IBM next week will take a big step forward in its project to create the fastest computer in the world when a small-scale version of the Blue Gene/L finds its way onto the Top 500 list of most powerful systems.
This version of Blue Gene/L—about the size of a dishwasher, according to its chief architect—is about 1/28th the size of the full Blue Gene/L supercomputer IBM currently is building for the Lawrence Livermore National Laboratory in California. That system is scheduled to be completed at the end of 2004 or in early 2005, said Al Gara, chief architect for the project. Gara said he was surprised but pleased that the small-scale version of the supercomputer made it onto the list compiled by the Top500 Supercomputer project. The list will be released at the Supercomputing 2003 show in Phoenix. Blue Gene/L will be ranked 73 on the list, IBM officials said.
“Its a great precursor of whats to come,” said Gara, in Yorktown Heights, N.Y. “This is just the tip of the iceberg.”
Blue Gene/L is part of a five-year, $100 million project at IBM to create the worlds fastest computer. That distinction is currently held by the Earth Simulator in Japan, built by NEC Corp. When completed, Blue Gene/L is expected to have a peak performance of 360 teraflops—or 360 trillion operations per second—will have more than 64,000 dual-processor compute nodes and will be housed in 64 racks. The Earth Simulator has a peak of about 40 teraflops, Gara said. The prototype Blue Gene/L has a peak of about 2 teraflops.
The Livermore laboratory will use the full supercomputer to simulate physical phenomena, including cosmology, according to IBM.
Blue Gene/L will not only be the fastest computer, but also will cost less and consume less energy than other supercomputers, Gara said. The computer, which will run Linux and be powered by IBMs PowerPC 440GX processor, will give users a familiar feel, he said. It will offer a lot of parallelism, which will make it energy-efficient and smaller than other supercomputers, but also will mean less memory. However, it will still offer about 16TB of memory.
All that also will bring supercomputing capabilities into reach of researchers who now cannot afford it, Gara said.
“One thing were trying to find out is which ways we can use this and what we can learn, and have that make its way into a roadmap” for future supercomputers, he said.
Blue Gene/L is one of two prongs of IBMs overall Blue Gene project. Researchers are working on the design for Blue Gene/C, which Gara—who doesnt work on that part of the project—said probably will be more cost-effective and aimed at more specialized applications. Eventually IBM researchers will look at the two designs and then decide on the direction for Blue Gene/P, which will be a machine with a peak performance of a petaflop.