American Scientist

The Supermen: The Story of Seymour Cray and the Technical Wizards behind the Supercomputer. Charles J. Murray. 232 pp. John Wiley, 1997. $24.95.

The Supermen is more than the biography of Cray and the other "wizards" implied by the subtitle. It traces the main line of supercomputer development from the early 1950s until Cray's death in 1996. Thus it covers the main developments in the field involving the hardware transitions from vacuum-tube logic elements, to transistors, to integrated circuits and to Cray's aborted use of gallium arsenide circuits, and emphasizes Cray's ground-breaking contributions to computer architecture involving parallel processing and the use of pipelined execution units.

Engineering Research Associates (ERA) was organized in 1946 to develop commercial applications for the U.S. Navy's digital code-breaking machines, special-purpose logic calculators involving more than 1,500 vacuum tubes. Soon after Cray joined the company in 1951, it was sold to Remington Rand, but its development of a business-oriented computer, the 1101, designed specifically for commercial applications continued. Shortly before, Remington Rand had bought the Eckert-Mauchly company (developer of the 1946 ground-breaking ENIAC vacuum-tube computer) and made it the Univac Division. Thus the purchase of ERA generated tensions for both groups. But these changes and tensions had little effect on Cray, who wanted nothing more than to work creatively and individually on ways to achieve greater computer power and speed. As the book shows, in spite of rapidly becoming the pre-eminent supercomputer designer of his time, Cray was unable to wholly isolate himself from the necessary corporate superstructure that supported his work.

Murray does a creditable job of describing the endless struggles computer designers faced, and still face, in achieving the small size needed to minimize signal transit time and in controlling the resultant heating. He properly emphasizes the wonderful achievements of M. Dean Roush, Cray's cooling engineer. But, in our opinion, he downgrades Jim Thornton, Cray's codesigner of the Control Data Corporation 6600 supercomputer, by calling him an "important cog." In fact, Thornton made some major contributions to the 6600, including devising a systematic attack on and solutions for the dependency problems associated with pipelining—now called scoreboarding. He and Roush, along with Cray, were the principal heros of supercomputer development.

In making his case for the far-reaching contributions of Seymour Cray, the author shows him primarily as a conservative who emphasized simplicity and was generally content to use proven technology. Thus, until his use of gallium arsenide integrated circuits in the ill-fated Cray IV, he was more a follower than a leader. Although Murray credits Cray with the invention of the magnetic switch, it was actually developed several years earlier by An Wang. Cray's genius was in computer architecture rather than in hardware.

Because this book covers an important part of the history of computer development and may become a reference in the area, it would be desirable to point out a few errors, misconceptions and misprints. For example, "low-voltage vacuum tubes" should be "high-voltage vacuum tubes," and "nuclear implosion" should be "nuclear explosion." Murray celebrates Cray's use, around the year 1990, of 78 layers in motherboard design and contrasts that with the use in supercomputers of only single-layer boards a decade earlier (p. 209). We know of no supercomputer with only single-layer motherboards in 1980, and as early as 1970 the ASC used 17 layers. There are many more examples we could cite if we had the space to do so.

The Supermen is an exciting tale, well told, of the Cray line of supercomputer development. It is reminiscent of Tracy Kidder's 1981 book, The Soul of a New Machine, which described in detail the development of the Data General MV8000 computer, but it fails to be as riveting as this work because it covers a much wider historical span and many different computer developments, and is necessarily more diffuse and less concentrated. Nevertheless, as participants ourselves in the development of the MIT Whirlwind large-scale vacuum-tube computer in the late 1940s (Macdonald) and the TI ASC (Cragon), we found this to be a valuable and interesting story of supercomputer development over the years. Murray succeeds in providing instructive insights into the character and style of the many engineers and managers who played important roles in the genesis and advancement of the series of Cray supercomputers that revolutionized the field.

Other excellent engineers around the world who worked on supercomputer development independently of Cray and his groups may find the title of this book a trifle exaggerated. Although the book is written at a popular, sometimes even superficial, level, its story of the genesis, evolution and flowering of supercomputing technology is a timely and instructive one. Read it and see what single-minded drive and genius (at least in the Edisonian sense of unremitting hard work and constant thinking about the problem) accomplished in a field that is becoming of greater and greater importance to our information-oriented society.—J. Ross Macdonald, Physics, University of North Carolina at Chapel Hill, and Harvey G. Cragon, Electrical and Computer Engineering, University of Texas, Austin