Document Type: 
Document Number (FOIA) /ESDN (CREST): 
Release Decision: 
Original Classification: 
Document Page Count: 
Document Creation Date: 
December 22, 2016
Document Release Date: 
November 29, 2010
Sequence Number: 
Case Number: 
Publication Date: 
March 1, 1985
Content Type: 
PDF icon CIA-RDP86R00995R000501130001-8.pdf1.66 MB
Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 . ~~~ Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Directorate of ~ SeCI'et Intelligence Soviet Computer Technology: Little Prospect for Catching Up (c) Secret SW 85-10038 March 1985 copy 3 5 5 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 ....~...6~..~~ Soviet Computer Technology: Little Prospect for Catching Up This paper was written b Office of Scientific and eapons esearc . Comments and queries are welcome and may be addressed to the Chief, Information Technologies Branch, OSWR Secret SW 85-10038 March i 985 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Secret Soviet Computer Technology: Little Prospect for Catching Up Key Judgments The United States leads the USSR in all fields of general purpose digital /rljormation available computer technology. This lead ranges from at least three years for as 4f /November /984 internal memory devices to more than 10 years in high-performance was used in this report. magnetic storage systems. In general, the outlook for the remainder of the 1980s will be for the US lead to increase slightly, although, for some high- priority applications, the Soviets may be able to reduce or design around a particular technology gap The Soviets have made progress both in computer technology and in computer production techniques; however, their progress has been over- whelmed by the rapid advances made in the West and Japan. The Soviets' status in seven important areas of computer technology is summarized in figure 1. These estimates are based upon the first delivery dates of functionally equivalent US and Soviet civilian computer products. If we were able to include computer production volume and quality in our measure of technology, then the United States would be at least several more years ahead We believe there are many reasons why the Soviets trail the United States in computer technology: ? The Soviets' centrally planned economy does not permit adequate flexibility to respond to design or manufacturing changes frequently encountered in computer production; this situation has often resulted in a shortage of critical components-especially for new products. ? The extraordinary compartmentalization of information in the USSR- especially on technologies with potential military applications; compart- mentalization not only restricts the flow of information, but also results in much duplication of work because of a lack of knowledge about other activities. ? The Soviet preoccupation with meeting production quotas, frequently at the expense of component and system quality control. ? The lack of adequate incentives for Soviet managers to take the risks associated with innovations or new technology. ? Poor coordination between separate design institutes and production facilities, sometimes resulting in products that have to be redesigned to fit a factory's production capabilities. iii Secret SW 85-/0038 March / 985 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Figure 1 Computer Technology: United States Versus USSR Minicomputers Mainframes I ligh-pcr~ixnwncc peripherals U 1 2 3 4 5 6 7 8 9 10 Il 12 13 14 IS Approximate length of US lead in years ? The Soviets' lag in computer-aided design and computer-aided manufac- turing techniques caused by a belated development start and also, ironically, by the Soviets' lag in computer technology. ? Concerns by Soviet officials that a computer is a powerful tool that could be used for antirevolutionary activity and that a proliferation of comput- ers might reduce the tight control of information in the USSR; these concerns tend to restrict access to and firsthand knowledge about computers as well as their applications. ? Provincial disputes within and between ministerial and institutional organizations. Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Secret ? Very poor customer support-including inadequate user feedback, poor installation support, and delayed maintenance-that frequently results in reduced efficiency and productivity for computers in use. Similar reasons also account for the Soviet lag in microelectronics technology as well as instrumentation and test equipment; these technology lags in components and basic electronic tools that are essential for modern computers contribute directly to the Soviet lag in computer technology. In our view, the entrenched Soviet bureaucracy would probably find it difficult to take the necessary steps in the foreseeable future to correct many of these well-recognized problems The Soviet lag in computer technology and production is resulting in a lag in both civilian and military computer applications. We believe that the Soviets have sufficient numbers of computers for high-priority, low-volume military and civilian projects. It is the remaining user community, including Eastern Bloc allies, who will experience shortages and delays in obtaining their desired computer systems. We expect the shortage of Soviet automation equipment to hinder seriously the modernization of their industrial base and also the growth of their economy The Soviets apparently lag the United States also in the application of computers in their fielded military systems. Historically, there has been a tendency in the USSR to avoid the complex multimission military systems-for which computers are an essential subsystem-that are frequently preferred in the United States. The generally conservative Soviet weapon design philosophy has probably not taxed Soviet computer capabilities in the past. However, this may be changing. We believe that the Soviets will be forced to incorporate more-advanced technology into their weapon systems in order to stay competitive with Western military development The Soviets' most significant hardware deficiencies are in supercomputers and high-performance magnetic disk technology. We do not expect the Soviets to have a supercomputer until 1985 at the earliest, whereas the first US commercial supercomputer was delivered in 1976. In magnetic disk systems, the Soviets are about a decade behind the United States. Lags in these critical areas will constrain Soviet computer system performance for applications requiring high-speed capabilities, such as ballistic missile defense, and applications requiring high input/output data rates, such as large real-time command, control, and communications systems. In the Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Secret NOFOKN software arena, the number of experienced Soviet programmers who are also cleared for classified projects may still be insufficient to complete all priority projects on time. We expect the Soviets' future progress in computer technology to be heavily dependent upon their advances in microelectronics and in second- ary storage technology, and upon their continuing activity in legally and il- legally acquiring Western and Japanese hardware and software. Judging