NVIDIA announced today that it has reached an agreement with HP to offer Tesla GPU's as an option to the HP Z800 workstation. NVIDIA launched Tesla several years ago as part of its attempt to broaden the perceived usefulness of its products (and the usefulness of GPUs in general). The driving force behind CUDA (and to some extent, GPU physics) has been the company's steadfast claim that massively parallel GPUs are capable of a level of efficiency (in terms of performance-per-watt) and scaling that's an order of magnitude higher than anything a CPU-based solution might hope to deliver.
GPUs, in other words, are kind of a big deal.
"The adoption of Tesla GPUs is the fastest of any new processor technology in the history of HPC," said Andy Keane, general manager of Tesla business at NVIDIA. "We are delighted to see a leader such as HP begin to ship Tesla GPU-enabled systems into the market and to help accelerate the work of their customers."
"HP’s workstation customers are professionals at the top of their game, pushing the limits of technology, for more creativity and innovation than ever before," said Jeff Wood, director of worldwide marketing, Workstations, HP. "The NVIDIA Tesla GPU takes our flagship Z workstations to extreme heights for floating point intensive applications."
Read NVIDIA's Tesla documentation
, and you'll inevitably end up wondering why anyone still uses those positively stone-age "CPUs" at all. The company's continuing emphasis on the performance potential of the GPU in non-gaming applications has resulted in significant advances in what NVIDIA calls "Visual Computing
." There's a Folding@Home NVIDIA GPU client, Adobe Photoshop 4 is capable of using the GPU, the TSUBAME super-computer relied on "detailed tuning for our GPU BLAS library" when it posted a 10 Tflop performance boost in June of this year, and the number of consumer-level games and applications capable of utilizing the GPU has been slowly but steadily increasing.
Being able to buy Tesla GPU's as part of a workstation package that starts at four figures (as opposed to five) theoretically opens the door to more developers who want to explore GPU computing (and CUDA), which, in theory, increases the likelihood that those lucky souls with NVIDIA GPUs will be able to take advantage of said capability at some point in the future. Keep in mind, NVIDIA is just one horse in a three-way race. Team Green has unquestionably sunk more resources and time into GPU computing than Intel or AMD, but early head starts do not, in and of themselves, reward later market dominance. With Larrabee and its promise of real-time raytracing edging closer, NVIDIA could one day find itself a niche player in a market it helped create.