dcsimg
RealTime IT News

Los Alamos Selects SGI for Supercomputer

The U.S. Department of Energy's (DOE) Los Alamos National Laboratory in New Mexico has purchased a Silicon Graphics supercomuter to help monitor the reliability and performance of America's nuclear stockpile without underground testing.

SGI's Onyx4TM UltimateVision, which the Mountain View, Ca.-based company calls the world's largest and most powerful advanced visualization system, is an 80-processor, 34-pipe system and offers "many times the visualization capability" of any other computing system available.

Scientists at Los Alamos will use the SGI system as part of the DOE's Stockpile Stewardship Program, a key component of which is the Advanced Simulation and Computing Initiative (ASCI), which is designed to accelerate the development of the computational power DOE scientists and engineers need to meet the visualization needs of virtual testing.

One of the most pressing problems facing LANL's ASCI program is managing and manipulating the terabytes of data required to transform data sets into visual representations. Traditional graphics-focused solutions are unable to meet the vast data requirements essential for visualization.

However, the new SGI system's scalable system architecture and large shared memory are specifically designed to rapidly digest multi-terabyte data sets while focusing dozens of microprocessors and graphics processors to tackle problems many times larger than ever before possible.

"Researchers at Los Alamos are confronted with enormous data sets. The scalability and throughput of Onyx4 provides new technology to visualize those data sets, and can rapidly deliver granular-level results," said Bob Tomlinson, simulation support manager of Los Alamos' ASCI Program.

The U.S. suspended underground nuclear testing in 1992 and established the Stockpile Stewardship Program in 1995 to continuously monitor the condition of America's nuclear weapons stockpile, assess the findings, and perform maintenance and refurbishment as needed.