Encyclopedia of 20th Century Technology

(Note: Sample material is taken from uncorrected proofs. Changes may be made prior to publication.)

Computer Memory, Early

Mechanisms to store information were present in early mechanical calculating machines, going back to Charles Babbage's Analytical Engine proposed in the 1830s. It introduced the concept of the 'store' and, if ever built, would have held 1,000 numbers of up to 50 decimal digits. However the move toward base-2 or binary computing in the 1930s brought about a new paradigm in technology - the digital computer, whose most elementary component was an On-Off switch. Information on a digital system is represented using a combination of on/off signals, stored as binary digits (shortened to bits): 0s and 1s. Text characters, symbols, or numerical values can all be coded as bits, so that information stored in digital memory is just 0s and 1s, regardless of the storage medium. The history of computer memory is closely linked to the history of computers but a distinction should be made between primary (or main) and secondary memory. Computers only need operate on segment of the data at a time, and with memory being a scarce resource, the rest of the data set could be stored in less expensive and more abundant secondary memory. This focus of this essay will be on primary memory technology, with attention to the early developments and the technological hallmarks.

A method of storing data on rotating "drums"- an electromechanical device- had emerged in the late 1930s. John Vincent Atanasoff successfully conjured a device consisting of a rotating drum in which 1600 capacitors were placed in 32 rows. Later, magnetic coatings were used on drums and during the War this storage emerged as a reliable, rugged, inexpensive but slow memory device. Engineering Research Associated (ERA) produced a commercial device in 1947 that could store 65000 32-bit words (over 2 million bits). Magnetic drum storage was to survive in later computers for secondary data storage, and later in disk platter form of continually shrinking proportions. It was the predecessor to the Hard Disk Drive, the ubiquitous component of all later computers.

Early digital computers made use of electromechanical relays for binary logic. Early relay computers included Howard Aiken's Automatic Sequence Controlled Calculator. Developed at Harvard University with the assistance of a group of IBM engineers, the Mark I was completed in 1944 and could store 72 numbers mechanically using electromagnetic decimal storage wheels, with the programming and data sequencing instructions fed into the machine via four punched tape readers. However the bulk, power requirements, and the delay time in the operation of relays were performance-limiting factors.

Thermionic valve (or vacuum tube) technology matured during the War. Valves could operate as simple switches and a switching speed increase in the order of 1000 times over relays was realised. Although there were earlier prototypes, the completion of the ENIAC (Electronic Numerical Integrator and Computor) heralded the era of electronic computers. Built for the US Army by the Moore School of Electrical Engineering at the University of Pennsylvania it was only completed in late 1945 after the war. It consisted of 18 000 valves in which its core logic was implemented, as well as 1500 relays which were used to store initial data loaded from a punched tape reader. 20 'Accumulators' to store binary numbers during calculations were implemented with valve logic. Subsequently, purpose-built binary data storage valve technology emerged, although not without difficulties in production (see Selectron valve, below).

The first generation of computers such as the Mark I and the ENIAC were purpose-built, and had the sequence of instructions 'hard coded' or pre-programmed, for example with paper tape input. The British Colossus computer was another special purpose machine using electronic valves that implemented algorithms developed by Alan Turing and colleagues to crack encrypted German U-Boat radio communications during the Second World War. However mathematician John von Neuman approached the issue of storing programming code at a more fundamental logical level and initiated the concept of the stored program, whereby program code could be stored in memory much like the data was, meaning the setup for a new calculation could be expedited without rewiring. The EDVAC computer (Electronic Discrete Variable Automatic Computer), first described in a draft report in June 1945 by von Neumann who was a regular visitor to the US Army project at the Moore School, University of Pennsylvania, was the first design to implement the stored program concept—what came to be known as the "von Neumann" architecture. EDVAC was however just an idea at that stage, and not constructed by the Moore Laboratory until 1949. Von Neumann and EDVAC's designers were keen to use purely electronic memory for stored program, not drums or relays.

From around the period of the mid to late-1940s, emerging digital computers experienced what historian Thomas Hughes describes as a "Reverse Salient". This was a result of the scarcity of capacious and reliable memory technology and held back the advance of digital computers. Several electronic innovations in response to this were adaptations from technology developed during the War.

The Ultrasonic Delay Line was developed for storing analog radar information, and was easily adapted to digital storage. The delay line worked by using a piezoelectric quartz crystal to convert an electrical signal into a sonic wave pulse, which then travelled through a liquid medium at the speed of sound in a long tube, and was then converted back to an electric signal. The difference in the speed of propagation meant that a number of bits of data could be stored. The delay line was a serial device - once a pulse had entered the tube there was no way to get at it until it emerged at the other end. All operations were considered as serial transfers of numbers

Delay lines were first built by at the Bell Telephone Labs by William Shockley, and later at the Moore School for the Radiation Laboratory at MIT in 1943. Mercury acoustic delay lines were adapted for use with computers by J. Presper Eckert and John Mauchly in 1946 and selected for the EDVAC, which was however not completed until 1952. Mercury delay lines realized a 100 times storage density saving over valve technology. The first use in digital computers was in Maurice Wilkes's EDSAC (Electronic Delay Storage Automatic Computer) completed in 1949 at the Cavendish Laboratory in Cambridge, UK, where 32 'tanks' of delay lines each stored 32 words of 18 bits (total about 2 kilobytes).