from past performance and current technology assessments, we expect the Soviets to fall further behind the United States throughout the 1980s. If the Soviets obtained turnkey production facilities or detailed production know-how from the West or Japan-as they have done in the past-they would be able to narrow, at least temporarily, a specific technology gap. Also, if they made a major technological breakthrough in areas where they appear to be investing heavily, such as in optical computing or optical storage-and chances are about even that they will-the Soviets could overcome some of their computer deficiencies, for applications such as ballistic missile defense or real-time reconnaissance 25X1 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Contents Page Key Judgments iii Introduction 1 Microprocessors 1 Internal Memory Technology 4 Semiconductor Memories 4 Magnetic Cores 5 Plated-Wire Memories 6 Minicomputers 7 Nairi-4 Minicomputer 8 PS-2000 Array Processor 9 General Purpose Mainframes 10 Large Scientific Computers 13 Magnetic Tapes 21 Magnetic Bubbles 21 Technological and Military Implications 21 Tables 1. Soviet Microprocessors 2 2. Characteristics of Soviet SM-I Minicomputers 8 3. Technical Specifications for Soviet Ryad-2 Mainframe Computers 15 4. "Standard" Soviet Elbrus-1 Configurations 18 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 1. Computer Technology: United States Versus USSR iv 2. US Intel-8080A Microprocessor and Soviet K580 3 3. - Soviet Copies of US Microprocessors, 1971-83 4 - __ -- -- 4. Semiconductor Memory Technology: United States Versus USSR, 1971-87 5 5. Magnetic Core Memory Technology: United States Versus USSR, 1955-85 6 6. Soviet Copies of US Minicomputers 9 7. Soviet SM-1420/SM-5 Minicomputer 10 8. Soviet Nairi-41 Minicomputer 10 9. Soviet PS-2000 Array Processor 11 10. Timetable: IBM and Soviet Ryad Mainframes, 1964-84 12 1 I . Soviet ES-1060 Twin Computer Complex 13 12. Mainframe Performance: United States Versus USSR 14 13. IBM System Software in Use in CEMA Countries 17 14. Soviet Elbrus-1 Computer Complex, Circa 1980 18 15. Magnetic Disk Technology: United States Versus USSR, 1965-88 20 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 secret Soviet Computer Technology: Little Prosp Catching U This report is an assessment of the state of the art of Soviet general purpose digital electronic computer technology. It is primarily a condensed version of a more detailed Technical Intelligence Report on the same subject.' The assessment is based on information about Soviet civilian computer systems; however, we believe that this information is indicative of Soviet capabilities in military general purpose computer technology A basic microprocessor typically consists of the inter- connection of an arithmetic and logic unit (ALU), a register set (very fast storage), a control unit, and interrupts. A microcomputer consists of a micro- processor plus a main memory, an input medium, and an output medium The current state of the art in Soviet microprocessor technology is a 16-bit single-chip capability in low- volume production and 16-bit chip-sets in serial pro- duction. The Soviets are four to six years behind the United States in microprocessor technology; however, we expect the US lead to increase in the near future as 32-bit monolithic processor technology matures. We have been able to identify 20 types of Soviet microprocessors (table 1). Although 20 is a small number relative to the number of microprocessors commercially available in the West and Japan, the Soviets have judiciously spread their resources across a wide variety of semiconductor devices and fabrica- tion processes. Thus, Soviet design engineers may choose a semiconductor device for a particular appli- cation on the basis of a wide variety of trade-offs in speed, power, radiation resistance, and cost The Soviets can, however, be expected to make in- creased use of complementary metal oxide semicon- ductor (CMOs) devices in the next few years. The best known advantage of CMOS technology is its low power, both in the standby and in the operating mode. The high immunity of a CMOs device to noise encourages design engineers to use a lower voltage power supply. In addition, special processing tech- niques can make CMOS chips more resistant than the widely used negative-channel metal oxide semicon- ductor (nMOS) chips to a specified radiation dose rate 25X1 or fluence. CMOS devices have other advantages over nMOS devices: ? Inherently faster switching times. ? Better resistance to "soft errors" caused by alpha- particle radiation. ? Higher tolerance to transistor-leakage problems. In the light of these advantages, we expect the Soviet military to direct major Ministry of Electronics In- dustry (MEP) resources toward the advancement of their CMOs fabrication processes during the remain- As in many non-Communist countries, US micro- processors have served as the model for many, and 25X1 probably most, Soviet products (figures 2 and 3). In general, Soviet microprocessors reflect various de- grees of similarity to US products (table 1, column 4). However, the Soviets have not copied US counterparts exactly, but rather have adapted the US designs to 25X1 Soviet fabrication processes. We expect other US counterparts to be identified in time. The Soviets also have demonstrated an indigenous design capability in microprocessors, according to evaluations of a K587 device that was received by a US firm in 197825X1 Perhaps the most striking aspect of the list of Soviet microprocessors (table 1) is the preponderance of bit- slice Zdevices. We believe that the Soviets' preponder- ance of bit-slice devices resulted from deficiencies in z Bit-slice devices and chip-set microprocessors implement the functions usually associated with a monolithic (single-chip) inte- 25X1 25X1 LJ/~1 25X1 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Table 1 Soviet Microprocessors Designator K532 Technology/Width (bits) CMOs/4 a Earli 1976 KS36 pMOS/8 a 1979 K555 TTL/2 a e 1976 K580 HMOs/8 1978 K581 HMOs/16 1979 K582 IIL/4 a 1979 K583 IIL/8 a 1980 K584 IIL/4 a 1977 K586 HMOs/ 16 1980 K587 CMOs/4 e 1978 K588 CMOs/16a 1977 __ K589 - --- STTL/2aa - _ 1977 K 1800 ECL/4 a 1982 K 1801 K 1802 HMOs/ 16 STTL/8 a 1980 1981 K 1804 STTL/4 a 1981 K1810 HMOs/16 1983 '? SOS/? 