ORNL Plans to Make 'Jaguar' Faster

The Cray XT3 supercomputer known as "Jaguar" already is one of the fastest machines in the world, but Oak Ridge National Laboratory researchers have plans to make it even faster. The Department of Energy facility hopes to quadruple Jaguar's speed by next year, from 25 teraflops to 100 teraflops, or 100 trillion calculations per second. "It's going to be a challenge, but I think we're going to get close," said Thomas Zacharia, associate lab director and computer chief. Part of the challenge is federal funding for the National Center for Computational Sciences at ORNL. Congress boosted the 2006 funding from the $25 million proposed by the Bush administration up to $55 million, and Zacharia said Oak Ridge officials were thankful for support of the leadership-class computing initiative. But the funding level is still short of earlier targets, he said. The Cray XT3 is ranked as the 10th fastest computer in the world and the fastest U.S. machine available for open scientific (nonmilitary) research. The lab plans to expand the supercomputer from its configuration of 56 cabinets to about 120 cabinets and double the number of processors. To date, the Jaguar has yielded the first 3-D numerical simulations of an ignition flame, which should help in the design of next-generation combustion devices, and uniquely precise views of molecular structures that may be used to combat genetic diseases. Supported by DOE's Office of Science, Zacharia is taking an important step toward investigating computer architectures for scientific discovery, even as ORNL deploys the system to test its effectiveness in solving important scientific problems in climate, biology, nanoscale materials, fusion and astrophysics. "Things are going very well," said Zacharia, who was honored with a top individual award earlier this month at the World's Supercomputing Conference in Seattle. "I think Oak Ridge is the place to be to do science in the United States," Zacharia said. For more information, go to the Oak Ridge National Lab Web site.