Questions Follow U.S. Buy of NEC Computer

The decision by the National Science Foundation last week to select a Japanese supercomputer for climate research has sent shock waves through the federal research establishment and raised serious questions about the competitiveness of U.S. firms.

It marks the first time that Japanese suppliers have beat long-dominant Cray Research in a competition for a major federal procurement, showing that the United States’ long superiority in supercomputing technology is now facing a serious challenge.

“What it indicates is that we are in a new era,” said Walt Brooks, the director of advanced computing and communication at the National Aeronautics and Space Administration’s Ames Center in Sunnyvale, Calif. “It is a wide-open competition now. Japanese processors are faster, so their fundamental building block is faster.”

Forest Baskett, chief technology officer for Silicon Graphics Inc., which purchased Cray Research in late February, called the decision “unfortunate.” However, he maintains that the United States continues to have a strong lead in supercomputer technology despite this win by Japan’s NEC Corp.

“It’s a complicated situation,” Baskett said. “The Japanese supercomputers are fast at processing a long string of numbers, but they are not very flexible when dealing with more complex problems. “We believe our approach will ultimately prevail,” he added.

Although the supercomputer industry is a relatively small and obscure sector of the U.S. electronics business--dwarfed by the market for personal computers, for example--it is widely regarded as a cornerstone of U.S. competitiveness.

The NSF decision is another blow to an industry already staggering from self-inflicted wounds as well as from cuts, brought about by the end of the Cold War, in government funding for supercomputer research. Not only has the government money used to underwrite supercomputer design evaporated, but also the universities, government agencies and national research laboratories--the customers for these machines--have seen their budgets shrink. In the last three years, four highly regarded U.S. supercomputer companies have filed for bankruptcy protection, including Cray Computer Corp., started by legendary supercomputer designer and Cray Research founder Seymour Cray.

Supercomputers are crucial to the design of aircraft and jet engines, not to mention other computers. The nation with the best supercomputers can decode other nation’s secrets, predict the weather with greater accuracy and better unravel the mysteries of genetics.

Moreover, the ability to design supercomputers--the fastest computers--has always been assumed to create a trickle-down effect that benefits leadership of everything from microprocessors to personal computers.

So important is the matter that some U.S. experts believe that NEC is illegally selling its computers below cost to gain a foothold in the U.S. market.

The Department of Commerce has reached a preliminary finding that there is evidence of dumping by NEC. If so, NEC may eventually have to pay special duties, but it would still gain the foothold that it is seeking.

Federal agencies have long depended on Cray to provide new generations of computers first to U.S. users and only later to foreign users, keeping the technical advantage on U.S. soil.

But if Japanese suppliers eventually develop superior machines, that advantage would eventually go to Japanese users, forcing U.S. laboratories to wait months or years for the newest machines.

The NEC computer would be used by the National Center for Atmospheric Research in Colorado and operated by a consortium of U.S. universities, with about 90% of the funding provided by the NSF.

Lawrence Rudolf, NSF general counsel, said the only criterion important to center was which computer could calculate its set of equations fastest, thereby making U.S. climate research preeminent in the world.

“We were not weighing national interest here, but we were evaluating the singular interest of our scientists to be at the cutting edge of climatological research,” Rudolf said.

Rudolf said no federal policy or law allowed the NSF to tilt the competition in favor of Cray. Indeed, federal laboratories--the biggest customers for supercomputers--are under such tremendous budget pressure that they are not inclined to do any favors for U.S. corporations.

Bill Buzbee, director of the scientific computing division at the atmospheric research center, said that the NEC’s machines operated substantially faster than did those of other candidates in tests run before the decision was made.

Asked if the federal government is doing enough to ensure U.S. leadership in supercomputing, Buzbee said Sunday: “We need to take a hard look at that. We may need to do more.”

Buzbee said that although the U.S. has not been eclipsed by the Japanese, it is for the first time seeing a real challenge. He noted that three major climate research labs around the world (including the Colorado center) have purchased supercomputers in the last year, with the awards going to Cray, Fujitsu and NEC.

Nonetheless, many senior leaders said the decision is a watershed.

“It is a very surprising decision,” said one senior federal technology official. “These people don’t have any loyalty to brand or country. They are like any consumer in the market who is pursuing their interest.”

NASA will be asking supercomputer vendors later this year for proposals for a new system the agency wants to install next year, Brooks said.

In many regards, the fluid dynamics calculations done for weather research are similar to the equations used in aerospace research. In both disciplines, researchers are attempting to create sophisticated models of how air behaves.

NASA ultimately wants to build a system that would enable aircraft designers to create a three-dimensional electronic model of an aircraft that would yield clues to its aerodynamic characteristics.

With both Europe and Japan pursuing the same goal, the nation that obtains that ability first would have a competitive advantage in designing aircraft. At the same time, however, NASA’s budget for supercomputers is being slashed by 50%, greatly increasing the pressure on the agency to make the right decision.

Brooks said that both Fujitsu and NEC processors operate about 50% faster than the basic processor used by Cray, giving the Japanese an edge in raw power. But Cray has designed faster systems for accessing memory, which has historically won competitions for the company.

* Vartabedian reported from Washington and Pitta from San Francisco.