CDC 6600

In search of a more relaxed environment to work on the CDC 6600, Cray, with his team of 30 engineers, moved to his childhood city, Chippeua Falls, in a separate laboratory in July 1962, away from the CDC headquarters in Minneapolis . In the CDC 6600, Kray used Fairchild Semiconductor planar silicon transistors instead of germanium. Thanks to the higher switching speed of the logic gates built on these transistors, it was possible to significantly increase the speed of the computer and greatly simplify its circuit. Despite the fact that silicon transistors withstand much higher operating temperatures than germanium, due to the high packing density of 400,000 logical elements of the computer, in order to reduce the length of the electrical circuits in the CDC 6600, Cray had to think about heat dissipation - air ventilation was not enough. Engineer Dean Roush developed a cooling system that removed heat from printed circuit boards with transistors to the tubes that circulated freon through aluminum radiators, like in a real refrigerator . The work on the CDC 6600 Cray was largely assisted by engineers Jim Thornton (CPU) and Dean Roush (cooling system).

By mid-1963, the CDC 6600 was ready, and early tests showed that it ran 50 times faster than the CDC 1604 . High speed was achieved thanks to several innovative solutions: in contrast to the then widespread scheme, in the CDC 6600 the main processor of the computer performed only logical and arithmetic operations. Work with peripheral devices was assigned to 10 “peripheral processors”, the main purpose of which was to “feed” data from input devices to the central processor and collect the results for sending to output devices . This made it possible to unload the central processor, reduce the set of its machine instructions to a minimum and make their execution very fast, that is, practically realize the idea that was later implemented in RISC processors later in the 1970s. In addition, the central processor had a command pipeline - a novelty in the computer industry, which before that had already been implemented in some computers of competing companies ( ILLIAC II , IBM 7030 Stretch ).

On August 22, 1963, a computer was introduced to a select group of journalists at Chippewa Labs. The presentation demonstrated the operational and technical capabilities of the computer, in particular, it was stated that the machine is capable of performing 3 million operations per second and at the same time provide 11 programs. The first model was intended for installation in the National Radiation Laboratory. Lawrence of the United States Atomic Energy Commission in the spring of 1964 ^[1] It was with reference to the CDC 6600 that the word " supercomputer " was first used. The CDC 6600 was 3 times faster than its main competitor, the IBM Stretch , at a much lower price — a little over $ 7 million ^[2] versus 13 million for the IBM Stretch . This made such a deep impression on IBM CEO Thomas Watson Jr. that the famous text of his internal memorandum of August 28, 1963 went down in history:

Last week, Control Data held a press conference at which it officially announced the release of its 6600 system. As far as I know, in the laboratory where the system was developed, only 34 people work, “including the cleaner”. Of these, 14 are engineers, 4 are programmers, and only one person has a Doctor of Science degree - a rather young programmer. The laboratory impressed the visitor with its efficiency, attitude to work and high motivation.

Comparing this very modest team with our own vast development team, I can’t understand why we lost our leadership, giving someone else the opportunity to offer the world's fastest computer in the market. At a meeting in Jenny Lake, I think, first of all, it will be necessary to discuss what we are doing wrong and what needs to be changed immediately ^[3] .

Original text

Last week Control Data had a press conference during which they officially announced their 6600 system. I understand that in the laboratory developing this system there are only 34 people, "including the janitor". Of these, 14 are engineers and 4 are programmers, and only one person has a Ph.D., a relatively junior programmer. To the outsider, the laboratory appeared to cost conscious, hard working and highly motivated.
Contrasting this modest effort with our own vast development activities, I fail to understand why we have lost our industry leadership position by letting someone else offer the world's most powerful computer. At Jenny Lake, I think top priority should be given to a discussion as to what we are doing wrong and how we should go about changing it immediately.

Having entered into a polemic with Watson in absentia, Cray replied: “It seems that Mr. Watson himself answered his own question” ^[4] , hinting that there are too many people working at IBM ^[5] .

The first serial CDC 6600s were unstable: despite all the efforts made to cool down, after 8-9 hours of operation, the computer overheated to 170 degrees and stopped working. Correction of deficiencies led to delays in the supply of computers to customers, which in turn was used by competitors and mainly by IBM , which decided to win the market of scientific supercomputers from CDC.

The assembly of the CDC 6600 computers was entrusted to the CDC factory in a suburb of St. Paul - Arden Hills . Cray insisted that the first five machines be assembled by his lab workers, and the Arden Hills factory workers were present to build and learn from. Cars with serial numbers 6 and above were already assembled at the Arden Hills factory ^[6] .

On the basis of the CDC 6600, a whole line of less expensive, but much slower-working computers under the general name "6000 series" were subsequently produced: CDC 6400 (April 1966), CDC 6500 (October 1967), CDC 6700 (October 1969) ^[7] . The CDC 6800 computer was supposed to be a continuation of the line, but Seymour Cray refused to simply improve the old model and created a new computer based on it - CDC 7600 , which became the next stage in the history of the development of supercomputers.

In 1968, the Soviet Union showed an informal interest in importing a CDC 6600 computer for the Institute of High Energy Physics in Protvino to analyze these experiments on the U-70 accelerator, which at that time was the most powerful in the world. As a response courtesy, the USSR was ready to share the results of the analysis with US scientists. The US government did not agree to this deal, fearing that the computer might be secretly used by the USSR to develop nuclear weapons ^[8] . As a result, British cars were purchased by , which in their characteristics were close to the CDC 6600.

In the same year, serial production of the BESM-6 computer was launched in the USSR, the productivity of which was estimated at 800 thousand operations per second using a mixture of Gibson-III commands, 8% higher than that of the CDC 6200 with a productivity of 740 thousand operations per second ^[9] . BESM-6 used 240 thousand semiconductor devices (60 thousand transistors and 180 thousand diodes) ^[10] .

In the spring of 1969, CDC applied to the US Government for an export license to supply the CDC 6400 to Yerevan to process the 6 GeV experimental data of the ARUS electronic synchrotron . After much hesitation, the license was refused in the summer of 1970 ^[11] .

In 1972, the CDC 6200 was installed at the Joint Institute for Nuclear Research in Dubna under the supervision of the Export Control Coordination Committee . In 1974, it was developed to the CDC 6400 model, and the next year to the multiprocessor CDC 6500 ^[12] .

• In 1966, the USA denied CDC the export license of the CDC 6600 computer for the French Atomic Agency to prevent France from its atomic program ^[13] .

•  

  Control console and modular racks in the background

•  

  Large control console and modular racks

•  

  Ferro magnetic memory module

•  

  Management console exhibited in Paris at Le Defense

• Thornton, J. Considerations in Computer Design - Leading up to the Control Data 6600 . - 1963. - 28 p.
• Thornton, J. Design of a Computer - The Control Data 6600 . - Glenview, IL: Scott, Foresman and Co., 1970. - 181 p. - ISBN 0673059537 .
• Grishman, Ralph. Assembly Language Programming for the Control Data 6000 Series and the Cyber ​​70 Series . - New York, NY: Algorithmics Press, 1974.- 240 p.
• David E. Lundstrom. A Few Good Men from Univac . - Mit Press, 1987 .-- 300 p. - ISBN 0262121204 .

• CONTROL DATA 6400/6500/6600 COMPUTER SYSTEMS Reference Manual
• Photos of CDC products, including the CDC 6600, on the Charles Babbage Institute website