Phenquestions
Cray's Anatomy
Seymour Cray was an American engineer and supercomputer architect who spent most of his life designing supercomputers and has been credited as the man who created the supercomputing industry. Widely recognized as the “father of supercomputing” [3], he was first an employee before becoming a businessman. He was a graduate of Electrical Engineering at the University of Minnesota in 1949 and completed his Master's degree in Applied Mathematics in the same institution in 1951.
From ERA to CDC
In1950, while still finishing his master's degree, Cray joined Engineering Research Associate (ERA), a new local company in Saint Paul, Minnesota. His expertise in digital computer technology led him to his very first project, ERA 1103, widely known as UNIVAC 1103, which then became the first scientific computer.[4] When ERA was purchased by Remington Rand and was merged with its UNIVAC department, many of its founders left to form Control Data Corporation (CDC). In 1958, Cray left ERA and joined his colleagues at CDC.
While at CDC, Cray set-up a lab in his own home in Chippewa Falls, Wisconsin, where he designed what came to be the first supercomputer, CDC 6600. It was released in 1964 and dominated the market for five years selling 200 units at $9 million each.[5] In 1968, with countless technical innovations, Cray completed the design of CDC 7600. Significantly faster than CDC 6600, CDC 7600 can process data at 36.4 Mflops.[6] CDC once again dominated the supercomputer industry with the release of CDC 7600. The success of the first two series of CDC supercomputers encouraged Cray to work on its third series, CDC 8600. It was, however, stalled when CDC went into some financial difficulties and prioritized another supercomputer project, STAR 100. It wasn't a workable arrangement for Cray, and so he decided to leave CDC to put-up his own company.
Starting-up Boldly
In the same year that he left CDC, Cray founded his own company, Cray Research Inc (CRI). With some doubts and still unaware of his reputation, he approached Wall Street for seed capital. To his surprise, investors lined up to back him up, and he conveniently acquired the funds he needed to put-up the company that provided the world's fastest supercomputers for decades.
Two years after his departure from CDC, they released STAR-100, which was three times faster than CDC 7600 and one of the first machines to employ vector processing, where the registers and memories are arranged to speed up the processing of a single operation on a large set of data.[7] However, poor implementation of the concept led to its poor performance and, eventually, its failure. With his knowledge and expertise in electronics and digital computer technology, Cray used a different approach in vector processing and replaced transistors with integrated circuits. With this and other enhancements in the design, Cray overcame the limitations of his competitors, and in 1976, CRI released its first vector supercomputer, Cray-1. Delivering 80MHz processor speed, and with a speed of 160 Mflops, Cray-1 surpassed the speed of any other computer at the time. The first system was acquired by Los Alamos National Laboratory after winning the bid at $8.8 million. Selling over 80 systems in the following years, Cray-1 is one of the most successful supercomputers in history. The success eventually made Cray a celebrity.
Following the success of Cray-1, the team at Cray Research, headed by principal designer Steve Chen, developed Cray X-MP. It was the first supercomputer by Cray Research that used multiple processors. Cray X-MP has a processor speed of 105 MHz and a speed of 800 Mflops. It became the world's fastest supercomputer from 1983-1985.
Seymour Cray, in the meantime, started working on Cray-2 and with 244MHz processor speed and 1.9 gigaflops (Gflops) system performance, took over Cray X-MP's spot as the fastest supercomputer of CRI in 1985. It, however, fell short to Russia's M13, which runs at a speed of 2.4 Gflops and the first one to break the gigaflop barrier.[8]
In 1988, Cray Research unveiled Cray X-MP's successor, Cray Y-MP. Another multiprocessor machine and improvement of Cray X-MP, it can handle up to 8 processors with a maximum speed of 2.667 Gflops. In addition, it has a higher memory bandwidth than Cray X-MP.[9]
Three years later, Cray Research released started releasing the Cray C90 series, another multiprocessor supercomputer that has double the capacity and speed of Cray Y-MP.
The Spin-Off
While Cray Y-MP was being developed, Seymour Cray was simultaneously developing Cray-3. Aiming to achieve 12 times the speed of Cray 2, he explored using gallium arsenide as semiconductors for the new machine. With Cray Y-MP underway, and because Cray 2's sales were lower than Cray X-MP, the company decided to put its development on hold. Undaunted, Cray left CRI and formed another company, Cray Computer Corporation (CCC), in Colorado Springs, Colorado, in 1988 and continued to work on the Cray-3 project. Because it was more ambitious than Cray-2 and various experiments were necessary, it proved to be more expensive than any of its predecessors. With numerous supercomputers emerging in the market, Cray-3 had no launch customer when it was completed in 1993. Its first and only model was instead sent to the National Center for Atmospheric Research (NRAC) for demonstration. [10] With no other sales prospect for Cray-3, CCC filed for bankruptcy in 1995.
SRC Computers and the Death of Seymour Cray
Cray used what remained of CCC to put up SRC Computers in 1995. With his unwavering passion for supercomputing, Cray went on to work on Cray-4, but it was interrupted when he died from injuries suffered from a car accident in 1996. After his death, Cray-4 was never completed.
CRI went on to release its C90 computer series up until 1996 when it was acquired by Silicon Graphics, which merged with Tera Computer Company in 2000. In the same year, Tera renamed itself to Cray, Inc.
The Legacy of Cray Supercomputers
Supercomputers play an important role in the field of computational science, from weather forecasting, pharmaceuticals, and nuclear studies, to name a few. To cater to today's demand for faster data processing and to lead the race in supercomputing, manufacturers are in constant pursuit of innovations. For decades, Seymour Cray's brilliance in supercomputing produced a series of supercomputers that have become pillars of today's giants. His works may no longer be in use today, but he undoubtedly built a legacy in the world of supercomputing.
Sources:
- Yevgeniy Sverdlik, June 22, 2020, “The World's 10 Fastest Supercomputers - in Pictures” https://www.datacenterknowledge.com/supercomputers/world-s-10-fastest-supercomputers-pictures/gallery?slide=1 Accessed 05 Oct 2020
- Wikipedia. “CDC 6600”, N.d., https://en.wikipedia.org/wiki/CDC_6600 Accessed 05 Oct 2020
- Wikipedia. “Seymour Cray” N.d., https://en.wikipedia.org/wiki/Seymour_Cray Accessed 05 Oct 2020
- “Cray Supercomputer”, N.d., https://history-computer.com/ModernComputer/Electronic/Cray.html Accessed 05 Oct 2020
- Wikipedia. “History of Supercomputing”, N.d., https://en.wikipedia.org/wiki/History_of_supercomputing Accessed 05 Oct 2020
- Wikipedia. “CDC 7600” N.d., https://en.wikipedia.org/wiki/CDC_7600 Accessed 05 Oct 2020
- Wikipedia. “Cray 1” N.d., https://en.wikipedia.org/wiki/Cray-1 Accessed 05 Oct 2020
- Google Arts and Culture. “The Cray 2 Supercomputer” N.d., https://artsandculture.google.com/asset/the-cray-2-supercomputer-seymour-cray/NQE7aCDl2Zb0dA Accessed 05 October 2020
- “The Cray Y-MP”. 14 Nov 1995, http://www.netlib.org/benchmark/top500/reports/report94/Architec/node9.html Accessed 05 Oct 2020
- Wikipedia. “Cray 3” N.d., https://en.wikipedia.org/wiki/Cray-3 Accessed 05 Oct 2020