In response to the high cost and slowness of delay lines, Frederick Williams and colleagues at Manchester University in the United Kingdom developed the technique of using electrostatic cathode-ray display tubes as digital stores. The cathode ray tube was another wartime development and the persistence of the phosphor display screen once illuminated by the cathode ray provided the memory. By 1948, a storage of 1024 bits was successfully implemented. William's colleague Tom Kilburn made improvements that increased the capacity to 2048 bits. The Williams-Kilburn tubes (commonly known as Williams tubes) were used on several of the early stored program computers, including the Manchester 'Baby' (1948) and the Manchester Mark 1 which became operational in 1949, and the Institute of Advanced Study (IAS) machine spearheaded by von Neumann at Princeton, finally completed in 1951. The Williams tube memory had a big advantage over delay line memory in that it allowed fast random access (any memory location could be addressed and read directly). The Manchester Mark I was the first to store both its programs and data in RAM, as modern computers do.

In 1946 Vladimir K. Zworykin, Jan Rajchman and colleagues at the Radio Corporation of America (RCA) labs built the Selectron valve (or selective storage electrostatic tube) that could store 256 binary digits (bits) of information, and was described by Herman Goldstine (who directed the outpost of the US Army's Ballistics Research Laboratory at the Moore School from 1942) as "a work of great engineering virtuosity". Like the Williams tube, the Selectron was also a random access storage device. The Selector design however took over 3 years to reach the market, and was scaled down from the initial specification of 4096 bits storage. The revised Selectron was used in the 1953 JOHNNIAC at the Rand Corporation. The JOHNNIAC was one of the many machines based on the Princeton Institute of Advanced Studies (IAS) report authored in 1946 by John von Neumann, Arthur Burks and Goldstine, based on their work on digital computers at Princeton following their involvement at the Moore School.

Magnetic core memory was first suggested by Jay Forrester of the Servomechanisms Laboratory at MIT and Andre Booth of the University of London, and was an important step in miniaturisation and speed of computer memory. Small ferrite rings that were laid out in a two dimensional matrix, with the magnetic property of hysteresis 'remembering' an electric pulse. Each element of data was independently accessible via a row and column addressing system. Forrester and colleagues developed a successful prototype in 1949, and core arrays storing 1024 bits were used in their Whirlwind computer installed for the Navy's SAGE anti-aircraft defense system in 1958. Core memory became the dominant means of primary storage and was widely used on large commercial and scientific computers such as the IBM 701 in 1955, which was a significant commercial success and led to IBM's establishment as the dominant player in large systems. The IBM 7094 in 1963 had a core memory with over a million bits.

Ultimately it was the solid-state semiconductor that was to provide the memory technology of the computer revolution. The transistor was invented in 1947 by William Shockley at the Bell Laboratories after 8 years of research on semiconductors for radar use. It was small and had very low power requirements and the switching speed was very quick compared to valves and core memory. Although it was a functional 'switch' replacement for the valve as discussed above, it did not appear as computer memory until the late 1950s. There was significant momentum with the success of magnetic core memory in the industry, and the reliability of initial mass produced transistors was a contributing negative factor. Furthermore, Bell Labs was a regulated monopoly and limited by a court decision to the telecommunications business, although they did use transistors in special purpose computers built for the military use in ICBM's around 1952. However information about transistor technology was available to other manufacturers, with Philco successfully mass producing reliable and high performance components. These were used in the SOLO, regarded as the first general purpose transistorised computer, completed by 1958. The commercialized version was called the TRANSAC which became available in 1960. Although the IBM 7090 (first installed in 1959) had a transistorised Central Processing Unit, it still used magnetic core memory.

The development of integrated circuit (IC) technology began in 1959 with Jack Kilby of Texas Instruments. Collections of transistors were miniaturized onto a silicon 'chip'. It was only in the early 1970's that semiconductor memory was commercially used in computers, after earlier pioneering use in purpose-built military and Apollo-era space guidance computers in the 1960s. The Super Nova produced by Data General in 1971 was the first commercial computer to use integrated circuit Transistor-Transistor Logic (TTL) IC chip memory, after the less-successful research computer the Illiac-IV broke the tradition of use of magnetic cores by using a 256-bit memory chip made by Fairchild. IBM went a proprietary route by using monolithic semiconductor memory in the System/370 Model 145 in 1971. Intel entered the IC memory market by producing a1024-bit memory chip in 1970.

Memory chips were rapidly embraced by the emerging minicomputers, resulting in both compact size and more importantly, low cost. Later in the 1970s, these IC memory chips together with microprocessors provided the basis of the personal computer. With the advent of Large Scale Integration (LSI) and Very Large Scale Integration (VLSI), the cycle of increasing density of semiconductor memory fabrication continued unabated. Although the concepts and architecture of computers remained consistent with early designs, these techniques, competitive markets and insatiable demand for personal computers globally, resulted in a doubling of memory density almost every year to the end of the century where up to a billion transistor elements on memory chips were commonplace.

See also: Computers, Early Digital; Computer Memory, Personal Computers; Integrated Circuits, Design and Use; Processors for Computers; Transistors; Vacuum Tubes/Valves

Bruce Gillespie

Further Reading

Paul E. Ceruzzi, A History of Modern Computing, MIT Press, 1998

Charles and Ray Eames, A Computer Perspective, Background to the Computer Age, Harvard University Press, 1973

Herman H. Goldstine, The Computer, from Pascal to von Neuman, Princeton University Press, 1972

Andrew Hodges, Alan Turing, The Enigma, Burnett Books, 1983

Christian Wurster, Computers, An Illustrated History, Taschen, Koln, 2002

Sample Entries