Hardware maker Nvidia has announced a new computer that has the power of a cluster of computers at a small fraction of the cost.
Nvidia, working with several partners, has developed the Tesla Personal Supercomputer, powered by a graphics processing unit based on Nvidia's Cuda parallel computing architecture. Computers using the Tesla C1060 GPU processor will have 250 times the processing power of a typical PC workstation, enabling researchers to run complicated simulations, experiments and number crunching without sharing a supercomputing cluster, Nvidia said.
The Tesla C1060 card, available on computers Tuesday, will sell for US$1,699, with desktop computer systems including the card selling for less than $9,995, said Andrew Humber, an Nvidia spokesman. The systems would run at a processing speed of four teraflops, or four trillion floating point operations per second.
A grid of computers can cost 100 times the cost of one of the Tesla-powered workstations, Nvidia said.
"This represents phenomenal price/performance for computational researchers who have typically had to compete for time on expensive and power-hungry clusters," he said in an e-mail.
Several institutions, including the Max Planck Institute, the University of Illinois at Urbana-Champaign and Cambridge University, are already using GPU-based personal supercomputers, Nvidia said.
"GPU based systems enable us to run life science codes in minutes rather than the hours it took earlier," Jack Collins, manager of scientific computing and program development at SAIC-Frederick's Advanced Biomedical Computing Center in Frederick, Maryland, said in a statement. "This exceptional speed-up has the ability to accelerate the discovery of potentially life-saving anti-cancer drugs."
While there have been claims of desktop supercomputers in the past, "this time it's for real," Burton Smith, a technical fellow at Microsoft, added in a statement.
Among the computer makers offering Tesla Personal Supercomputers are Dell, Lenovo, Asus, Western Scientific and several others.