Supercomputer

In 1964, the CDC 6600 arrived at Lawrence Livermore National Laboratory, three times faster than any computer on Earth. Designed by Seymour Cray at Control Data Corporation, it could execute three million floating-point operations per second—and the term 'supercomputer' was coined to describe it. This machine was not merely an incremental improvement; it represented a new category of computing, one where raw performance enabled previously impossible calculations in nuclear physics, weather prediction, and cryptography.

The adjacent possible for supercomputing emerged from a specific confluence: the hydrogen bomb and the transistor. Thermonuclear weapon design required simulating physical processes too complex for hand calculation and too dangerous for physical experiment. The Atomic Energy Commission laboratories—Los Alamos, Livermore, Sandia—had an insatiable appetite for computing power and virtually unlimited budgets to acquire it. Meanwhile, the transistor computer had matured enough by 1960 to provide the raw components for machines far faster than anything built with vacuum tubes.

Why did Seymour Cray, working in a small laboratory near Chippewa Falls, Wisconsin, build the world's fastest computer when IBM had vastly greater resources? The answer lies in architectural philosophy and focus. IBM spread its effort across commercial, scientific, and military markets; Cray concentrated exclusively on raw speed for scientific computing. In 1957, a group of engineers left Sperry Corporation to form Control Data Corporation in Minneapolis. Cray joined them in 1958 and by 1960 had decided to design the fastest computer possible.

Cray's key innovations were architectural. The CDC 6600 introduced peripheral processing: instead of a single CPU handling all tasks, ten smaller 12-bit processors managed input/output operations while the main processor focused purely on computation. This division of labor—later called offloading—freed the central processor from slow I/O operations. The machine used Freon refrigerant cooling to manage heat from its 400,000 transistors, running at a clock speed of 100 nanoseconds—among the fastest of its era. The console used a cathode-ray tube display, a commercial first.

The first CDC 6600s were delivered to Livermore and Los Alamos in 1965, but the machine quickly found customers beyond government laboratories. CERN acquired one for particle physics calculations. Universities ordered them for scientific research. About 100 CDC 6600s were sold at approximately $7 million each—proving that a market existed for extreme computing performance.

The CDC 6600's reign as fastest computer lasted until 1969, when Cray's own CDC 7600 replaced it. This established a pattern: supercomputing would be defined by continuous architectural innovation in pursuit of speed. Cray left CDC in 1972 to found Cray Research, where he would design the iconic Cray-1 (1976) and Cray-2 (1985), becoming known as 'the father of supercomputing.' His legacy was not just the machines he built but the understanding that some problems—weather modeling, nuclear simulation, protein folding—require computational power beyond ordinary computers, justifying a separate category of machine designed for speed above all other considerations.

By 2025, the fastest supercomputers exceed an exaflop (10^18 operations per second), roughly three hundred million times faster than the CDC 6600. But the lineage is direct: machines designed for computational problems too large for any ordinary computer, built by engineers willing to pursue every architectural advantage.