1979 '? HMOs/8 1983 ? HMOs/8 1983 ~+ Bit-slice device. ~ It is not certain that K555 is a microprocessor. Soviet open-source literature in 1977 identified it as atwo-bit-slice microprocessor; but a 1984 open-source catalog equates the K555 family to the Texas Instruments (TI) SN74LS series, which does not have a micro- processor product. Soviet Agat is modeled after US Apple microcomputer architecture. ~ May have been originally TTL. A G 1 l2 microprocessor was mentioned in a 1978 Soviet publica- tion; we suspect that it is actually a microcomputer. MOS =metal oxide semiconductor CMOs =complementary MOS pMOS =positive-channel MOS HMOs =negative-channel MOS TTL =transistor-transistor logic STTL =Schottky-clamped TTL ECL =emitter-coupled logic SOS =silicon on sapphire IIL =integrated-injection logic Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 secret Soviet microelectronic fabrication capabilities during the 1970s and early 1980x. nlthough single-chip microprocessors are cheaper, smaller, and more reli- able, they also place more-stringent demands than do multichip microprocessors on the production equip- ment and the overall fabrication process. The Soviets' deficiencies in the production of semiconductor de- vices would also explain their usage of bit-slice archi- tectures in metal oxide semiconductor (MOST technol- ogies -something that is not done in the West or Japan, because it is simply not efficient or cost cff~cctivc We believe that all of the Soviet microprocessors listed in table 1 are in at least limited production. A US market analysis firm estimated that over 60 million microprocessors were shipped by companies in the non-Communist world during 1983. On the basis of fragmentary information, we suspect that the pro- duction volume of Soviet microprocessors lags this figure by two to three orders of magnitude. Ironically, law production volumes of microelectronic devices may hinder advances in Soviet microelectronic pro- duction technology. Major US manufacturers have attributed a significant portion of their high yield and production technology advances to their very large production volume, which quickly exposes the manu- facturing processes that arc the leading causes of rejection There arc reliable reports reflecting Soviet interest in or development of microprocessor applications for their military. We do not have information at this time that a Soviet microprocessor is currently de- 25X1 ployed in or designed into any specific Soviet military system. There is a great potential for using small fast microprocessors wish low power requirements in mili- tary applications, and we believe that it is simply a matter of tune before we obtain firm evidence that the Soviets arc so using them 25X1 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Secret Figure 3 Soviet Copies of US Microprocessors, 1971-83 United States USSR Intel-8088 ~~ 4 years ~ (?) Intel-8086 ~ 6 years K1810 Intel-8085 ~ 7 years (?) Motorola-10800 ~ 7 years K1800 AMD-2901 ~ 6 years K1804 GI CP-1600 ~~ 4 years ~ Intel-4004 World's first microprocessor 1971 72 74 76 78 80 82 83 Semiconductor Memories The Soviets are three to five years behind the United States in semiconductor random-access memory (RAM) technology (figure 4), but when production capability and quality are considered we assess that the US lead can be extended by at least several more years. The Soviets' literature indicates that they are even further behind the United States in read-only memory (ROM) technologies, including programma- ble ROMs (PROMS) and erasable PROMS (EPROMs) Over 50 semiconductor RAMS and over 50 ROM- type memory devices, most in more than one version, have been identified in Soviet catalogs and open literature. As with their microprocessors, the Soviets have spread their semiconductor memories across a variety of technologies, including to - and high-speed emitter-coupled logi the Soviets began using sma -capacity semlcon uctor RAMS in the late 1970s. There were "...adequate supplies of accept- able quality ..." of 16-Kbit (1K = ],024) RAMS in mid-1980 The Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 secret Figure 4 Semiconductor Memory Technology: United States Versus USSR, 1971-878 Capacity in bits United States ~ USSR 1,024 K 1971 72 74 a R;aed on meta Loxide semiconductor drnumlr random-access memories. 4 years - >~ 25X1 303319 12-84 Soviets probably had a 64-Kbit dynamic RAM in sample production by the early 1980s. In the United States and Japan, the 256-Kbit RAM is now being produced and prototypes of a 1-Mbit RAM have been made with improved photolithographic techniques in- stead of X-ray or E-beam lithography, contrary to what was frequently forecast in the technical litera- ture. We expect that the Soviets will not put 256-Kbit RAMS into production until the late 198 Magnetic Cores Magnetic core memories have several features that are attractive to the military. First, magnetic core is a random-access memory; the time to retrieve data is the same no matter where the data are stored in the memory. Second, core memories are nonvolatile; when power is disconnected or interrupted, core memories do not lose their information as many semiconductor memories do. Third, core memories are available in systems that have been hardened against nuclear radiation. Fourth, cores require no power in order to retain data in a standby mode. On the other side of the ledger, core memories require much more physical 25X1 space and more power to operate, and cost much Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Secret Figure 5 Magnetic Core Memory Technology: United States Versus USSR, 1955-85 1?xternal core diameter in mils - United States Ussk can be stored at each intersection of these wires. Plated-wire memories have several characteristics that are attractive to military-system designers. Plated-wire memories provide protection against elec- tromagnetic pulse (EMP) and have good radiation hardening properties. These memories are nonvolatile and can be made with a nondestructive readout (NRDO), thus providing additional protection for the stored data. However, plated-wire memories are ex- tremely expensive-on the order of $1.00 per bit, whereas most other mPmorv technologies are just pennies per bit or less. reported in the mid- 25X1 s on t e ov~ets pure ase of a turnkey factory more, than a semiconductor memory of like capacity. The trend in US military applications is toward semiconductor memories, with backup or shadow memories in critical applications Magnetic core technology is frequently gauged by the external diameter of the basic ferrite toroid'-the smaller the diameter, the greater the memory capaci- ty per unit of area. Using this core diameter as a figure of merit, we find that the Soviets lag the United States in magnetic core technology by about five years (figure 5). We believe that this lead was maintained during the 1970s, even though the empha- sis on the development of magnetic cores in the United States decreased as fast as the popularity and usage of semiconductor memories increased Plated-Wire Memories A plated-wire memory is arandom-access memory that consists of a plane of parallel wires electroplated with a thin film of magnetic material and overlaid by a set of transverse word lines. One bit of information from the Japanese to produce plated-wire memories. Included in the agreement was extensive documenta- tion that should enable the Soviets to build duplicate plants if they so desired. The Japanese turned the new factory in Yerevan over to the Soviets in December 1976-the same year that the Japanese firm discon- tinued plated-wire memories. The output capacity of the Soviet plant was rated at 120 million bits of wire memory per year. This was actually higher than the capacity of the plated-wire plants in Japan, because of the high requirements set forth in the contract with the USSR in 25X1 the first 18 months t e Yerevan plant pro uce ewer than eight million bits of plated-wire memory. This shortfall was attributed not to the Japanese equip- ment but to the low quality of Soviet base metals. [n the United States, plated-wire