
08/13/2006 | 04:50 AM
Nvidia Corp., one of the world’s largest supplier of graphics processors, said during a conference call with financial analysts that the company could imagine a chip that combines general-purpose processing features and graphics processing capabilities. Such devices will be eventually available from Advanced Micro Devices and ATI, however, Nvidia does see value in the approach as well.
“In some segments, we could imagine GPUs integrating with CPUs, or CPUs integrating with GPUs. But we’ll approach those segments as it makes sense. Because often times it’s not necessarily one CPU. It could be x86. It could be ASICs, it could be PowerPC. It could be Cell. It could be ARM. It could be something else,” said Nvidia’s chief executive and president Jen-Hsun Huang.
The head of Nvidia was responding to a question regarding plans by combined AMD-ATI company as well as of Intel Corp. to release chips that converge central processing unit (CPU) and graphics processing unit (GPU) for certain markets. Nvidia believes that while such a microprocessor would be needed for some market segments, the plan of the company is to create GPUs that offer value to end-users.
“The reason why the GPU has been separated from the CPU historically is because the two work off of completely different rhythms. And because of the different segmentations and the velocity of innovation, and all of those things, the GPU continues to be a standalone device for the segments in the marketplace that we serve. Our job isn’t really to worry about which is going to integrate which; our job is to make the GPU valuable,” said Mr. Huang.
Microprocessors integrating graphics capabilities and features of core-logic sets may provide companies like AMD and Intel an opportunity to target yet untapped markets with cost-effective, yet more-or-less powerful computers. Presently only AMD and Intel have competitive x86 processors, the chips that power the absolute majority of personal computers in the world. Presently there are no desktop-class Windows operating systems to run on PowerPC, Cell or ARM central processing units, which means that the world’s two leading CPU makers have advantage over rivals who do not own x86 technology.
At this time neither AMD or Intel officially announced roadmaps for converged CPU-GPU devices, which may be an indicator that such chips are planned for distant future only. Still, the fact that even Nvidia’s chief executive talks about such products, means that the idea makes some general sense.