Communications of the ACM
Seymour Cray and the Development of Supercomputers
The name Seymour Cray is almost synonymous with supercomputing.
Credit: Cray Research, Inc.
Computers in the 1940s were vast, unreliable beasts built with vacuum tubes and mercury memory. Then came the 1950s, when transistors and magnetic memory allowed computers to become smaller, more reliable, and, importantly, faster. The quest for speed gave rise to the "supercomputer", computers right at the edge of the possible in processing speed. Almost synonymous with supercomputing is Seymour Cray. For at least two decades, starting at Control Data Corporation in 1964, then at Cray Research and other companies, Cray computers were the fastest general-purpose computers in the world. And they’re still what many people think of when they imagine a supercomputer.
Seymour Cray was born in Wisconsin in 1925, and was interested in science and engineering from childhood. He was drafted as a radio operator towards the end of World War II, went back to college after the war, then joined Engineering Research Associates (ERA) in 1951. They were best known for their code-breaking work, with a little involvement with digital computing, and Cray moved into this area. He was involved in designing the ERA 1103, the first scientific computer to see commercial success. ERA was eventually bought out by Remington Rand and folded into the UNIVAC team.
In the late 1950s, Cray followed a number of other former ERA employees to the newly-formed Control Data Corporation (CDC) where he continued designing computers. However, he wasn’t interested in CDC’s main business of producing low-end commercial computers. What he wanted was to build the largest computer in the world. He started working on the CDC 6600, which was to become the first really commercially successful supercomputer. (The UK Atlas, operational at a similar time, only had three installations, although Ferranti was certainly interested in sales.)
Cray’s vital realisation was that supercomputing – computing power – wasn’t purely a factor of processor speed. What was needed was to design a whole system that worked as fast as possible, which meant (among other things) designing for faster IO bandwidth. Otherwise your lovely ultrafast processor would spend its time idly waiting for more data to come down the pipeline. Cray has been quoted as saying, "Anyone can build a fast CPU. The trick is to build a fast system." He was also focussed on cooling systems (heat being one of the major problems when building any computer, even now), and on ensuring that signal arrivals were properly synchronised.
From LinuxVoice
View Full Article
No entries found