Supercomputing Comes to Desktop

SAN FRANCISCO ~ US technology firm NVIDIA rolled out high-performance “personal supercomputers” this week that let desktop workstations handle mind-boggling tasks once far beyond their capabilities.

Computers built with innovative NVIDIA graphics processing units (GPUs) are capable of handling calculations typically relegated to expensive supercomputing “clusters” – a technology breakthrough the company says could soon bring lightning speeds to the next generation of computers aimed at the consumer market.

NVIDIA’s Tesla Personal Supercomputers deliver approximately 250 times the processing power of current computer workstations for similar prices, according to the California-based company.

“This changes everything. This supercomputing power is being brought to the workstation,” Tesla computing products general manager Andy Keane said.

Massachusetts Institute of Technology and other universities and research facilities are already using GPU-based personal supercomputers.

“GPU-based systems enable us to run life-science codes in minutes rather than the hours it took earlier,” said Jack Collins of the Advanced Biomedical Computing Center in the US state of Maryland.

“This exceptional speedup has the ability to accelerate the discovery of potentially life-saving anti-cancer drugs.”

NVIDIA, founded in 1993, became renowned for GPUs that drive sophisticated computer game and video hardware.

Its graphics chips have been evolving to augment and even supplant central processing units (CPUs) at the heart of most computers.

While CPUs typically handle tasks in a linear style, zipping from start to finish in series, GPUs work on tasks simultaneously in order to do things such as get color pixels together on screens to present moving images.

Sets of NVIDIA chips built for speed, power, and superior graphics production are built into upgraded Macintosh notebook computers recently rolled out by Apple.

GPUs that could do parallel computing mathematics at lightning speed were engineered by NVIDIA to make Tesla chips that put desktop workstations on par with supercomputers at one 100th of the price.

“You can’t ignore the GPU; you can’t say it is just a game chip,” Keane said. “The GPU is now a processor. It is not just for graphics.”

Supercomputers based on Tesla chips are being built by Dell and other NVIDIA partners with starting prices of less than US$10,000 – still out of reach of most consumers but far more affordable than existing supercomputers.

NVIDIA says the muscular machines enable researchers, software developers and others requiring massive computing power to handle heady tasks at their desks instead of having to queue for time at supercomputer clusters.

As supercomputing becomes more common, the technology is apt to work its way into consumers lives and products.

For example, supercomputers could be used to enable online picture searches based on face recognition or speed research into viruses.

“Hearing this kind of stuff, you get goosebumps,” Keane said while discussing what scientists might do with supercomputing at their desks.

“When you start distributing these tools to incredibly creative people, good things happen.”

Filed under:

Leave a Reply