In a recent NVIDIA analyst conference, company VP Michael Hara spoke on various topics pertaining to the GPU business. Hara spoke out on the increasing power of the GPU, which would lend itself exceptionally well to applications outside of graphics. In fact, the NVIDIA VP believes that GPUs can now offer better performance to cost value than x86-achitecture chips.
â€œToday GPUs are used for graphics, and GPUs in the future will be used for computing,â€ said Hara. â€œIn certain markets that are mathematically intense, the GPU has proven to be superior technology to a standalone x86. When you put an x86 next to a GPU, the order of magnitude you get in the increase in performance is mind-boggling at this point.â€
Hara went on to give the example of the financial sector being keenly interested in the computational power of modern GPUs. Such mathematically intense industries currently still rely on expensive servers and specialized hardware, such as math co-processors. NVIDIA feels that their products will soon be able to offer performance approaching one teraFLOPS to these markets.
â€œWe're now approaching teraflops of performance, which is more than Cray. If you think about the Cray, it was still in the gigaFLOPS, we're pushing teraFLOPS,â€ Hara stated. He also added that companies such as Cray were at an immediate disadvantage because its research and development costs were very high, but its products were aimed at a very small market â€“ something that NVIDIA believes it doesnâ€™t have to worry about. â€œWe think about the fact that there's this tremendous floating point capability that is existing in these GPUs, the question is how to tap it.â€
AMD currently delivers one teraflop of computing power in a single system with two ATI Radeon HD 2900 XT-derived Stream Processors. Intel plans to deliver over a teraflop of computing power with derivatives of its Teraflops Research Chip. As for specialized, non-graphics applications for GPUs, ATI Radeon cards are providing the greatest performance per chip in Stanford Universityâ€™s Folding@home program.
At the current time, the core of NVIDIAâ€™s business is still in the PC market. In response to a question asking Hara where NVIDIA is looking to next to grow its GPU business outside of the PC space, he said that the next frontier for the company will be on the road â€“ inside cars.
â€œWeâ€™ll talk to you every quarter more and more about our automotive market. As our customers in the automotive market look for GPUs, youâ€™re going to find at the end of this year NVIDIA making announcements with large car manufacturers putting GPUs in cars,â€ Hara revealed.
As LCD screens find their way inside more automobiles, NVIDIA is exploring ways to fit its technology in those spaces. Although the in-car GPU may not yet have to power a game of Need for Speed, GPS navigational systems are slowly evolving from simplistic 2D-maps to more detailed and elaborate 3D representations. â€œIn this case, itâ€™s more for information, but that car will need a GPU,â€ Hara added.
Certainly, growth for NVIDIA appears to be outside of its traditional roots. â€œYouâ€™ll find that every new market we add will be non-PC,â€ the VP said. â€œWeâ€™re not looking for a way outside of the PC market, but weâ€™re looking for a way to expand the utility of the GPU ... We look at platforms that are trying to deliver an experience to the customer.â€