The Supermen

Murray, an engineering journalist, wrote a history of American supercomputing centred on Seymour Cray. It runs from Cray's work at Engineering Research Associates and Univac in the 1950s, through the founding of Control Data Corporation and the CDC 6600 and 7600, to Cray Research and the Cray-1 and Cray-2, and finally to Cray Computer Corporation and its failure with the gallium arsenide Cray-3.

The technical argument concerns design philosophy. Cray built machines by minimising the distance signals travelled, which drove the physical layout of his computers, including the cylindrical shape of the Cray-1 that shortened wire lengths. He used relatively simple processor designs at extremely high clock speeds rather than complex architectures, and he pushed cooling technology hard, eventually immersing the Cray-2 in fluorocarbon liquid. Murray also documents Cray's working method: small teams, isolation from corporate management, refusal to use the previous machine's design, and starting each project from a blank sheet.

The business argument is about the limits of that method. Cray Research prospered while single-processor speed determined supercomputer performance. When massively parallel machines built from commodity microprocessors became competitive in the late 1980s, the economics changed, and Cray's approach could not match them on cost. Cray Computer Corporation went bankrupt in 1995 and Cray died in a car accident the following year.

The book is a good technical and personal history but it is not analytical about the industry, and it ends before the cluster architectures that now dominate high-performance computing. Readers wanting the market history should supplement it.