memories have been developed for the guidance computer in the Polaris and Poseidon missiles. Plated-wire memories have been used also in the US Minuteman weapon system computer as well as briefly in a few US and Japanese commercial computers. Ina 1982 list of US space computers being used by the National Aeronautics and Space Administration, nine of 17 systems used plated wire for their main memory. It is reasonable to expect that the Soviets would also use their plated- wire systems in ruggedized mobile applications with modest memory capacity requirements 25X1 25X1 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Secret The Soviets are four to six years behind the United States in 16-bit minicomputer technology. They may realize their first 32-bit superminicomputer by the end of 1985 Following along the lines of the successful CEMA cooperative program in mainframe computers, the Council of Principal Constructors of Minicomputer Systems was created in 1974 in an attempt to coordi- nate minicomputer development within CEMA coun- tries. The Soviet Union assumed the major role and developed four new minicomputers: the SM-1, -2, -3, and -4. This group, and possibly the SM-5, constitutes the first eneration of SM minicomputers, SM-I (table 2). The SM-1 and the SM-2 are modeled after the Hewlett-Packard HP-2100 architecture and are pri- marily for process and production control as well as real-time applications. The SM-2 is essentially an SM-1 with an improved central processing unit and more main memory. These two machines are unusual examples of Soviet plagiarism in that the resulting Soviet systems are not compatible with HP software. The Impuls Scientific Production Association devel- oped the SM-1 and the SM-2; the SM-1 is produced at the Ministry of Instrument Making, Automation Equipment, and Control Systems (Minpribor) plant in Orel, and the SM-2 is produced by the Impuls association in Severodonetsk. Impuls is currently pro- moting modernized versions called the SM-1M and the SM-2M. The great majority of the publicity has been given to the SM-2M, which has been identified in a Soviet brochure with 25 different configurations including 22 dual-processor versions. In an open- source article, the deputy general director of Impuls states that different computer architectures will be used to overcome the "disadvantages" of traditional minicomputers such as the SM-2. He then describes the SM-1210 multiprocessor and the PS-3000 array processor, which Impuls may now have in production. As in the United States, the trend in the USSR is toward multiprocessor systems to avoid the through- put bottleneck caused by sequential processing on a uniprocessor system and for having put these machines into serial produc- tion. The SM-3 and the SM-4 are modeled after the low end of the US Digital Equipment Corporation (DEC) PDP-11 minicomputer line and are intended primarily for small scientific and engineering applica- tions. The SM-3 and SM-4 can execute DEC soft- 25X1 ware without modification. The newer SM-4 with 256 Kbytes of main memory can execute DEC's RSX- 11 Moperating system. The popular UNIX operating system, which was originally written in the United States for DEC PDP-11 minicomputers, also is known to be available in the USSR. Copies of PDP-11 minicomputers are also being produced in Bulgaria, Czechoslovakia, East German Hungary, Poland, Romania, and Cuba 25X1 Being copies of US systems, the Soviet SM systems 25X1 provide a good basis for a comparison of 16-bit minicomputer technology. According to the year of first installation for the SM-1 through the SM-5 (figure 6), the Soviets are about four to six years behind the United States in general purpose 16-bit minicomputer technology. However, if we were to go by the quality and quantity of production, several more years could be added to this US lead. In 1979 Soviet official were having "yield and reliability" pro ems m t eir SM production line. They hoped to resolve these problems and to be producing 1,000 SM units per year by 1982, the bulk of which were to be SM-3 and SM-4 models. In the fall of 1981 ~ the Kiev production plant was ,500 SM-4 units would be produced that year. Even the more optimistic forecast is quite modest, considering that the SM is the primary minicomputer series for the entire Soviet Union. By comparison, after about a decade of production, DEC had almost 100,000 PDP-11 minicomputers installed worldwide by the end of 1981. In late 1982 a reliable source reported that the SM-4 would be replaced by the newer SM-1420-also called the SM-S-mini- 25X1 25X1 25X1 computer in 1983 (figure 7) 25X1 25X1 the Ministry of Electronics 25X1 n ustry as eci ed to develop, manufacture, and sell its own line of minicomputers in the Soviet 25X1 In 1981 the USSR State Prize in Technology was awarded to 10 Soviet managers and engineers for having developed the SM-3 and SM-4 minicomputers Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Table 2 Characteristics of Soviet SM-I Minicomputers Characteristic SM-1 SM-2 SM-3 SM-4 SM-5 Average speed in kops a 130 155 135 215 400 Main memory, Kwords 4 to 32 32 to 128 16 to 28 16 to 124 128 to 2,097 Instruction time in microseconds Fixed point Addition Multiplication Division Floating point Addition Multiplication 110 b 23 Division - 40 ~~ These speeds, except for the SM-5, whose speed is estimated, were cited in Soviet literature and seem to be more realistic than the 250 kops for the SM-1 and 800 kops for the SM-4 which are frequently quoted in open literature. n Probably implemented in software. Implemented in software. kops: 1,000 operations per second. Kwords: 1,024 words with 16 data bits per word. Not known. Note: Soviet open literature has placed the SM-5 in the SM-I family; however, a Soviet export brochure (circa 1983) claims that the SM-5 is in the SM-II family of minicomputers. 410 ~ 34 Union. The Soviet official said that MEP had made "exact replicas of the DEC minicomputers." We do not know at this time whether the Soviet official was referring to DEC's 16-bit PDP-1l minicomputer line or to its newer 32-bit VAX superminicomputer fam- ily. Minpribor has been the primary manufacturer of minicomputers in the USSR to date. If MEP begins producing minicomputers, serious interministerial conflicts could easily arise between MEP and Minpri- bor, because MEP is also the primary (possibly the sole) source of microelectronic com onents for Min- pribor minicomputers At an international conference in 1981, an East European scientist stated that the Soviets were devel- oping aminicomputer that would be compatible with software for DEC's VAX superminicomputer. Al- though information is very sparse, we believe, on the basis of past Soviet accomplishments, that the Soviets will produce their first 32-bit minicomputer by the end of 1985, and that this machine will be compatible with DEC VAX software. DEC's first VAX, the 1 1 /780, introduced in 1978, is a complex machine requiring 23 printed-circuit boards for its central processor. We believe that the Soviets will have an easier time copying the DEC VAX 11/750, which, with its four-board processor, is much less complicat- ed than the 1 I /780. Nairi-4 Minicomputer Although the SM series has no known special versions for military applications, at least one civilian general purpose minicomputer, the Nairi-4, has been modified Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Secret Figure 6 Soviet Copies of US Minicomputers ~ United States ~ USSR PDP-11/44(?)a ~~-- 3 years --~ SM-5 PDP-11/34A ~~ 4 years ~{~ SM-4b PDP-11/40 ~~ S Years --!!