by William Hoffman
Like a good wine, Seymour Cray improves with age.
He has always stressed that speed comes with simplicity, so his new machine must be the simplest yet. The Cray 2 supercomputer now being tested runs through a cycle in four nanoseconds--that's four billionths of a second--and is capable of doing a billion calculations a second.
The Cray 2 is 6 to 12 times faster than the revolutionary Cray 1, has more than 10 times the memory, and takes up a fraction of the space. Cray has yet another machine in mind that will clip along at one nanosecond, the time it takes light to travel about a foot. Those who say it can't be done don't know much about Cray.
In this the computer age, Cray is at once the classic individualist and the ultimate engineer. His individualism is legend. He has lived and worked most of his life in Chippewa Falls, Wisconsin, keeping a low profile. He prefers slacks and sweaters to suits, dislikes press attention, is an avid wind surfer, and compares building computers to composing music.
In 1972, Cray founded his own company after twice bolting from larger computer firms when they no longer suited his technological style. But even Cray Research, Inc., now a $150 million-a-year company employing more than a thousand people, couldn't keep him in the front office. In late 1981 he resigned as chairman and became an independent contractor so that he could devote full time to engineering tasks. Under a contract with Cray the company has dibs on his new designs at least through 1985 and will provide support for his research.
Cray's prodigious engineering gifts can be accurately judged only by his peers, of which he has none. Perhaps closest is Gene Amdahl, formerly of IBM, founder of the Amdahl Corporation and Trilogy Ltd. and himself no small whiz at designing computers. Amdahl called Cray "the most outstanding high-performance scientific computer designer in the world." An official at the federal research center at Los Alamos, New Mexico, described him as "an outstanding national resource." Fortune magazine once said of Cray that "in a field where genius is almost taken for granted, he is a towering figure."
Cray has always been driven by science. He could be a wealthier man if he'd remained in the corporate hierarchy or if he'd taken full advantage of patent rights over the years. But he says money is a means, no more or less important then the metals, materials, and mathematics that make up a computational engine. Faster computers enable other scientists to push back the frontiers of knowledge. Cray finds satisfaction in that.
Computational speed--and that's what supercomputers like Cray's are all about--is essential to tackle some of the most difficult problems in science, those involving the behavior of moving fluids. Fluid dynamics can be simulated numerically in three-dimensional grids, but it takes a powerful and lightning fast computer to do it well. Parallel processing enables supercomputers to do many calculations at the same time rather than in sequence.
Supercomputers are used in weather prediction, aircraft and automotive design, seismic analysis, oceanography, petroleum exploration, and astro- and nuclear physics. They also have application in electric power distribution, integrated circuit design, and chemical and biomedical research. Their value to science and technology is hard to exaggerate.
Last year the Japanese embarked on a program to develop a computer more powerful than any now available by the end of the decade. Some U.S. scientists fear that loss of leadership in supercomputers would have dire consequences for the nation's technology and defense. Indeed, President Reagan's "Star Wars" defense system presumes U.S. leadership in supercomputers.
The so-called Japanese threat does not much concern Cray. He is confident that the best machines will continue to be the creations of engineers working independently, and there's no telling how many even faster computers he has tucked away in his imagination.
At the outset, Cray wasn't entirely certain he was in the right field. "I had a lot of doubts because things were pretty primitive in those days," he said in an interview for Update. "They taught things like electric motors, which I didn't care about.
"There were a few courses in electronics, but most of the instructors had their eye in the corner of the book all the time they were talking," he said, adding that one teacher, LeRoy T. Anderson, made a deep impression on him. "He was an outstanding instructor. I still see him once in a while."
Cray said he really appreciated the "broad aspects" of education at the University, particularly in mathematics. "I took another year and got a master's degree in applied mathematics. I found that very instructive because the engineering school at that time had a minimum amount of basic mathematics. I appreciated things like the theory of complex variables, which to me was wholly new."
There were no computers around when Cray graduated, but he went to work at a pioneer company in St. Paul, Engineering Research Associates, Inc., where he met Frank C. Mullaney, now a consultant, and William Norris, now head of Control Data. Mullaney supervised a project that Cray worked on.
"Seymour was involved in circuit, logic, and software design almost from the start," Mullaney said. "This combination of interests and abilities is rather unusual to find in one person."
The company was bought out by Remington Rand, which became Sperry Rand (now Sperry Corp.)., the country's first computer company. It was there that Cray worked on his first computer, an early machine that became the Univac 1103. In 1957, Cray, Mullaney, and Norris set out on their own and formed Control Data Corporation.
At Control Data, Cray designed the 1604 computer, which was "critical to the early success of Control Data," Mullaney said. It was one of the earliest machines to use transistors in place of vacuum tubes. Then Cray designed the CDC 6600, an early mainstay of scientific research, and the CDC 7600, the first supercomputer, which made its debut in 1969. He also designed the CDC 8600, but the company decided not to market it.
