Nvidia: We Are Going to Have a Full Series of PetaFLOPS Supercomputers

Nvidia Pins Hopes on Supercomputers, Intends to Attract New Talent

by Anton Shilov
09/30/2010 | 11:21 PM

Nvidia Corp. clearly has a vision for the market of high-performance computing as at its latest GPU Technology Conference the company made a prediction that loads of  PetaFLOPS-capable supercomputers would actually feature Nvidia's compute cards known as Tesla.

 

"We are going to have a full series of PetaFLOP computers. Soon there will be more PetaFLOP-class computers outside the U.S. than there are inside the U.S. So, for me that was a worrisome trend because here you have a clear technology advantage that provides a lot of benefit," said Andy Keane, said Andy Keane, general manager of Tesla business at Nvidia, reports InsideHPC.com web-site.

At present Nvidia does not have a lot of Tesla deployments, at least in terms of units, despite of their advantages. In fact, the world's two largest HPC GPU deployments are Nebulae supercomputer with 4640 Tesla C2050 cards as well as Tianhe-1 that sports a total of 2560 dual-chip ATI Radeon HD 4870 X2 cards (5120 GPUs).

It is widely discussed that ExaFLOPS supercomputers will have to rely on non-x86-based processors or some kind of hybrid chips  due to power consumption requirements. As a result, HPC customers are resting a wide array of technologies that could enable maximum performance per watt and naturally Nvidia Tesla 2000-series is one of such technologies.