A key topic at this week's SC09 supercomputing conference, which takes place Nov. 14-20 in Portland, Ore., is how to reach the exascale plateau in supercomputing performance.
"There are serious exascale-class problems that just cannot be solved in any reasonable amount of time with the computers that we have today," says Oak Ridge Leadership Computing Facility project director Buddy Bland. Today's supercomputers are still well short of exascale performance. The world's fastest system, Oak Ridge National Laboratory's Jaguar, reaches a peak performance of 2.3 petaflops. Bland says the U.S. Department of Energy (DOE) is holding workshops on building a system 1,000 times more powerful. The DOE, which is responsible for funding many of the world's fastest systems, wants two machines to reach approximately 10 petaflops by 2011 to 2013, says Bland.
However, the next major milestone currently receiving the most attention is the exaflop, or a million trillion calculations per second. Exaflop computing is expected to be achieved around 2018, according to predictions largely based on Moore's Law. However, problems involved in reaching exaflop computing are far more complicated than advancements in chips. For example, Jaguar uses 7 megawatts of power, but an exascale system that uses CPU processing cores alone could take 2 gigawatts, says IBM's Dave Turek. "That's roughly the size of medium-sized nuclear power plant," he says. "That's an untenable proposition for the future." Finding a way to reduce power consumption is key to developing an exascale computer. Turek says future systems also will have to use less memory per core and will require greater memory bandwidth.
View Full Article
Abstracts Copyright © 2009 Information Inc., Bethesda, Maryland, USA