Fujitsu Unveils Innovative Supercomputer : Technology: The Japanese firm is hoping to compete head-to-head with American companies.

Upping the stakes in one of the most politically contentious areas of high technology, Fujitsu Ltd. of Japan on Wednesday unveiled an innovative supercomputer that could challenge the United States’ longstanding lead in high-performance computing.

Japanese companies including Fujitsu, NEC and Hitachi have struggled for years to compete with market leader Cray Research and other American firms in selling the extremely powerful computers, which can solve complex problems in geology, chemistry, atmospheric science and other areas.

Though they have done well in Japan and Europe, the Japanese have never succeeded in selling systems in the United States. They complain that the United States has a de facto ban on foreign supercomputers, particularly in the government sector, which accounts for a large portion of supercomputer sales. But Cray contends that it is Japan that discriminates; the American company has only 20% of the Japanese market.

The new Fujitsu machine is important because it marries traditional supercomputer designs--pioneered by Cray--with new parallel processing technologies. Parallel processing involves tying hundreds, or even thousands, of computer processors together to attack a problem simultaneously, and many experts consider it the wave of the future in supercomputing.

American companies such as Intel, Ncube and Thinking Machines have developed machines that use thousands of comparatively simple processors. But the Fujitsu design calls for between seven and 222 high-powered processors, comparable to those used in a Cray, to be tied together.

This approach makes it possible to use modified versions of existing supercomputer software, rather than writing complex new programs from scratch. Cray and IBM are pursuing similar designs, though they have not yet announced products.

Fujitsu’s vice president for U.S. sales, Thomas M. Miller, said the company will concentrate on selling machines to the petroleum, chemical and pharmaceutical industries rather than fight a seemingly hopeless battle for government contracts. Just last week NEC lost out to Cray in a bitterly contested fight for a $75-million NASA supercomputer contract.

Gary Smaby, a Minneapolis-based supercomputer consultant, said: “Commercial customers are much more likely to make a decision based on price/performance considerations.”

And if Fujitsu’s new machine performs as advertised, he said, it should be a viable competitor for business in the United States as well as an attractive option for existing Fujitsu customers elsewhere in the world.