Mr. Cray Goes To Washington, Hat In HandRussell Mitchell
Suddenly, the reclusive Seymour Cray is talkative. No wonder. His company is almost broke. And he wants Uncle Sam to lend him a hand.
At 67, Cray is supercomputing's revered patriarch. His inventions made Control Data Corp., Cray Research Inc.--indeed, the U.S.--leaders in blazingly fast computers, key to building nuclear bombs, assessing global climate change, and crash-testing cars in cyberspace. But his current company, Cray Computer Corp. in Colorado Springs, Colo., says it has only enough cash--just $27.6 million--to last until August. Cray can't count on selling any multimillion-dollar Cray 3 computers before then, so he's desperate for new financing. So, hat in hand, he's heading for Washington in search of money--or, at least, a firm order or two.
That will be a first for Cray, who has never asked for direct government aid in three decades of building supers. Too much red tape, he always said. But Cray is a pragmatist. "Government money is better than no money at all," he says now.
Even with the new Administration's technophilia, it's unclear how Cray will be received. He has yet to show a working model of his Cray 3, which is five years behind schedule and has soaked up $300 million. Now, Cray says, a first public demonstration is planned within a month. The one Cray 3 order that was made, by Lawrence Livermore National Laboratory, was canceled more than a year ago.
VECTOR POWER. One of the Cray 3's big problems is that it's a "vector" supercomputer, based on a design Seymour developed in the 1970s at his previous company, Cray Research. Many analysts view the vector market as mature, soon to be surpassed by massively parallel processor (MPP) supercomputers. These machines spread large problems over dozens of cheap microprocessors and their makers claim they can often outperform vector machines for less money.
Cray's response: Contrary to popular belief, vector machines are far from reaching their speed limits, and MPP machines have yet to prove themselves at the speeds the Cray 3 will deliver. "Software problems have kept them from being competitive," he says. Cray argues that his company's recent decision to bet on MPP is simply "diluting" its support for future vector-processor machines.
With Steve S. Chen, this country's only other high-end vector designer, out of business since closing his Supercomputer Systems in February, Cray says his machines are the best bet for pushing vector technology to the limit. "I'm going to take the core business away from my old company," he says. Cray Research says it remains fully committed to the vector market.
The Cray 3's technology is impressive. It uses 16 ultrafast processors made from gallium arsenide chips, not the usual silicon. Each measures just 4 inches by 4 inches by 1 inch, and is assembled by robots specially designed by Cray's engineering team. The entire computer fits in three-foot-high octagonal box.
STREET PERFORMANCE. But all that will be for naught if Cray Computer doesn't survive. In addition to visiting Washington, Seymour Cray has gone to Wall Street to make his sales pitch and seek financing that would do minimal damage to Cray Computer stockholders, the majority of whom have taken a fierce beating on their original investments. Since 1991, the company's stock has fallen from the high teens to 3 3/4.
For most of his adult life, Seymour Cray has known precisely what he wanted--a chance to build the world's fastest computers. Now, for the first time, he may need a benefactor to pursue his dream.