In a profile of Seymour Cray that ran on May 15 in The Washington Post Magazine, an incorrect name was given for a founder and the current chairman of the board of Control Data Corp. His name is William Norris.
Most every day, in a Midwestern town surrounded by corn fields, a middle-aged computer engineer emerges from his house carrying sheets of quadrille paper covered with mathematical equations. He takes them to a nearby laboratory where he is designing an extraordinary machine, a mass of wire, copper and silicon submerged inside a Plexiglas vat, with blinking lights and bubbles rising eerily among the circuits. It's a supercomputer, a so-called "number-cruncher" capable of making millions of calculations per second. Both it and its maker are indirectly linked to the security of the United States, but that is only a small part of the story.
The computer will be the fastest on earth, and will cost about $15 million when it is marketed next year. Already customers are lining up. Other computers designed by the same man--and looking like round Victorian settees with metal backs--sit in intelligence and defense establishments in Washington; no official will say how many. The government is such an important customer that Cray keeps a permanent representative in Washington.
Cray computers inform the calculations of astronauts, weather forecasters, aeronautical engineers, nuclear physicists, oil men and movie producers. The machines have a capability beyond most humans' imagining, in a tenuous realm where nanoseconds can mean the difference between profitablity and decline, and perhaps between life and annihilation.
The man is Seymour Cray, 57, the J. D. Salinger of CPUs (central processing units), a recluse who almost never grants interviews and has twice resigned from flourishing companies he helped found. Cray seldom leaves Chippewa Falls, Wis., his home town and work place, except to ski precipitous slopes or to scuba off Australia's Great Barrier Reef-- pursuits that make company executives and government bureaucrats uneasy. Now Cray's--and the United States'--number-crunching supremacy is threatened by the Japanese, who are dumping $200 million into an effort to surpass the man from Chippewa Falls and everything he represents.
His new computer, the Cray-2, will have a 256-million-word memory and a computing speed of about 2 billion transactions per second. (A mere 800 million calculations per second is roughly equivalent to 154 persons working with pocket calculators 24 hours a day, seven days a week, for an entire year.) At least one of the Cray-2's predecessors--and some say many--sits in the basement of the National Security Agency at Fort Meade, Md.; another is in NSA's Communications Research Division of the Institute for Defense Analysis at Princeton University. The NSA is so proud of its Crays that it pictured one on the cover of a recruitment brochure.
Los Alamos National Laboratory in New Mexico bought the first Cray-1 ever made, in 1976. The scientists liked it so well that they have since bought four more. The Crays are used to simulate atomic blasts, alleviating the need for real testing, and to lead physicists deeper into fission's mysteries. The Lawrence Livermore Laboratories in California has four Cray machines, the National Center for Atmospheric Research in Boulder, Colo., has two, as does the National Aeronautics and Space Administration. ARCO, Exxon, Chevron, Shell and Sohio have one each; so does Bell Laboratories in New Jersey. What all these users share is a need for high-speed correlation of an almost infinite number of mathematical variables.
A supercomputer can microcook every bit of pertinent information regarding, say, the secret detonation of a nuclear device in the south Atlantic, and serve it up for instant analysis. Theoretically it can assist in monitoring the millions of conversations that pass electronically through the atmosphere each second.
By quickly computing all possible physical variations, a supercomputer can replicate a three-dimensional model of an improved airplane wing that may save billions of dollars in fuel costs. It can produce computer models of global cloud patterns, brimming oil reservoirs a mile beneath the earth, the intensity of a laser or the hidden aberrations of your own heart.
Hollywood's Digital Productions recently bought a Cray X-MP, a refinement of the Cray 1, for $12.6 million to create full-screen computer simulations. The company will provide action scenes for a new film called Starfighters, a mere intimation of what's coming: A company official believes "we will eventually be able to simulate the physical morphology of humans." This advance theoretically could do away with human actors.
It all goes back to Cray, a legendary eccentric in an industry full of them. Until recently Cray designed and built his own sailboats, then burned them at the conclusion of the sailing season so he would have an excuse to design and build more. He once told a shareholders meeting that his company's stock was overvalued. An introverted student of mathematics and electrical engineering from the University of Minnesota, he went to work for the predecessors of Sperry Univac in the early 1950s with a vision of a computer that would perform perhaps 200 million calculations per second. But the strictures of a burgeoning corporation grated on him. He quit.
With Frank Norris and other entrepreneurs, Cray cofounded Control Data Corporation, today Cray Research's chief domestic rival in supercomputer technology. Cray became the director of CDC, a curious title for a man who worked in a laboratory in Chippewa Falls, leaving management details to executives in Minneapolis. Cray's purview was computer design and Boolean algebra. He created an early transistorized computer with unprecedented speed and capacity that scientists clamored for. Working with other engineers, he produced a bona fide supercomputer--a mere 36 million calculations per second--in 1969.
But Control Data was becoming too big for Seymour Cray, and the firm was interested in commercial applications, so Cray resigned in 1972 to form his own company. He objected in principle to raising capital by borrowing money; Cray Research floated stock on the basis of its founder's genius and an insurance policy on his life with Lloyd's of London.
Cray's business practices had more in common with Eli Whitney than with Henry Ford: He insisted upon building and selling one machine at a time. Each development had to be a compatible extension of an existing machine, and all those machines had to be supercomputers. Small groups of innovative engineers worked on specific projects, bearing sole responsibility for them. An unusually large amount of revenue was put back into research, in a new facility a football field's length away from the old Control Data lab.
