TOKYO — Sequestered inside a well-guarded compound on the edge of Tokyo, Mikio Watanabe's SX-2 supercomputer cuts a singularly unimpressive figure.
For starters, it is bulky and gray, with all the visual appeal of a filing cabinet.
Watanabe, an engineer at NEC laboratory here who has spent every working moment for the last seven years on this numerical Goliath, senses his visitor's disappointment.
"OK, comparisons speak for themselves," he says. Out comes a scrap of poster board covered with numbers comparing the performance of the SX-2 with its Japanese and American competitors. It reads like a paean to the SX-2.
Fastest in World
"This," Watanabe announces with unconcealed pride, "is the fastest computer in the world."
In the race to develop ultra-high speed computers, Japan doesn't plan to be left in the dust.
Five years ago, the supercomputer field was primarily the stomping ground of Cray Research, an American firm. Then Japan launched its national superspeed computer project. In the ensuing three years, Japan has narrowed much of the lead the United States built over the last 15 years.
Japan now boasts the world's fastest computer, capable of executing 1.3 billion separate calculations each second. Its production and marketing know-how has helped poise the market at the edge of decisive expansion. U.S. technological supremacy in this field is being challenged for the first time.
"The era of playing catch-up is ending," says Genya Chiba, research director of National Research & Development, one of the government research organizations included under the national supercomputer project's umbrella. "Now we are the ones to be caught up with."
Effect of Largess
Super-computers are the Arabian steeds of the information-processing industry. They are expensive--$5 million to $15 million apiece--and fast. As a result, both in terms of manufacturers and clientele, the world supercomputing market became a highly exclusive club. A corporate duopoly comprised of Control Data and Cray Research dominated the field from its early days in the 1970s. From the first, its best customer was the U.S. government.
"The supercomputing industry in the U.S. was nurtured by the largess of DOD (U.S. Department of Defense) and DOE (Department of Energy)," says George Lindamood, a technology watcher in Tokyo for Burroughs.
Since supercomputers were being used for highly specific tasks, there seemed to be no need to compromise their performance by designing them to work with the general-purpose mainframe computers employed by many large businesses. So Cray and Control Data produced highly specialized machines for a narrow little market that few other manufacturers appeared interested in.
Enter Japan. Through its computer manufacturers, it is instigating a wholesale transformation of the industry. Their thrust in the field is based on the conviction that virtually all the industries critical to its economic future--among them biotechnology, semiconductor development and aerospace science--will increasingly demand the sort of computing power that only supercomputers can provide.
"Supercomputers will become the workhorses of the '90s," says Shoji Tanaka, a University of Tokyo semiconductor physicist who helped launch his country's drive to catch up with Western semiconductor technology in the 1970s.
The Japanese also see supercomputers as the guinea pigs on which future generations of computer devices will be tested. Faster computers will require faster electronic circuitry attainable only through the development of new microelectronic technologies. So not only will the actual computing power of supercomputers prove vital to the design of advanced microcircuits, explains Fujitsu's supercomputer project leader, Hiroshi Uchida, but "their very existence will help extend the state of the art."
New Type of Contestant
Japan is also bringing a new type of corporate contestant into the race. Unlike the U.S. firm Cray Research, the Japanese companies--Fujitsu, Hitachi and NEC--are all multifaceted electronics companies. Each boasts its own line of computers, from personal-sized machines to big mainframe computers. And each is among the world's top 10 semiconductor manufacturers.
At the same time, they have embarked on a global strategy to take advantage of IBM's absence from the supercomputing market. Fujitsu and Hitachi have designed their machines to be compatible with IBM equipment--meaning that users of the Japanese computers can therefore move to more powerful machines without the trouble and expense of adapting to a special system.