Supercomputers: A Way To Screen Out The Hype?
Never has the market for high-end computers been more confusing for the scientists and engineers who use them. A plethora of supercomputer hopefuls is flooding the market with unusual and incompatible designs--each of them claiming to offer superior performance. With bewildered buyers sitting on their hands, the competitiveness of the U.S. supercomputer industry is threatened by stagnation.
To inject some rationality and get things moving again, computer scientist David J. Kuck is lobbying federal agencies to create a "national performance center" that would cut through the hype and help customers evaluate the real performance trade-offs of competing designs. Kuck, who until recently headed the Center for Supercomputing Research & Development at the University of Illinois, figures the proposed center would cost less than $2 million a year to operate. Individual agencies say they're interested--but only if they run the center. Kuck insists that won't fly, because the organization he envisions will require wide collaboration to spur progress.