Seymour Cray Charts New Waters in Supercomputers
The story goes that Seymour Cray builds a new sailboat for himself by hand every year. But at the end of each summer, Cray burns the boat to the ground.
He does so, those who know him say, because he doesn’t want to become a prisoner of his old boat design. He wants to start with a blank sheet of paper the following year.
For Seymour Cray, the 63-year-old father of the supercomputer and the founder of Cray Research, that story is an allegory for his life’s work. Throughout his long and legendary career--which roughly parallels the history of the computer--Cray has always been obsessed with building the world’s fastest computers.
And he has done so time and time again. Most computer scientists alive today have never known a time when the world’s fastest, most powerful computing machine used in advanced scientific research wasn’t a Seymour Cray design.
“It’s like saying one man has designed all of the fastest aircraft ever built,” observes Neil Davenport, vice president of Colorado operations for Cray Research.
“I’m 40, and Seymour Cray has designed every scientific computer I’ve used my entire life,” adds Larry Smarr, director of the National Center for Supercomputing Applications at the University of Illinois.
“He is the Thomas Edison of supercomputers,” Smarr adds. “In retrospect, we will look back on Seymour Cray the way we look back on Edison. He has transformed society.”
But after each new machine, each time that Cray has successfully pushed the outer limits of computer technology, he has started over, seeking new challenges.
This time, he is beginning anew with both a revolutionary design and a new company. Cray has just split off from Cray Research, the Minneapolis-based company he founded 17 years ago, to form Cray Computer in Colorado Springs, Colo.
He is pursuing the enormously risky goal of once again leapfrogging existing computer technology with a powerful new super machine, the Cray 3, which industry watchers say will be about the size of a small coffee table. With his new computer, Cray is attempting to revolutionize computing by dispensing with the silicon integrated circuit, the basic component in all computers today, the chip that made Silicon Valley famous.
New Chip Material
Instead, he is attempting to become the first to build a computer based on chips made of an untried material, gallium arsenide, which is extremely brittle to fabricate but could lead to a computer three times as fast as the fastest silicon-based machine.
To do all this, he had to leave Cray Research, which is spinning off Cray Computer to its shareholders. His old firm decided to concentrate on its more cautious efforts to develop a new and extremely fast generation of silicon-based supercomputers.
So Seymour Cray is once again staring at a blank piece of paper, just as it is said he does with his boat each year.
Whether that sailboat allegory is true is hard to determine, because Cray almost never talks to the press, industry analysts or, for that matter, anyone outside a small coterie of computer designers and supercomputer users. He is, in fact, a recluse, a single-minded genius who has consistently shunned the fame that his prodigious contributions to American high technology should have brought him. As a result, he remains a virtual unknown outside the computer community, which is apparently the way he likes it. He not only rejects interview requests, he once even refused to show up at the White House to receive a national science award.
“He avoids publicity at almost all costs,” notes William Graham, President Bush’s science adviser, who knows Cray.
But Graham and others who know him argue that Cray may be the most important non-famous person in America, a man who almost single-handedly created the supercomputer industry, perhaps the highest of today’s high-technology fields. He has also kept the United States in the lead in supercomputers--which are used for complicated tasks such as simulating nuclear reactions, designing aircraft wings and cracking secret codes--despite increasing competition from the Japanese.
‘Like Hero Worship’
Indeed, within the tightknit world of computers, he is widely revered. When he walks into a computer conference, he often receives a standing ovation. “It’s like hero worship,” observes computer industry analyst Gary Smaby of Needham & Co.
Top computer scientists remain in awe of him for his legendary ability to hold an entire supercomputer design concept in his head. “When you talk to him, you feel like you don’t want to talk long because you don’t want to waste his time,” says Smarr of Illinois. “You wonder why you are wasting his time because it is hard to come up with something that he hasn’t already thought of.”
In an increasingly complex field dominated by vast teams of designers, Seymour Cray may be the first--and last--man to design supercomputers essentially on his own.
“I had to introduce Seymour at a conference and I kept trying to think of a way to do it,” recalls Norman Morse, division leader for computing and communications at the Los Alamos National Laboratory in Los Alamos, N.M., which used Cray Research’s first supercomputer in the 1970s. “First, I was going to say he is like the Albert Einstein of supercomputing or the Thomas Edison of supercomputing. But then someone in my group said I should say he is really the Evel Knievel of supercomputing, because he has pushed the boundaries beyond where anyone thought we could go.”
Doing just that is what has motivated Cray since he left the University of Minnesota with a master’s degree in applied mathematics in 1951.
After graduation, he went to work for a new Minneapolis company, Engineering Research Associates, which eventually became a part of Sperry-Univac, where Cray designed some of the early vacuum tube computers. At Sperry he designed the Univac 1604, which at the time was the world’s most powerful computer.
Helped Found Control Data
In 1957, he left Sperry to help found Control Data, where he eventually solidified his reputation within the computer industry as a brilliant computer designer. There, he developed the Control Data 6600 and later the 7600, the fastest computers of their times.
But by the early 1970s, Cray felt that Control Data was losing its focus on fast computers designed specifically for the scientific community, and he left in 1972 to found Cray Research. When his Cray 1 finally came out in 1976, it was 10 times faster than his Control Data 7600--which at the time was still the fastest computer on the market.
But Cray had no interest in running his own company--he just wanted to make supercomputers. So in 1981, John Rollwagen, originally Cray’s vice president for finance, took over as chairman. Cray officially became a consultant to the company but continued to run the firm’s laboratories in Chippewa Falls, Wis., Cray’s hometown.
In Chippewa Falls, he often worked out of his home, where he sometimes--so the story goes--cleared his mind by working on a deep tunnel he dug by himself in his back yard.
Since last year, he has worked out of Cray’s new research center in Colorado Springs, a facility with a work force of about 200 that will form the basis for his new company.
But Cray is not leaving to run his own business again. Davenport, now the vice president for Cray Research’s Colorado operations, will do that. Instead, the new arrangement, Cray hopes, will let him focus on what he does best and do it as quickly as possible. He has said the first Cray 3 will be ready for delivery in 1990.
With Cray gone, analysts wonder what will happen to Cray Research, still the nation’s premier supercomputer company. Although Cray himself never sought publicity, analysts and reporters who follow the firm say Rollwagen often invoked the Seymour Cray legend to help sell the company to the computer community. The firm will now have to sell its customers on the talents of a more anonymous and less colorful design team.
Perhaps the most easily understandable barometer of Seymour Cray’s importance came May 16, when Cray Research announced that Seymour was leaving. On that day, Cray Research’s stock fell $6 a share, an 11% decline.
