It would have seemed appropriate for Kartsev to take it easy, or to at least take a short break after working so hard, but it just didn't happen that way. Back in 1966, Kartsev proposed a plan for a multi-computer complex consisting of machines that were especially designed to work together. Preliminary research showed that such a complex could achieve an operational speed of one billion operations per second. At the time, no machine in the world was capable of reaching that speed. That goal inspired Kartsev and his subordinates, and by 1967 they had completed a draft for the design of the M-9 Computing Complex. The Defense Ministry quickly approved the project.
The M-9 Complex consisted of a control processor and four types of computers: a functional-operator, a numerical computer, as associative computer, and a peripheral calculator. The M-9 was supposed to work not just with single numbers, but with groups of numbers that were the approximate representations of functions, or multi-dimensional vectors. In other words, Kartsev designed the M-9 to analyze more in-depth relationships between the data than the contemporary machines were able to do at that time.
The main distinction between this machine (Kartsev named it the functional operator) and the typical computer was in the structure of the arithmetic units' interface, which were timed by the sameclocked circuit. At the end of every operation - each computer performed its command during one or two clock cycles - and beginning of the next, the exchange of information between the output from an arithmetic unit and the input into a memory unit (writing down previous commands), and between input into an arithmetic unit and the output from a memory unit (reading the next set of data), occurred without a significant loss of time.
The numerical vector machine, which was part of the M-9 complex, carried out operations on partial functions and multidimensional vectors. The high performance associative machine carried out most of the routine work of sorting and organizing information arrays. The numerical computer worked with an independent program and also with the programs that were synchronized with other computers in the M-9 complex. It coordinated the work of multiple computers and allowed the complex to maintain high productivity while processing heterogeneous information and creating a universal digital means for solving a wide class of problems that demanded very high performance computers.
Unfortunately, the M-9 complex was not mass-manufactured, although its design and the successful demonstration of the prototype were important achievements for Kartsev's team. 1967 was an excellent year for the M-9's designers because NIIVK was founded. Kartsev was appointed as its director and his department became the backbone; it was an official endorsement of Kartsev's scientific school.
In 1969, the Soviet government ordered the construction of the M-10 electronic computer, which was to be based on the already proven M-9 vector computer. Doctor of Technical Sciences Leonid Vasilievich Ivanov recalled, "The event was preceded by a compelling meeting to consider the future of two projects already in the making: the Elbrus, managed by Lebedev, and the M-10, managed by Kartsev. Lebedev vigorously argued against a multi-processor version of the Elbrus, insisting on a single processor version for maximum productivity. Academician Glushkov supported both directions, and they were approved."[3]
Early in 1970, the production factory began setting up to assemble its first M-10 model. Later that same year the construction blueprints were finalized and by August 1971 the prototype of M-10 had been manufactured and ready for adjustment. At the same time, the construction plans for the industrial model of the computer were undergoing revision in preparation for their mass-manufacture. 1971 turned out to be a very hard year for Kartsev and the strenuous work took its toll: a heart attack left him bedridden for several months. Fortunately, he recovered.
By June 1973, every system component of first M-10 model was assembled and tested according to the technical specifications, and the machine finally came together as a complete unit. In September, the first industrial model of the M-10 successfully passed all technical tests and was placed into trial operation for additional software debugging.
By December, the factory had finished testing the second model and begun its mass-manufacture; the M-10 was produced for more than 15 years. Dozens of these computers were made and many of them are still in service today. Several powerful computer complexes had been built using the M-10 computer as the base. In 1976, an M-10 model from one such computer complex, and its software, successfully passed the rigorous state testing.
A group of NIIVK specialists and the factory were awarded the Soviet Union's State Prize for the M-10 in 1977. Among those from the NIIVK receiving the prize were Deputy Senior constructor Leonid Ivanov, Alexander Alexandrovich Krupsky, Leonid Yakovlevich Miller, Yuri Rogachev, Rene Shidlovsky, and software designer Alexander Karasik, along with Senior Engineer Anatoly Shishilov and Deputy Senior Constructor Valeri Alexandrovich Mushnikov from the factory. As chief of the project, Kartsev was awarded the Order of Lenin. Over a hundred of NIIVK specialists and factory workers were awarded other special Soviet orders and medals.
