Small Computers of the 1950s

Abstract

The first computers filled rooms, but by the mid-1950s manufacturers were selling “small” machines: desk-sized or refrigerator-sized tube computers with a magnetic drum for memory, priced at a fraction of a UNIVAC. The Librascope LGP-30 and the Bendix G-15 sold about four hundred to five hundred each; Burroughs, Elliott and a crowd of short-lived Californian firms built others, and IBM’s 650, counted as small at the time, sold some two thousand. Two conferences, in Washington in 1952 and Philadelphia in 1954, defined the category. Several of these machines survive, and one, the Harwell Dekatron, still works. The transistor minicomputers of the 1960s grew out of them (see DEC and the Minicomputer Era).

What Counted as Small

On 14 May 1952 the Office of Naval Research, whose mathematics and computing work was led by Mina Rees, held a symposium at the Pentagon on “commercially available general-purpose electronic computers of moderate price,” meaning under $100,000. Seven firms presented machines: Jacobs Instrument’s JAINCOMP-B1, a refrigerator-sized parallel machine with core memory; Monroe’s MONROBOT III, with 800 tubes, a drum and twenty-digit decimal numbers; the CADAC of the Computer Research Corporation in Hawthorne, California, 1.4 metres high with 195 tubes; the Circle computer from New York, built on the architecture of von Neumann’s IAS machine; the three-metre-wide Elecom 100; the planned ElectroData 30-201, driven by Clifford Berry, Atanasoff’s former partner; and the MINIAC of the Physical Research Laboratories in Pasadena. A MONROBOT called the 1952 presidential election for NBC (see UNIVAC and the Eckert-Mauchly Computer Corporation); most of the seven sold only a handful. NCR bought the Computer Research Corporation in 1953, Underwood bought the Electronic Computer Corporation, and Marchant took over the Pasadena laboratory and built only two MINIACs.

From 8 to 10 December 1954 the Eastern Joint Computer Conference in Philadelphia, organised by three engineering societies, took small digital computers as its theme. Alan Perlis defined the class there: a price under $150,000, a memory of at least a thousand words, and access to a library of programs. Other definitions at the conference set limits of twenty kilowatts of power. The speakers covered Axel Wenner-Gren’s ALWAC, NCR’s CRC 102, the MINIAC, Bell Labs’ transistorised TRADIC, the Burroughs E101 and the IBM 650, whose first delivery to a customer happened to fall on the conference’s opening day.

The Burroughs E101

The Burroughs E101, presented at Philadelphia by the mathematician Alex Orden, who had helped build the SEAC, was the size of a desk. It borrowed the keyboard and printer of Burroughs’ Sensimatic accounting machine and added a drum with a hundred storage locations, 163 tubes, 1,500 diodes and twenty relays. It was programmed with pins: each perforated board held sixteen rows, one instruction per row, and pins pushed into its three fields selected the operation and the registers or memory addresses; up to eight boards could be chained. An addition took a twentieth of a second, a multiplication a quarter of a second. It cost $32,500, or $850 a month to rent, went on sale in 1955 and did not sell well. The Catholic University of Louvain bought one in 1958 for Georges Lemaître, the physicist-priest who had proposed the expanding universe and the “primeval atom” that became the Big Bang; it was the university’s first computer and survives in Leuven. Burroughs entered the market properly in 1956 by buying ElectroData and its Datatron, which became the Burroughs 205.

The LGP-30

The physicist Stanley Frankel, born in Los Angeles in 1919, had run the hand-calculation group at Los Alamos during the war and was among the first users of ENIAC, computing for Edward Teller’s hydrogen-bomb studies. In 1949 he lost his security clearance; at Caltech he built a simple machine called MINAC, the Minimal Automatic Computer, in 1954. Its commercial version, the Librascope LGP-30, came out in September 1956. It had 113 tubes and 1,350 germanium diodes in a cabinet of 112 by 66 by 84 centimetres, about the size of a chest freezer, and weighed 365 kilograms; a drum turning at 3,600 revolutions a minute held 4,096 words of 32 bits (31 of them usable, see The Story of Mel). It had sixteen instructions, added in a quarter of a millisecond and multiplied in fifteen, and used a Friden typewriter with paper-tape reader and punch for input and output. It cost $47,000.