The company was pleased enough with Cray's designs early on that in 1962 it built him a laboratory within walking distance of his home in Chippewa Falls, but as Control Data few and diverted more of its resources into commercial applications, Cray felt increasingly constrained. He and several associates, including Mullaney, broke away and formed Cray Research with an eye to developing and marketing supercomputers and nothing else.
Cray was opposed to borrowing, so he sent his chief financial officer, John Rollwagen, to Wall Street, where the company went public in early 1976, raising $10 million overnight. A month later the company sold its first Cray 1 computer to Los Alamos at a price of $8.8 million.
Further sales the following year enabled the firm to show its first profit. Earnings for 1982 were $19 million on revenues of $141 million, increases of 38.9 percent and 4.6 percent respectively over 1981. Rollwagen, now chairman, attributes the slower-than-usual growth to price cuts on the Cray 1S system and to higher development costs for the Cray 2, according to a newspaper article.
Cray was more than happy to turn over his managerial responsibilities to Rollwagen and others, though it caused something of a stir among stockholders who feared a rift. Cray quickly went public with reassurances. After more than a year he feels his status as independent contractor has worked well.
"I think it's been very successful," he said. "The goal as you might imagine is for me to get rid of all the official functions--paper signing and things. That was accomplished well with the contract simply because I don't have to sign anything anymore. I'm still very involved in the company, and I think there's a feeling that it's a mutually satisfactory arrangement."
An article in Corporate Report magazine last December highlighted a company-wide effort to reaffirm Seymour Cray's values in an increasingly sophisticated and entrepreneurial environment. "The Cray Style is meant to convey something of the founding genius's inspiration," at a time of expanding markets and potential profits, the article noted.
Last year Cray Research closed down its research facility in Boulder, Colorado, not long after it announced that it would build the Cray 2 system. Early enthusiasm for the work at Boulder gradually waned once Cray got the technology for the new machine on track.
Which is one reason why Cray Research is probably not as vulnerable to a corporate takeover bid as some believe. Cray himself might not take kindly to new management, and even though "independent" he continues to be the company's chief asset.
The truth is somewhat less fanciful. Cray does retire t his cottage, where he can concentrate hours at a time in solitude. But computer building is a highly abstract exercise. He uses only pencil and paper--"about a pad a day" of 8 1/2 by 11 inch quadrille-ruled paper. His calculations are review by a 30-person development team, modified if necessary, and converted into a computer module of microcircuit chips.
What Cray calculates are Boolean equations. Invented by Englishman George Boole in 1854, Boolean algebra lay dormant until the middle of this century. Suddenly it became the most important mathematical tool in computer science.
That's because Boolean has some of the characteristics of a binary system, using only the digits 0 and 1, which in turn can correspond to the presence or absence of an electric pulse. Boolean lends itself admirably to the study of the vast network of switches, lines, and storage elements in a computer.
Cray said he never heard of Boolean algebra while he was at the University, but whenever it was the he started using it he was truly in his natural element--mathematical and systems logic. Cray has built into his machines the same virtue as the mathematical system they employ: simplicity. His machines add, subtract, multiply, and divide, and do nothing more.
Of course, it's not all that simple. Cray must know the electronic organization of the whole machine and its millions of circuit gates that must read either 0 or 1. Then he has to figure out what to do with the excess heat generated by the densely packed microcircuits to prevent their failure.
In the Cray 1, the problem of head was solved by building a refrigeration system into the machine. But the Cray 2 is much more densely packed, with the central computing section standing 26 inches tall and 38 inches long. Cray's solution: immerse it in an inert fluorocarbon liquid, the same liquid used as artificial blood. It is, in a sense, a computer in an aquarium.
Cray was able to pack the Cray 2 more densely because he is using three- rather than two-dimensional circuit modules, another of his innovations. No one can package computer modules better than Cray. The longest wire in the Cray 1 is four feet, the longest in the Cray 2 sixteen inches. Shorter wires make for faster processing because an electric pulse can travel no faster through wire than the speed of light. In the world of nanoseconds, a long wire is like an anchor.
That Cray can visualize an entire new computer in his head, build it, and have it perform as conceived is astonishing. Peter C. Patton, associate professor and director of the University Computer Center [now the Minnesota Supercomputer Institute], said that Cray reminds him of William A. Miller, who designed the Offenhauser racing car engine.
"Miller could design castings, have them made, machine them, assemble his engines, and expect them to work right the first time and produce the intended horsepower," Patton said. After more than 40 years of team development efforts in internal combustion engine technology, Miller's engine "is still a standard, a mark to be beaten," he said.
But now Cray is showing signs of flexibility if not downright enthusiasm. The reasons are metal-oxide-semiconductor (MOS) technology, which allows for a much-expanded computer memory, and a chemical compound called gallium arsenide, which Cray believes is the chip of the future because it has a distinct advantage over silicon in speed.