~ SM-4b PDP-ll/OS HP-21MX ~f HP-2100 ~~- 1971 72 74 76 78 80 82 83 Year of production May he a copy o(PDP-I 1/44, less [he cache memory. n The SM-4 is known to have been a copy of DEC's PDP-11/34A since a[ least [he early 19gOs. The original SM-4 may have been a copy of [he PDP-Il/40. for such use special Nairi-4 with gold contacts was being built in 1977 for the Soviet military. Indeed, the Nairi-4 has several other conspicuous characteristics that would make it useful for many fixed land-based military applications. One is the use of plated wire to provide a nonvolatile main memory having a nondestructive readout. Plated-wire memory also has good radiation- resistance characteristics The Nairi-4 minicomputer has used a magnetic drum for bulk storage. Magnetic drums were used in many early automated subsystems for the US military, but they have largely been replaced by fixed-head mag- netic disk systems, by core memories, and, more recently, by semiconductor memories. The physical size of the Nairi-4 computer would limit it to applica- tions at fixed ground-based sites or on large mobile platforms. A new version called the Nairi-41 was briefly mentioned several times in the open literature in 1982. A nonoperational Nairi-41 was displayed at the 1983 Leipzig Spring Fair (figure 8). According to a technical brochure, the Nairi-41 has a 540-nanosec- ond cycle time for register-register instructions and up to 256 Kbytes of semiconductor memory PS-2000 Array Processor The PS-2000 is a microprocessor-based array proces- 25X1 sor system developed in the late 1970s at the Control Problems Institute, Moscow, in coordination with the Impuls Scientific Production Association in Severo- donetsk (figure 9). With Impuls' involvement, it is likely that the PS-2000 will operate only with the SM-2 and the SM-2M minicomputers. Soviet litera- ture states that the PS-2000 consists of eight, 16, 32, Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Figure 7. Soviet SM-1420 SM-5 Minicompute or 64 "processing elements" that can be dynamically rrrrranged. In Soviet literature, "processing element" usually refers to a bit-slice microprocessor component. The word lengths in the PS-2000 (12, 16, or 24 bits) suggest a 2-bit or a 4-bit device as the basic building block. This hypothesis, in conjunction with the opera- tional speed of the PS-2000 and the time at which it was developed, suggests that the K5894 microprocessor is used in this machine n Suviet article announced that the PS-2000 was able to halve the computation time of a modeling problem executed on a uniprocessor minicomputer-probably an SM-2. This increase in performance seems more realistic than the extraordinarily high speeds claimed in the Soviet press since 1981. Even so, the PS-2000 is important because it reflects the Soviets' interest and progress in array processor technology The term "mainframe," which originally referred to the central processing unit and sometimes the main memory, is now generally used to describe a class of computers exemplified by the IBM large-computer line. nlthough far surpassed in numbers by the Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Secret microcomputers and minicomputers sold today, the mainframe class still accounts for over 50 percent of the dollar volume in computer sales worldwide Historically the Soviets, together with their CEMA partners, have placed great emphasis on, and have invested significant resources in, the development and production of their series of software-compatible mainframes known as the Unified System (ES: Edin- aya Sistema) or as the Series (Ryad). the Soviets patterne t e first Ryad family after the highly successful IBM System/360 line; Ryad-2 developers used the IBM System/370 as a design basis. The adoption of a proven commercial system was aloes-risk decision enabling the Soviets to circumvent much of the R&D costs associated with the development of new comput- ers as well as most of the software development costs. In both Ryad-1 and Ryad-2, Soviet models were first installed approximately seven to eight years after their IBM counterpart (figure 10). Several improved versions of the Ryad-1 series were put into serial production during the 1970s in Bulgaria (ES-1022B) and the USSR (ES-1022, -1033, and -1052). Other Ryad-1 machines included the Czechoslovak ES- 1021, which was not compatible with IBM Sys- tem/360 software, and the Polish ES-1032~~ The Ryad-2 family initially consisted of five members that entered production in the late 1970s-the ES-1025 (Czechoslovakia), ES-1035 (USSR, Bulgar- 25X1 ia), ES-1045 (USSR), ES-1055 (GDR), and ES-1060 (USSR;--plus a sixth, the ES-1065 (USSR), which was in production by 1982.` The ES-1060 slipped from the Ryad-1 program because of technological problems including overheating of its fast integrated circuit logic, and is now considered part of the Ryad-2 scrics (figure 11). Three improved Ryad-2 models .5X1 were in or nearing production in 1984: the Czechoslo- 25X1 vak ES-1026, the East German ES-1056, and the Soviet ES-]061 Figure 12 shows a performance comparison of Ryad-2 mainframes and some IBM System/370 computers. The values for the speed-operations per second s (ops)-and the memory size are taken from Soviet and Western literature. Although lagging US mainframe Hungary produces the ES-1015 minicomputer, which is also listed as a Ryad-2 machine but is not compatible with IBM System~370 Operations per secon and other single measures of efTectiveness are an oversimplification, as system performance is actually a comnlex fir~f many factors, especially the specific applica- 25X1 25X1 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Figure 10 Timetable: IBM and Soviet Ryad Mainframes, 1964-84 ~ IBM ;, Ryad S/=Sys[em S/360 delivered Ryad- I revealed S/370 delivered 303X announced 7 years ~ Ryad-2 revealed Ryad-1 Ryad-I delivered announced Ryad-3 Ryad-3 revealed announced Ryad-2 delivered Ryad-2 announced - -I- L---~-L~ -1_ i_ _.. i_- 1964 66 68 70 72 74 76 78 80 82 84 technology, the Ryad-2 family still offers the Soviets and their allies a fairly wide range of computing capability. Table 3 shows some Ryad-2 technical specifications. We believe that the Soviets are produc- ing Ryad-2 machines in sufficient numbers to satisfy at least priority users. However, when one considers quantity, quality, and performance/cost ratios real- ized by the general-user population, the Soviets are about four years further behind the United States than the seven to eight years indicated by the dates of first delivery Based on open literature, figure 13 illustrates those IBM system software products that we believe are in use, with some name changes, in the Soviet Bloc. Open literature suggests that the Soviets are using most IBM system software products released prior to 1978; the most ,notable exception is Multiple Virtual Storage (MVS) In