Supercomputer designer Steve Chen has finally broken cover after some three years during which people had started wondering whether he was even still in business, and has been talking guardedly about the SS-1 machine that his Eau Claire, Wisconsin-based Supercomputer Systems Inc is developing with IBM Corp’s money. He promised that the SS-1 will be faster and more useful than anything available from Cray Research Inc, and that while it would be faster, speed is no longer a good measure of a supercomputer’s usefulness – he accused competitors of talking in MachoFLOPS. He stressed that what really mattered were the software tools that companies like his could come up with that inspire creativity in the engineers and scientists that used the machines – if you can solvethe problem you have today with your present computer, don’t bother to get a supercomputer, he said, suggesting that users wait till the next generation, which would drastically shorten the time to design things like cars and aircraft. He says that the SS-1 will combine elements of massively parallel and vector processing, and will be built out of clusters of processor and memory modules. He will start building production models of the SS-1 – for testing only – next year. A single-processor version is planned, but observers say a full SS-1 may cost as much as $70m. The machine is already two years behind schedule, in part because of several bad chips IBM is making the chips to Chen’s design. Despite no immediate prospect of any sales or revenue, the company employs 300. Chen broke cover at a Bloomington, Minnesota symposium sponsored by Research Consortium Inc of Eden Prairie and the Gartner Group.
Content from our partners
