Hewlett-Packard has bagged a $24.5 million Linux supercomputer contract with the Department of Energy, in part because the system features fast connections among its 1,400 CPUs, says a department computer scientist.

HP is building the supercomputer--benchmarked at a peak speed of 8.3 trillion floating-point operations per second--at the department's Pacific Northwest National Laboratory in Richland, Wash. Installation should be complete by year's end. The lab plans to use the computer to study toxic-waste storage, materials design, geology, and genomics problems that require low latency between processors. "This is the first major Linux supercomputer," says Scott Studham, technical group lead at the laboratory's molecular sciences computing facility. "A lot of this system's resources come from HP spending a lot of time on the storage area network. It's very tightly coupled."

The system will run on upcoming Intel 64-bit Itanium chips code-named McKinley and Madison. It will replace an IBM RS/6000 SP system that, when purchased five years ago, was among the world's fastest. That system now ranks 190th on Top500.org, a list of the world's fastest supercomputers. The new HP system, if live today, would rank second on the widely watched list, Studham says.

HP, the world's No. 2 computer maker, says customers who need high-performance technical computing increasingly want to run their programs on industry-standard Intel servers, which can help pare operational costs. HP says the contract shows off the company's expertise in technical computing. Says Martin Fink, general manager for Linux systems at HP, "This deal demonstrates to the world that there's more than one player" in that market.