November 1981 the Soviets announced a new Ryad mainframe, the ES-1036; a scale model of this com- puter was exhibited at the Budapest Spring Technical 308X delivered 308X announced 43XX delivered Fair in 1983. A Soviet export brochure obtained in 1984 states that the ES-1036 represents the first stage in developing Ryad-3 computers. According to open literature, the ES-1036 can execute up to 400,000 operations per second, has a main memory of 2 to 4 megabytes, has an 8-kilobyte buffer (or cache) memo- ry, and will have a virtual machine operating system. We suspect that Ryad-3 computers will be copies of the IBM 43XX and 303X families. In the spring of 1982, the Soviets also briefly mentioned that they were developing a prototype ES-1061 computer, which will be a modernized version of the ES-1060. The ES-1061 was to enter serial production in 1984, according to Soviet open literature. Other new Ryad designators include: ? Hungary: ES-1016, -1017. ? Czechoslovakia: ES-1026, -1027. ? Poland and/or USSR: ES-1034, -1047. ? GDR: ES-1056, -1057. ? USSR: ES-1046, -1066, -1067, -1077, -1087. Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Secret We believe that at least the ES-1036 and -1061 were in production in 1984, and that most of the machines listed above will enter production over the next two to three years Status We believe that the Soviets will not develop their first digital electronic supercomputer until 1985 at the earliest. The lack of a modern supercomputer can restrict or slow advanced R&D programs as well as civil and military applications, such as energy explo- ration and strategic missile defense, that require a Elbrus Computer The USSR does not have a supercomputer in a class with the US Cray-1 or Cyber-205. The machine most likely to become the first Soviet supercomputer will probably come from the Elbrus project at the Institute for Precision Mechanics and Computer Technology (ITMiVT) in Moscow. The Elbrus-1 multiprocessor computer was created and fostered during the 1970s by V. S. Burtsev, the director of ITMiVT. The Elbrus-1 system employs a tagged architecture with a stack organization and an addressing structure similar to those of the Burroughs B-6700 system first deliv- ered in 1970, in the United States. However, Elbrus-1 is much more ambitious in that it reportedly has from huge number of computations. Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Figure 12 Mainframe Performance: United States Versus USSR -- IBM Ryad 195 168 165 135 izs 1 111.11 I Ill Main memory in kilobytes 1.1.1111 1060 1045 1055 (East German) 11111111. 1 I I1 one to 10 central processing units (CPUs) operating asynchronously ? with up to four input output proces- sors and 32 memory modules interconnected through a series of crossbar switches (figure 14). In 1982 a " nn asynchronous multiprocessor assigns tasks to difTerent proces- surs, using a set of indicators to designate which processors are free and which arc busy. Typically, a processor will operate on a task until it has completed the task or until it is interrupted by the "all" of the Elbrus-1 com- puters o w lc e was aware are single-processor models except one machine that has two CPUs. We suspect that this small number of processors in Elbrus computers being delivered may be due to the lack of a generalized operating system or to troubles with such Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 secret Tsble 3 Technical Specifications for Soviet Ryad-2 Mainframe Computers Estimated date of first delivery 1980 1977 1979 1979 1978 1984 1982 Production plant Cakovice Minsk-Brest/Sofia Kazan Dresden Minsk Minsk Minsk Country Czechoslovakia USSR/Bulgaria USSR GDR USSR USSR USSR Processor Speed (1,000 operations/ secontn 35 160 650 450 1,000 2,000 3,000 Fixed add time (?s) 5-13 4.5 0.7-0.85 0.6-2.7 0.25-0.30 a 0.12 Fixed multiply time (?s) 95-220 23 2.8-3.4 3.4-5.2 1.5-1.8 _ a 0.6 Floating point add time (?r) 50.0 95.0 1.9 1.6 0.80 a 0.24 Floating point multiply time (?r) 9.7 19.8 2.8 2.7 2.3 a 0.30 Main memory Capacity (Mbytes) 0.1-0.5 0.25-1 1-4 0.25-4 0.5-8 1-8 2-16 Cycle time (ns) 1,250 800 840 1,140 _ 800 a e Access time (ns) 500 550 650 a e e 870 Bytes fetched per cycle 8 8 8 8 8 e e Microprogram control memory Capacity (Kbytes) 48RW 7R0 + 1RW 8 48 e a Cycle time (ns) 380 200 120-380 135 a e a Access time (ns) a e a 140 65 e e Length of word accessed (bytes) a a 8 8 16 e e Cache (scratch pad) memory Capacity (Kbytes) X X 8 X 8 e 32 Cycle time (ns) X X 120 X 135 e e Access time (ns) X X 72 X 65 e a Length of word accessed (bytes) X X 8 X 8 e a Channels Maximum number 2 5 6 5 7 8 e Total transfer rate (kbytes/s) a 1,200 5,000 6,000 9,000 a 15,000 Maximum number 1 4 (5) 6 (4) 6 (6) 6 a (16) Transfer rate (kbytes/s) 33 740 (1,500)6 (1,500)6 (1,300)6 a (1,500) Byte-multiplex channels Maximum number 1 1 2 2 2 e e Transfer rate (kbytes/s) 24 40-280 ~ 40-160 ~ 40-1,500 ~ 110-670 ~ a 110-? Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Table 3 Tecbnical Specifications for Soviet Ryad-2 Mainframe Computers (continued) Block-multiplex channels Maximum number X 5 Transfer rate X 1,500 (kht~res/s) Class, per State Standard GOST 16325-76 ll ~~ Data not available. n On these models the block-multiplex channel can be operated as a selector channel. Speed varies depending on numbers and types of operational channels in system. ?s =microsecond = 10-? second; ns =nanosecond = 10-? second. Byte = 8 bits (8 binary digits); Kbyte = 1,024 bytes; kbyte = 1,000 bytes; Mbyte = 1,048,576 bytes. X =equipment not available on model; RW =read/write; RO = read only. ote: peel canons or ya computers vary, sometimes great y, in CEMA literature. None of the values in this table have been confirmed by direct access to a Ryad-2 computer, and we believe that they tend to be overly optimistic. Ryad-2 systems introduced in the late 1970s had ferrite-core main memories; these were upgrad- ed [o semiconductor memories in the early 1980s. Operational parameters for semiconductor devices are used in this table. The performance of the ES-1065 is based upon a uniprocessor configuration. The ES-1026, -1056, and -1061 are modernized Ryad-2 versions of the ES-1025, -lOSSM, and -1060, respectively. an operating system for Elbrus-1 computers having more than two processors. By analogy, the first US commercial supercomputer, the Cray-1, initially was delivered in 1976 with only the most primitive soft- ware support for system management Table 4 lists four Elbrus configurations identified as "standard" in a Soviet brochure. All of the through- put values are quite optimistic; and the maximum main memory capacity is modest relative to US state of the art. On the basis of comments by Soviet scientists and the size of the Elbrus machine, we estimate that between five and 10 Elbrus computers have been built each year since 1979. Cray Research Corporation in the United States delivered an average of seven Cray supercomputers each year between 1976 and 1984 there would be two versions of Elbrus: one for civilian use, and one for the militar the only difference between these two computers would be the method of testing. Ballistic missile defense is an application frequently cited by emigres for the Elbrus computer. It was