"The really dramatic thing that's happening right now is the relative performance of various kinds of memories," Cray said. "Because of personal computers the large MOS memory chips have really been pushed. It's quite dramatic what one can buy today in MOS memory compared to five years ago.
"Part way through the Cray 2 program I decided this was too good a thing to pass up, so I made the switch to MOS. We're still going through the mechanics of making that change," he said.
For the Cray 3 ("although we don't have firm designations for what we're going to call things") the company is building its own gallium arsenide integrated circuit plant. Cray admits it's "out of character," but he insists it's a desperation move.
We've been working at arm's length relationships with [chip] vendors for so long that you get used to that mode. But the turnaround times have now turned out to be five or six years instead of one year, which would be reasonable. It seems like all the [chip] companies get bigger and the turnaround gets worse.
"This will be the first time we've gotten involved in the actual integrated circuit fabrication. That's the whole scary world I never thought I'd find myself in. There's a great deal of worldwide enthusiasm right now for gallium arsenide. It seems like everyone's made a mistake or we've picked a good time to get involved," Cray said, adding that because basic materials are available, actual chip development "will proceed rather quickly."
The "Cray 3" will use the same basic architecture as the Cray 2, with individual processors about a cubic foot in size, Cray said. "We're taking about 16 of them in the multiprocessor configuration, so we tend to tend up with sizes that are like a cubic yard for the end product."
Cray said he expects the first Cray 2 to be shipped sometime next year. The machine will be sold "in multiprocessor configurations" of up to four, he said. A single processor, "which is a perfectly reasonable thing to have," should cost a quarter as much as the original Cray 1 (which has a single processor) and "have somewhat better performance."
Cray would have reason to be less optimistic today about that application of "number crunchers," as supercomputers are called, but he is pleasantly surprised at a new one.
Last year a Los Angeles film company leased a Cray 1 for "digital scene simulation," a high-quality film process. Computer-created films like Tron are catching on with moviegoers.
"That one came out of the woodwork," Cray said. "Apparently it's really going. I hear that other people are anxious to get involved."
The potential market for supercomputers has grown in recent years to several hundred clients, even though a buyer can expect to pay in the neighborhood of $11 million for a machine. Of some 70 supercomputers now in operation in the United States and Western Europe, abut 50 were manufactured by Cray Research. All but a few of the rest are Cyber 205s made by Control Data. The Cyber 205, released in 1981, is roughly comparable to the Cray 1 in speed and power. Each is better at solving certain kinds of problems.
The Japanese National Superspeed Computer Project, which involves six major companies--Fujitsu, Hitachi, Nappon Electric, Mitsubishi, Oki, Toshiba--together with the government-funded Electro-Technical Laboratory has alarmed portions of the U.S. scientific community. "Collaboration among manufacturers, academia, and government laboratories will be necessary" to maintain U.S. leadership in supercomputer technology, according to a report published last December in Science.
Cray is not amused. "I have really strong feelings about that," he said. "I feel the bigger the group that works on the project, the lower the chances for success. I'm appalled at our trying to make a country-wide coordinated effort. I just can't imagine it ever being successful.
"I believe you want a lot of independent people thinking their own thoughts and trying their own things. We're not going to participate in any national effort, and I don't want any money from the government. We've got competition within the company. I've got a group here five miles away who I know are trying to outdo me."
In Tracy Kidder's Pulitzer Prize-winning nonfiction book, The Soul of a New Machine, a computer engineer in the midst of putting together a research team comes across a videotape: "In the movie, an engineer named Seymour Cray described how his little company, located in Chippewa Falls, Wisconsin, had come to what are generally acknowledged to be the fastest computers in the world, the quintessential number crunchers.
"Cray was a legend in computers, and in the movie Cray said that he liked to hire inexperienced engineers right out of school, because they do not know what's supposed to be impossible."
When asked what has surprised him most about how his products are used, Cray replied: "I just design these things for myself. I'm always surprised when other people use them. I don't know what all this supercomputer talk is about. They certainly aren't supercomputers; they are kind of simple, dumb things."
But then, anyone who has to keep company with precise arrangements in strings of zeros and ones and in circuit modules can be forgiven for being facetious now and then. The fact is Cray is proud of what he does and determined to do it better yet. The only limitations are "the visions of the designer," he believes.
Cray escapes from the computer world several times a year. He and his wife vacation briefly in warmer climes, often in Mexico or the Caribbean. And on summer days, when the wind is up on Lake Wissota, Cray puts his pad and pencil aside, grabs his wind surfboard, and sets out to meet the challenge.
He alludes to his "getting old" on occasion, but, as might be expected, retirement is not in his program. "I hope I die with my boots on," he said. "I'm looking forward to that possibility."