NEWPORT, R.I.—The Energy Department’s Oak Ridge National Laboratory will start to take delivery on a computer at least as three times more powerful than any now in operation.
The system should be operational by 2008, according to Thomas Zacharia, ORNL associate lab director for computing and computational sciences. Zacharia spoke at the High Performance Computing and Communications Conference this week in Newport, R.I.
Cray Inc. of Seattle will supply the system, named Baker. It will run approximately 24,000 2.6 Ghz quad-core Opteron processors made by Advanced Micro Devices Inc. The nodes will be housed in 187 liquid-cooled cabinets. The system, which is still in early stages of design, will have either 187 or 400 terabytes of working memory (depending on the cost of memory modules) and from one to 11 petabytes of storage.
Such a computer, if operational today, would be considered the world's fastest; it may even be the first to break the 1 petaflop limit. A PFLOP is 1 quadrllion floating-point operations per second.
According to the Top500.org list, today's most powerful computer is the Blue Gene/L, run by Lawrence Livermore National Laboratory, another Energy Department outfit. That system currently runs at about 350 teraflops (350 trillion floating-point operations per second), though the lab and IBM Corp. plan to expand its capability to 1 PFLOP, also by 2008.
The laboratory will rely on the Tennessee Valley Authority for the power needed to run this behemoth. The power company is building a 170 megawatt substation to support this and other ORNL projects.
Like current ORNL systems, this one will be used by some of the country's best scientists to further explore the limits of their fields. Researchers must demonstrate the need for using the computer, and the lab will look for those jobs that are too large for other systems.
"We're entering new realm of scientific discovery, where we are looking at things at a much smaller scale. Computing expands upon our understanding of the natural world," Zacharia said. "Greater computer capacity allows researchers tackle more complex problems."
Read the Entire Article here