rumored in Soviet scientific circles around 1978 that an Elbrus was to be installed on an aircraft carrier A new model, Elbrus-2, has been under development at ITMiVT. According to Soviet literature, this ma- chine will exceed 100 million operations per second. Elbrus-2 was mentioned as early as 1977, but we suspect that Burtsev has been busy debugging Elbrus- 1 and is still trying to perfect an Elbrus-2 prototype. A Soviet scientist stated in September 1983 that no Elbrus-2 machines had been produced as of that date. If Elbrus-2 is realized, it will be, we expect, the Soviets' first entry into the supercomputer realm. (S NF~ M-10 Computer In May 1979 M. A. Kartsev published a description of a synchronous multiprocessor system called the M- 10 that he had designed at the Institute of Electronic 25X1 25X1 25X1 25X1 25X1 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 secret Figure 13 IBM System Software in Use in CEMA Countries Release dates of IBM software products 1966 Batch Single task 1967 Spooling Multiprogramming 1968 Multitasking 1969 1970 1971 Teleprocessing Timesharing 1972 Virtual storage Virtual machines 1973 1974 Parallel operations 1975 1976 1977 1978 Program products 1979 High-performance option 1980 System product 1981 1982 1983 Extended architecture 1984 Extended virtual machines System/360 announcement DOS/VSE Program products OS/VS1 Program product CP/CMS CP/40 OS/VS2 R2 MVS MVS/SE Program product VM/SP System product MVS/SP V2 MVS/XA Systems believed to be in use in CEMA countries Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 "fable -t "Standard" Soviet }~:Ibrus-1 ('onfikurations Number of (~PUs I ' 4 10 Throughput in Mops 1.5 3.0 5.5 12.0 Storage capacity, 576 1,152 2.304 4,608 Kb~'tcs M1lenwn blocks I ' 4 8 h1cnwn~ commutators I 2 4 8 Input~autput ~ ~ 2 4 processors U:ua communication (Il ~ 2 16 processors (~Pl~: central processing unit. Mops: millions of operations per second. Kb~tc: 1,024 bytes 8,192 bits. 1 t uption;il. ('omputcrs ~Il?UM1 in Moscow.' Kartsev said that up to seven M-10 computers can be joined together in a ' In a synchronous multiprocessor system, the processors operate in a lockstep manner usually timed to a worst-case operation. This procedure greatly reduces the management overhead associated with asynchronous systems but can Icad to inefficiencies for very operations per second single synchronous complex. Another open source suites that Ryad peripheral equipment can be used with the M-10. According to Kartsev, this 32-bits word computcr has an average speed of over 5 million Kartsev described the control unit of the M-10 as being able to dynamically adapt the number of pro- cessors under program control as a function of the word length. This approach is similar to a technique used in the US Illiac-IV supercomputer, which made it possible either to execute with 64 processors on word lengths of 64 bits or to use 128 processors on 32- bit words. Having these alternatives is useful in applications that arc suited to parallel algorithms and have variable numerical range requirements computcr available in the Soviet Union during the Although the new Ryad-2 ES-1065 computer may be faster, the M-10 may have been the most powerful 25X1 25X1 25X1 25X1 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 ~ecre~ 1983, we suspect that the design philosophy of this domineering personality is well entrenched at the institute that he directed. We expect that improve- ments and variations on the basic M-10 architecture t oug artsev ie m pn The number of experienced Soviet software program- 25X1 mers who also are cleared for classified projects may still be insufficient, thus probably leavine manv mili- 25X1 will continue through the 1980s Ye. P. Velikhov, vice president of the USSR Academy of Sciences, stated in late 1983 that he was the focal point for an accelerated program on the development of supercom- puters. This pronouncement is interesting because Burtsev's institute and the Elbrus program are under the control of the USSR Academy of Sciences. Until now, Burtsev seems to have had an autonomous reign in pursuit of his high-performance Elbrus computer. Many sectors of Soviet society, especially the military, are known to be anxious for a supercomputer Velikhov's appointment a oca point or managing a supercomputer program may be the first step in opening up the development of these machines to other or anizations within the academ We estimate that, in general, the Soviets are five to 10 years behind the United States in the implementation of large multiuser and real-time software systems as well as in - chniques for various industries There are numerous causes that contribute to the Soviet software lag. Some of the problems frequently cited by Soviets with access to programmers in the USSR are: ? The Soviet hardware lag. ? A belated appreciation of, and belated emphasis on, software. ? A poor or nonexistent vendor-user feedback loop in the USSR. ? Low pay for programmers relative to other technical personnel. ? Poor software development tools. ? A Soviet preoccupation with meeting quotas-usu- ally at the expense of quality control. ? Duplication of work due to the excessive compart- mentalization of software routines written at many facilities tary projects not completed on time Software is one area where technology transfer from the West can help the Soviets close specific gaps with 25X1 quantum jumps. Software programs are conveniently stored on relatively small media such as diskpacks, floppy disks, or magnetic tapes-or on solid state 25X1 memory devices, which are even smaller. It is not just classified military software that is of interest to the Soviets; they also can use many commercial software programs to improve their industrial base or to imple- ment military subsystems. Programs are available from thousands of commercial outlets in the non- communist world. Thus, the United States has a major technology transfer problem. If the Soviets were able to obtain a microprocessor from the United States, a team of engineers and technicians would need from one to four years to reverse engineer the device. However, if the Soviets obtained just one copy of a software program, it would be a minor project for even a novice to turn out copies of this program immediately. With the increasing number of comput- ers available to the Soviets that are functional equiva- lents of Western systems, we can expect the Soviets to 25X1 2ox~i continue, and probably to increase, their legal and illegal acquisition of Western software systems~~ 25X1 Magnetic Disks The Soviets are about 10 years behind the United States in high-performance magnetic disk technology. This is one of their most serious computer hardware deficiencies and it is limiting the performance of their computer systems in many applications Figure 15 illustrates the significant lead that the United States has in magnetic disk devices. The Soviets have announced their own 200-Mbyte disk drive (ES-5080)-about four years after the Bulgari- ans began low-volume production of comparable equipment (ES-5067) and about 10 years after the Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Secret Figure 15 Magnetic Disk Technology: United States Versus