2 min read

Intel, SGI To Work On NASA Supercomputer

The plan is to build a 1-petaflop system by 2009 and a 10-petaflop supercomputer by 2010.
Intel, NASA, and SGI have signed an agreement to work together on a supercomputer capable of 10 petaflops, or 10,000 trillion operations per second, by 2012. The high-powered machine will be used for modeling and simulation at the space agency's Ames Research Center in Moffett Field, Calif.

Under terms of the agreement announced Wednesday, the three organizations will first build a 1-petaflop system by 2009. The project for the two systems is called Pleiades, and builds on the 2004 deployment of Columbia, another supercomputer that was built by the three organizations.

Columbia generated a tenfold increase in supercomputing capacity for NASA, but falls short of what's needed for future space missions, the agency said. In 2009, the first Pleiades machine is expected to increase computing capability 16 times, with the second supercomputer providing an additional tenfold increase.

"This additional computational performance is necessary to help us achieve breakthrough scientific discoveries," Ames director S. Pete Worden said in a statement.

NASA last month marked 25 years in supercomputing at the Ames Center's NASA Advanced Supercomputing Division, which was founded in 1983 under the name of the Numerical Aerodynamic Simulation (NAS) program.

NAS supported the space shuttle's return to flight and subsequent missions. It also hosts simulations critical to the design of space exploration vehicles, Ares and Orion. NAS supercomputers support aeronautics research and simulate experiments in nanotechnology, combustion, atmospheric chemistry, ocean and climate modeling, hurricane prediction, solar and black hole physics, and cosmological structure.

The division recently installed the world's largest known visualization engine, the Hyperwall-2, which has more than a million times the graphics rendering power of the original workstations at NAS.