In 1975 Cray had lunch with a bright young Minneapolitan who had an engineering degree from MIT and a graduate degree in business from Harvard. His name was John Rollwagen. At the time, Rollwagen was a vice president of International Timesharing Service who had once worked for Control Data, the mother ship for supercomputer talent. Cray recognized a rare blend of technical and administrative ability. After lunch, he said, "Why don't you think about it?" "Think about what?" asked Rollwagen. "About coming to work for us," said Cray.
Today Rollwagen is president and chief executive officer of Cray Research. The founder is just a consultant to the firm, albeit a highly paid one. Some of what drove Cray from Control Data--the administrative detail necessary in running any successful corporation--also drove him from the tiller of his own company. His research facilities are provided by Cray Research, which retains options on all his technological developments. Cray is not part of the company's hierarchy, but he is part of its soul.
That poses unique problems for Rollwagen, who must deal with Cray's legacy and his ongoing influence on the company's product. Rollwagen isn't bound to respect Cray's wishes, yet he talks to him by telephone every week. Their relationship, according to observers, is mutually supportive. Cray is known for his unconventionality inside an unconventional company. For instance, shortly after moving into his new lab, Cray hired as his personal secretary a waitress in his favorite Chippewa Falls restaurant instead of an experienced Cray employe. But Cray's overriding interest is supercomputers, and friends say he is pleased that his creations contribute to the stability and security of the nation.
Cray Research is a public company, listed on the New York Stock Exchange. No individual or group has a con- trolling interest (the largest block of stock, 2.1 percent, is held by Cray). Stockholders must be convinced that 15 percent of revenue should go into research, another of Rollwagen's tasks. The company is bound up with technological advance and, in his view, with national pride. America is presently the leader in supercomputer construction, but the Japanese intend to take over the market. Both Hitachi and Fujitsu have announced that they will unveil machines faster than Cray's this year.
"It's a shot across our bow," says Rollwagen, who doubts that the Japanese machines will be as fast as a Cray. "We'll see how big a splash it makes."
Cray Research occupies part of the old Pillsbury Building in downtown Minneapolis. Its renovated interior space includes "pods" --three working offices around a communal sitting area, with sliding glass doors and soft earth colors. Rollwagen shares his pod with the company's chief financial officer and legal counsel. He works at a stand-up desk. On the walls hang an abacus and photographs of the mountains Rollwagen has skied. The titles on his bookshelf include Design and Behavior, The Soul of a New Machine and former CIA director William Colby's Honorable Men.
Rollwagen still remembers hearing the bleep emanating from Sputnik broadcast over the radio in 1957. "That sound got me into electrical engineering." He has an easeful authority suited to the sparkly, still-youthful computer industry.
He has codified something he calls "the Cray style," a statement of corporate identity that reflects the individualism of the company's founder. For instance:
At Cray Research, we take what we do very seriously, but don't take ourselves very seriously.
"I want to keep this an open, creative environment where unusual things happen," says Rollwagen. "We don't have a security system --the need-to-know system is terrrible! The best stuff is so simple that you can't keep it down, anyway. It's much better to be off-the-wall, where everybody knows everything."
People also have fun at Cray Research. There is laughing in the halls as well as serious discussion.
"If Hitachi wants to know our stuff," Rollwagen says, laughing, "they can come and get it . . . That would put them three years behind us."
Cray people feel that they are on the winning side. They feel successful, and they are.
"There's a lot of responsibility. You have to have guts to work here. There's no support system, no place to hide. We have to forget what we did before and start over again. That's scary for some people."
The Cray approach is informal and non-bureaucratic . . . "Call--don't write" is the watchword.
"We think the best work isn't done with big teams and lots of money, but with small, adventuresome groups and a limited amount of capital," says Rollwagen. "The environment works."
In Chippewa Falls, 100 miles east of Minneapolis, the company built Seymour Cray's research facility, called Riverside. Located aross town from the company's main industrial plant, it is the official off-the- wall entity. Seymour Cray's decision to cool his latest creation, the Cray-2, in an inert liquid fluorocarbon was considered radical by his competitors. Equally radical is Cray's decision to develop his own microchip out of something called gallium arsenide, and escaping dependency on Japanese chip suppliers.
Cray Research is the second largest employer in Chippewa Falls (first place goes to an automated shoe factory), a town of 17,000 where ice-fishing and wind-surfing are considered more important than high-speed circuitry. In one of three prefabricated industrial buildings east of town, women in blue coats use mi- croscopes to solder some of the 6 million connections that go into one Cray computer. Each machine made by Cray contains about 70 miles of wiring and takes about 12 weeks to build and 18 weeks to check out. The company sells about one a month.
The company is essentially developing two separate lines of compatible computers. Its engineers stay a few billionths of a second ahead of the competition. Here an engineer from Taiwan named Shyh-shing "Steve" Chen designed the Cray X-MP, the first computer developed at Cray Research without the help of Seymour Cray himself. The X-MP is three to five times faster than the Cray-1 and uses many of the same components, including Freon circulated in metal tubes, as in a refrigerator, to keep the machine from burning up during operation.
Chen may eventually replace Cray as the company's preeminent computer designer. "Our next machine," he says, in an office piled with papers, the products of his research, "will be as fast as a Cray-2, but with different architectural features." He hopes it will be faster than a Cray-2, using less radical components. For instance, he will continue to use silicon chips, and Freon, instead of bathing his creations in fluorocarbon.
"The important thing for us as a company," Rollwagen says, "and as a natural resource for the country, is to eventually develop supercomputers without Seymour. We have to show we can do it."
How does a small American corporation stay ahead of its domestic and foreign competition, in the gathering tempest of supercomputer technology?
A colleague of Cray's tells an illustrative story about Cray when he worked for Control Data. The engineers were required to produce book-length reports of their projects, but Cray's report consisted of only one single page. Under the heading, Marketing Strategy, he had written, "Proceed with confidence."