updated 01:05 pm EST, Tue November 18, 2008
NVIDIA Tesla supercomputer
NVIDIA on Tuesday announced the availability of its GPU-based Tesla architecture Personal Supercomputer, which, the company promises, has as much as 250 times the computing performance of a PC thanks to NVIDIA CUDA parallel computing architecture and up to 960 parallel processing cores. Along with its partners, buyers can now opt for desktop PCs with the sort of computing power previously reserved for supercomputing clusters.
The higher performance is made possible to heterogeneous computing, where GPUs such as NVIDIA's Tesla C1060 Computing Processor, works in conjunction with CPUs. The C1060 is based on NVIDIA's CUDA parallel computing architecture that enables developers to take advantage of the Tesla design via the C programming language. Engineers, designers and researchers in various industries can perform complex computations without having to head into the lab, including in CAD/CAM design, computational fluid dynamics, imaging or other tasks.
NVIDIA's GPU-based Tesla chips will be available at 28 OEMs, system builders and resellers, including Dell, Lenovo, Asus and Western Scientific. Dell promises its Precision R5400 and T7400 PCs will use the Tesla C1060 GPU to give the,=m as much as two teraflops of extra performance. Systems with four-GPU Tesla chips are available now from the likes of Colfax, Tycrid and Boxx, while 3-GPU Tesla Supercomputers can be had from AMAX, Microway, Penguin, Velocity and Western Scientific, all for under $10,000.