USSR, 1965-88 ? US Bulgaria USSR 3380-2~ 8 years -- 5066 10 years ~ 5061 7 years ~ 5052 __I I I I I I I I I I I I I I I I I I I I I I I I 1965 66 68 70 72 74 76 78 80 82 84 86 88 Year of first delivery advent of IBM's 3330-11 counterpart. Bulgaria and into computer systems believe that system perfor- the USSR have, however, been able to adopt some mance above approximately 5 million operations per mechanical features of disk drives, such as voice-coil second would be severely hampered without further motors, in a very timely fashion advances in disk technology. We believe that this Although the low performance of Bulgarian and Soviet disk drives may impose some inconveniences now, Western engineers who integrate disk memories Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 ~ecrcr situation is currently slowing or negating many appli- cations on the Elbrus-1 multiprocessor system and will also hinder system performance on Ryad comput- ers beyond the current top of the line, the ES-1065. the Soviets are p acing a tg prtortty ono tatmng now-how for the production of high-performance magnetic disks, probably via Western Europe or Japan. We believe that the Soviets also are seriously pursuing optical storage technology to alleviate this bottleneck in system performance Magnetic Tapes According to open literature state of the art in magnetic tape drives in CEMA countries is 1,600-bits-per-inch (bpi) density with a data transfer rate of 189 kilobytes/second. IBM first released comparable equipment in 1966-an 18-year diH'erential. A density of 6,250 bpi at 1.25 mega- bytes/second has been used in the United States since 1973. In March 1984, IBM announced its new high- performance magnetic tape drive, Model 3480, sched- uled for delivery in 1985. The new 3480 will have a linear density of approximately 19,000 bpi and a data transfer rate of 3.0 megabytes per secon~ Magnetic Bubbles The Soviet Union possibly had a prototype 64-kbit a magnetic bubble memory (MBM) by 1980 and a 92- kbit prototype by 1981. By comparison, at the same "MBM size has had a confusing evolution. Early MBM products used "bit" quite loosely, generally rounding a number to the closest 1,000 bits. Later products reverted to the "normal" powers-of-2 sizing for memories. In this MBM section, we use lk as ap rp Oxi_ time in the United States, 256-kbit MBMs were in production and 1-Mbit MBMs had been developed in the laboratory. MBM is an attractive storage technol- ogy for military applications because bubble memo- ries exhibit very good performance in severe environ- 25X1 ments presenting extremes in dust, shock, heat, humidity, and radiation. Bubble memories are nonvol- atile and have a reputation for high reliability relative to magnetic tape and disk a ui ment which use electromechanical drives 25X1 25X1 Today, the Soviets are trailing the United States in all aspects of electronic computer technology. If we include the quantity and quality of computer produc- tion, the US lead averages several years more than is indicated by just comparing the dates of first installa- tion of functionally equivalent US and Soviet systems. As a result of the more advanced microelectronic technology and computer packaging techniques in the United States as well as the poor state of the art in Soviet peripheral equipment, we expect the US lead to 25X1 increase by one to three years in all ma'or electronic computer technologies by 1986 25X1 It is difficult to assess accurately the impact of the Soviets' lag in computer technolo on their develo - ment of military systems 25X1 25X1 It is rare 25X1 when computer 25X1 tec no ogy is In ertng t e eve opment of a specific military program, However, at the high- 25X1 per ormance en o computer technology, at least, we can speculate with reasonable confidence that mili- tary systems requiring high-throughput computers have been negated, delayed, or reduced in capability because of the Soviets' deficiency in this area. The impact would have been serious on large high-speed 25X1 computational problems such as ballistic missile de- fense and on high-volume, high-speed data transfer applications such as real-time command, control, and communications systems requiring large data bases. Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 The Soviet scientific community has frequently ex- pressed the opinion that the lack of a supercomputer is hampering many R&D projects such as in computa- tional physics and chemistr Apart from large scientific computers, the impact on military systems of the Soviet lag in computer tech- nology is more difficult to judge; here the lack of information is more of a barrier. One may argue that the tr~iditionally conservative design philosophy asso- ciated with Soviet military systems has not stressed their computer technology. Another possibility is that system requirements were kept modest in line with the Soviets' knowledge of the limitations of their comput- ers. The truth is probably a mixture of both hypotheses The Soviets tend to have less reliable computer sys- tems than the United States or Japan because Soviet microelectronic components are less reliable and Sovi- et quality control is generally weaker. An example of how this reliability can affect a critical system is ICBM design. The Soviets use triply redundant com- puters on board their ICBMs. Although individual computers have failed during flight tests, there have been no mission test failures to date attributed to the onboard computer complex. By contrast, the United States has used a single computer for navigation, guidance, and control functions on board its Minute- man and MX missiles. Ironically, today US contrac- tors arc reportedly going to redundant computer systems in many designs for increased reliability. For example, the F-16 flight control system and the navigation system on the Navstar satellite will both have triply redundant processors on board The Soviets understand and appreciate the potential impact of high technology on weapon systems. Auto- mation in the Soviet military sector will grow steadily and become an integral part of new system designs. We suspect that the Soviets during the 1980s are following the US approach from the 1970s; that is, expanding the use of mil-spec minicomputers for tactical military applications, while continuing to decrease reliance on special-purpose computers. As the reliability of Soviet microprocessors in severe environments improves, they will become more preva- lent in Soviet tactical systems. Although automation in the Soviet military is expected to increase through- out the 1980s, the rate of increase is expected to be slower than in the United States, especially for mobile tactical systems. Over the next three years we expect that the Soviets: ? Will improve the quantity and quality of their semiconductor memory devices and microprocessors. ? Will phase in the production of Ryad-3 mainframes. ? Probably will build their first 32-bit minicomputer. ? Probably will build their first supercomputer. ? Will fall further behind in all areas of computer technology Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8 Secret Secret Sanitized Copy Approved for Release 2010/11/29 :CIA-RDP86R00995R000501130001-8