Smaller and Faster: The Cray-2 and 3

Cray-2 CPU and cooling tower

The Cray-2's unusual cooling scheme immersed dense stacks of circuit boards in a special non-conductive liquid called Fluorinert™, which in turn was cooled in the tank at the right. This machine was used to simulate nuclear fusion.

Smaller and Faster: The Cray-2

For many supercomputers, swifter meant bigger. But the Cray-2 supercomputer, though smaller than the Cray-1, ran 12 times faster. It computed in one second what would have taken ENIAC several months.

The Cray-2 took nine years to develop. Released in 1985, each of the 27 made sold for between $12M and $17M.

Seymour Cray (left) and Cray Research CEO John Rollwagen with a Cray-2 cooling model

Rollwagen, an electrical engineer and businessman, became CEO in 1980. He led Cray Research, Inc. through its most successful period, but also through its decline.

View Artifact Detail
1 2 3 4 5 6 7 8
Cray-2 marketing brochure

Behind the CPU is the Cray-2 cooling tower which re-circulated Fluorinert cooling fluid throughout the computer.

View Artifact Detail
CFT77 Fortran manual

Fortran is the standard programming language for mathematics, science and engineering. This brochure describes the Cray version of Fortran which was the language of choice for supercomputer customers.

View Artifact Detail
Hitting the Wall: The Cray-3

The Cray-3 was to be 10 times faster than the Cray-2. But Seymour Cray, unwilling to join the trend toward using many slower processors, switched to exotic gallium arsenide chips, packing more than a thousand in each 4-cubic-inch module.

This risky path to speed backfired. Cray sold only one partially completed machine.

1 2 3
Cray-3 CPU section

Unlike Cray’s previous designs, the Cray-3 used unproven technology—gallium arsenide—and Cray had to invest in the company making them. The clock speed of 474 Mhz was impressive for the time, but the machine was difficult to manufacture.

View Artifact Detail
1 2
Designer Seymour Cray and the Cray-3

The only Cray-3 made was delivered to the National Center for Atmospheric Research in 1993, but was soon taken out of service because it was unreliable.

View Artifact Detail