The M-10 computer was actually a synchronized multi-processor system and was part of the third generation of Soviet computers: its basic logic elements consisted of the 217-series Posol microprocessors. The computer was intended to support complex automated control systems in real time and to solve a variety of scientific problems. Having inferior microelectronics for its technological construction base, it did not perform as fast as the CRAY-1, which appeared at about the same time. Yet the M-10 possessed some architectural potential advantages in terms of the average number of processor cycles per single executed operation. The lower the number, the better the computer's architecture: the M-10's average number of processor cycles per single operation ranged from 0.9 to 5.3. The Cray-1's ranged from 0.7 to 27.6. The minimum values for both computers were close, but the M-10's maximum value was significantly less than that of the Cray-1.[4]
The M-10's value becomes even more apparent when considering why it was originally built: Kartsev and his colleagues designed the M-10 in absolute secrecy for the Soviet Missile Attack Warning System [in Russian, the Sistema preduprezhdeniya o raketnom napadenii, or SPRN] and for general outer space surveillance.[5] The system provided the Soviet Union's military leaders with comprehensive information regarding a possible threat of missile attacks and continuous observation of the cosmos. In space, via satellite, the SPRN detected missiles launches. On the ground, the system was composed of nine powerful radar-tracking stations located along the Soviet Union's borders near Riga, Murmansk, Pechera, Irkutsk, Balkhash, Mingechaur, Sevastopol, and Mukachevo, which were supported by a network of M-10–based computing complexes.
Up until the early 1980s, the M-10 reigned as the highest-performing computer in the Soviet Union in speed (it ran at about 20-30 millions operations per second), internal memory capacity, and data transmission in a multiplex system. For the first time in the world, its design allowed for seven computers to be connected and have a direct information interchange (without multiplex channels) between individual computer programs. Also, the system featured automatic reconfiguration of a field of processors, a second level internal 4 megabyte random access memory, and external access to both levels of internal memory.
These innovative technical features received eighteen invention certificates and five industrial model certificates. Beginning in 1980, Kartsev and his team gave the system new storage devices and renamed it the M-10M. The M-10 and M-10M computers had fully compatible software and hardware. In his presentation at the NIIVK's 15th anniversary, Kartsev discussed these memorable years:
In 1967, we made an audacious proposal – to build the M-9 computing complex. Because it was the 50th anniversary of the October Revolution, I nicknamed the complex "The October." The Ministry of Devices building, where we were set up, was too small for us, but the officials told us: 'Since you already work for Kalmykov, go see him.'[6]
The M-9 project was never realized. But in 1969, we started the M-10 project, and it was up and running by 1973. For many years, it was the most powerful computer in the Soviet Union. This complex is also responsible for some unique scientific discoveries, particularly in the field of physics. Yet, the project was not greeted with open arms, and frankly speaking, the authorities told us that we were crazy, that we could never make a computer out of a heap of metal, and that the whole thing would never work. Only now we've got them convinced, and subconsciously they understand that a big computer needs a huge amount of equipment. But back then, nobody could picture it. The work was extremely difficult; our team had to work at a number of sites around Moscow: at the Sokol-1 enterprise and on Great Pochtovaya Street, plus in a number of sub-basements: on Great Vasilevsky Lane, on Burdenko Street, Plyushchikha, and on Shchukina Street.
After the establishment of the Electronic Control Computer Institute, the team acquired 590 square-meter premises of a former cabinet-maker's shop at Sokol square. In order to accommodate the whole team, we still had to lease other premises – mainly sub-basements – all around Moscow. The Institute constructed a separate building for us in 1975, and then added the laboratory wing as a special project, but not until 1985 through 1986. Nevertheless, the Ministry's managers were always friendly and supportive, as were our customers. They helped us get down to business and we came of age.
It was not difficult to understand the skeptics' position, especially considering some numbers: the BESM-6 computer operated with 60 million transistors, 180 thousand semiconductor diodes and 12 million ferrite rings. The M-10 computer used 2 million microchips, 1.2 million transistors and 120 million ferrite rings. It was not a "heap of metal" as Kartsev called it, but an unimaginable number of electronic elements that was supposed to be seamlessly integrated with complex circuitry. When all of the bugs were finally worked out and the machine became operational, its total annual loss of productivity amounted to only 10 minutes!
[3] Ivanov's article appeared in the Russian-language journal Questions of Radioelectronics [Voprosii Radioelektronniki], Vol. 2, 1993.
[4] Author's note: For more on this, see B.A. Golovkin, "The Evolution of Parallel Architectures and the M Series Computers," in Questions of Radioelctronics, No. 2, 1993.
[5] Editor's note: The first public information about this was disclosed only on April 1, 1990 in Pravda, with the publication of A. Gorokhova's "Stoyanie pri Pestryalove," or "The Problem at Pestryalov." Pestraylov was a top secret Soviet military site.
[6] Editor's note: Valeri Dmitrievich Kalmykov was Minister of Minradioprom, the Soviet All-Union Ministry of the Radio Industry, or Obschesoyuznoe Ministerstvo Radio Promishlennosti.