Seymour Cray (1925-1996) was an American engineer and supercomputer architect. He designed a series of computers that were among the most powerful in the world for many decades. Cray is sometimes known as the father in supercomputing.
Seymour Roger Cray, the son of Seymour Cray and Lillian Cray, was born in Chippewa Falls in 1925. Seymour, the elder, was a civil engineer who encouraged his son’s interest in engineering and science. Seymour, a young man often, was able to construct a device from Erector Set parts. This converted punched paper tapes into Morse codes signals. Cray was given the basement of his family home as a laboratory. Cray was a keen student of electrical engineering and preferred to spend as much time as possible in his laboratory.
After graduating from Chippewa Falls High School, Seymour enlisted in the US Army in Europe in 1943 as an infantry communications platoon radio operator. He then moved to the Pacific military theater in the Philippine Islands where he worked on cracking Japanese codes. He returned to the United States in 1945 and enrolled at the University of Minnesota. In 1949, he received a Bachelor’s degree in Electrical Engineering. A Masters in Applied Mathematics followed in 1951.
Cray remained at the university and joined Engineering Research Associates (ERA), a local company in Saint Paul, Minnesota. ERA built specialized cryptographic equipment to support the US Navy. Cray worked with a variety of computer technologies while at ERA. He used everything from magnetic amplifiers and vacuum tubes to transistors. He quickly became an expert in digital computer technology after his design work on UNIVAC 1103, the first commercially viable scientific computer. (See the photo below).
The console of ERA 1103 (UNIVAC 1103) includes the Input/Output section.
ERA 1103 was the first commercial computer to use RAM-random access memory. It used 36 Williams tubes with a capacity of 1024 bits each. This gave it a total memory of 1024 words (36 bytes each). The magnetic drum memory contained 16384 words, magnetic tape storage, and four Raytheon/Potte units. The instruction set included 41 logical and arithmetic operations. Standard input and output equipment included a paper reader, typewriter, and paper punch. Optional equipment could include a line printer or oscilloscope display unit.
Cray quit ERA in 1958 to join the Control Data Corporation (CDC) which had just been established by his colleagues at ERA. Cray had already completed the design for the CDC 1604, a low-cost, improved version of ERA 1103, and it had impressive performance considering its price. Cray then designed the CDC 6600 commercial supercomputer. It outperformed all other computers at the time by a large margin. Cray put a lot of effort into the design of this computer to make it run as fast and efficient as possible, even though the hardware was not the most advanced. Cray, unlike most high-end projects at the time, realized that performance was more than just processor speed. I/O bandwidth also had to be maximized to not “starve” the processor of data.
The CDC 6600 (see lower photo) was a large-scale solid-state, general-purpose computing system. It was a distributed architecture with a central scientific processor and ten fast peripheral machines. This machine was a reduced instruction set (RISC), long before the term was even invented. The CDC 6600 was announced in 1964. It was very expensive (the base model cost $6891300). It was also the first computer to be developed in the Chippewa Falls laboratory in Cray.
The input to the computer was via punch cards or digital magnet tape. Output was two line printers. A card punch, a photographic planter, and standard magnetic tape. Operator input was done via a keyboard. Interactive display consoles allowed users to see graphical results while data was being processed. The computer contained 65000 60-bit words and had a large disk storage device. Six high-speed drums were used as storage intermediate between the central core storage unit and magnetic tapes. It supported FORTRAN 66 and a program library. On the CDC 6000 series of computers at ETH Zurich, Pascal, a still-popular programming language was created.
Cray was able to work on 6600 computers after he had been successful. Cray decided to leave CDC in 1972 and start his own business. Cray established Cray Research with the help of the CDC and a group of former CDC employees. He used the same Chippewa Falls lab and set up a new production facility.
Cray Research was initially unable to finance a supercomputer because the CDC, which is now a large organization, couldn’t support more than one. Cray was surprised to discover that Wall Street was well-respected when he approached Cray to seek seed capital. Cray was able to get all the funding he needed from the financial world.
The 80 MHz Cray-1 was first announced in 1975. The first box machine was delivered in 1976. Although it had a theoretical performance of 160 MIPS (80MHz x 2 instructions per cycle), floating-point performance was generally around 136 MFLOPS. Surprisingly enough, there was so much excitement that a bidding war broke out between Lawrence Livermore National Laboratory (now Los Alamos National Laboratory) for the first system. The latter won and received serial number 1, in 1976, for a six-month trial. Cray-1 with serial number 3 was given to the National Center for Atmospheric Research, paying US$8.9million ($7.9m plus $1 million for disks) in 1977.