New book on Cray, supercomputing era

Rohit Khare (khare@pest.w3.org)
Tue, 11 Feb 97 17:05:30 -0500


-> http://www.businessweek.com/1997/07/b351434.htm

"It is Cray's personality and approach--the wide-ranging mind that was not
above grappling with small, tactile problems--that make him such a compelling
figure. Moreover, writes Murray, ''the performance of his machines--coupled
with his well-known eccentricities--created an extraordinary mystique.'' (By
contrast, consider today's Internet entrepreneurs, with their ethereal
software products, and think how colorless their life stories may seem in the
future.)"

"While IBM's committees jammed 735 instructions into a high-performance
computer called STRETCH--which bombed spectacularly in the early 1960s--the
6600 sped by on only 64."

-------------------------------------------------------------

THE SUPERMEN
The Story of Seymour Cray and the Technical Wizards Behind the Supercomputer

By Charles J. Murray
Wiley 232pp $24.95

In 1963, BUSINESS WEEK helped touch off one of the splashiest legal battles
in American business history. A story in that year's Aug. 31 issue described
how a certain Seymour Cray, working with a few engineers in the woods of
Wisconsin, had created what was by far the world's fastest computer. Control
Data Corp.'s 6600, as it was called, cranked through an astonishing 3 million
instructions per second. The article raised the question: How could such a
machine ever be kept busy? Evidently, the news incensed IBM Chairman Thomas J.
Watson Jr.: In a now famous memo, he rubbed his lieutenants' noses in certain
of the story's details: ''I understand that in the laboratory developing this
[CDC] system there are only 34 people, including the janitor.... Contrasting
this modest effort with our own vast development activities, I fail to
understand why we have lost our industry leadership position....''

The giant was aroused. Within months, Big Blue was promising customers
souped-up models of its new System/360 computers, claiming they would
outperform CDC's machine. In fact, those high-end 360s never existed--and
never would--but the mere pledge was enough: For 18 months, CDC could not book
a single order for its very real and deliverable 6600. In response, Control
Data sued IBM in 1968 as being a monopoly selling ''paper tigers.'' Big Blue
eventually settled, but not before Watson's scathing memo was made public--or
before more than a dozen other companies, and even the U.S. government, filed
similar antitrust actions. IBM was kept busy in court until 1982.

As it turns out, the rather shy engineer who so angered Watson--and who is
the focus of Charles J. Murray's interesting new book, The Supermen--later
came to be recognized as one of digital computing's few heroic figures.
Seymour R. Cray was a reclusive and somewhat eccentric but brilliant engineer
whose name is now synonymous with supercomputers. Always pushing the outer
limits of computer performance, Cray worked with every major hardware
technology, from vacuum tubes in the 1950s to exotic gallium-arsenide
microchips in the 1990s. Through it all, he enjoyed the glory of great
technical and commercial success and the admiration of the entire computer
industry. Sadly, it all ended on a tragic note.

Among the things Murray describes best is how Cray stands out as one of the
few computer designers who had the individual genius and independence to
conceive new computers from scratch--not once, but again and again. Nearly all
other successful computer families, from IBM's System/360 to Intel Corp.'s
x86 line of microprocessors, have been parented and nurtured by committees
that struggled to exploit new technologies while maintaining compatibility
with past designs.

Cray, by contrast, always started with a clean sheet of paper, producing one
record-breaking machine after another. Cray's triple challenge was always
this: make digital circuits run faster, which means hotter, too; cram more of
those circuits closer together so they can communicate faster; then, extract
the heat from that dense, toaster-hot package so it won't burn up. To cool his
machines, Cray used everything from compressed Freon to automotive fuel
injectors spraying oil.

It is Cray's personality and approach--the wide-ranging mind that was not
above grappling with small, tactile problems--that make him such a compelling
figure. Moreover, writes Murray, ''the performance of his machines--coupled
with his well-known eccentricities--created an extraordinary mystique.'' (By
contrast, consider today's Internet entrepreneurs, with their ethereal
software products, and think how colorless their life stories may seem in the
future.) Cray even helped design the basic logic, or architecture, of his
machines. The 6600, for instance, was the first computer to employ an approach
that 20 years later was hailed as revolutionary--so-called reduced
instruction set computing, or RISC. While IBM's committees jammed 735
instructions into a high-performance computer called STRETCH--which bombed
spectacularly in the early 1960s--the 6600 sped by on only 64.

Murray does a fine job of tracing Cray's career through a succession of
companies: Engineering Research Associates, Remington Rand, Control Data, Cray
Research, and finally, Cray Computer. At each step along the way, he explains
the technical and commercial challenges that confronted and finally defeated
his subject, whose final creation, the Cray-3, never made it out the door.
Unfortunately, the author fails to probe much below the surface of either the
business of supercomputing or, more disappointingly, of Cray as a person. What
really motivated this man? Where did he get his inspiration? What did he
think about when not working, and so forth? Murray thanks Cray and his many
colleagues for allowing interviews. But nowhere, alas, do we hear Seymour
Cray's own voice.

Sad to say, we no longer have the chance. In 1996, just months after shutting
down his last company, Cray Computer Inc., which burned through $200 million
to produce only a single prototype, Seymour Cray died, at the age of 70. The
man who fathered some of the most advanced technology ever conceived died as a
result of a primitive mechanical catastrophe--a head-on automobile collision.

By JOHN W. VERITY