For many American colleges the LGP-30 was the first computer. Dartmouth College bought one in 1959 and let undergraduates use it; John Kemeny and Thomas Kurtz, who ran it, later created BASIC. Librascope sold about five hundred, followed in 1963 by the transistorised LGP-21. Frankel went on to design the electronics of the Combitron desktop calculators for Diehl in Nuremberg, and died in Los Angeles in 1978.

From 1958 the Minden firm Schoppe & Faeser, founded on 17 June 1948 by Hermann Schoppe and Hugo Faeser and until then a maker of large analog computers (see Analog Computers), built the LGP-30 under licence and sold it through Eurocomp; the HNF puts the run at fifty, the first computers built in North Rhine-Westphalia. Schoppe & Faeser followed with the transistorised RPC 4000 and LGP-21, about two hundred computers in all, before turning to industrial automation.

The Bendix G-15

Bendix Aviation of Los Angeles bought Northrop’s computer department in 1952 and in the mid-1950s brought out its own machine, the G-15, designed by Harry Huskey, who had worked on ENIAC and on the Pilot ACE in Britain and built the SWAC in Los Angeles (see Alan Turing and the ACE). Huskey and David Evans, later co-founder of Evans & Sutherland, described it at the Wescon trade fair in August 1954 as “small in physical size”: 1.83 metres high on a footprint of 82 by 91 centimetres. It had 450 tubes and 3,000 germanium diodes, and its drum held 2,160 words of 29 bits, organised to imitate the delay lines of the Pilot ACE, whose serial design it borrowed. A converted IBM typewriter served as terminal. In 1958 the basic unit cost $49,500 or $1,485 a month. About four hundred were built before Bendix sold its computer division to Control Data in 1963; one at the System Source Museum near Baltimore is described as the oldest working computer in the United States.

The IBM 650

IBM announced the IBM 650 in July 1953 and delivered the first in December 1954. With about two thousand built it was, if counted as a small computer as it was at Philadelphia, by far the most successful of the class. On 20 January 1956 Allianz presented one at its Munich headquarters, the first series-produced computer in West Germany, ordered on 6 May 1955 after the insurer’s head, Hans Goudefroy, had taken advice from the Darmstadt pioneer Alwin Walther; it ran until October 1961. Five more 650s were running in West Germany by the end of 1956, at the federal salaried employees’ insurance in Berlin, IBM in Sindelfingen, Volkswagen, the Deutsche Edelstahlwerke in Krefeld, and at a hotel in Bad Neuenahr where one computed for the Bundeswehr, and IBM’s German subsidiary began building them that year (see IBM: The Company).

British Machines

The instrument makers Elliott Brothers, founded in London early in the nineteenth century and builders of analog fire-control computers for the Royal Navy since the 1910s, set up a research laboratory at Borehamwood in 1946. Its Elliott 401, finished in March 1953 with state support, was 4.19 metres long, weighed over a tonne and held 500 tubes, nickel-wire delay lines and a magnetic disc, later replaced by a drum; at the time this counted as small. Shown at the Physical Society’s exhibition at Imperial College in April 1953, it was Britain’s third computer from a private maker after the Ferranti Mark 1 and LEO. It worked at the Rothamsted agricultural research station from 1954 and has been on display at the Science Museum since 2016. Ten of the similar Elliott 402 followed, one of which replaced the Zuse Z5 at Leitz in Wetzlar in 1958, and 33 of the larger 405, sold with NCR; in 1959 an Elliott 402F worked in the BBC’s election studio.

At the Atomic Energy Research Establishment at Harwell, a computer of 138 tubes, hundreds of relays and nine stores of dekatron counting tubes (each module holding ten eight-digit decimal numbers) ran its first operations in April 1951. The Harwell Dekatron was no faster than a person at a desk calculator, but it could run for days unattended. Replaced in 1956, it went to the college in Wolverhampton, where as WITCH (Wolverhampton Instrument for Teaching Computing from Harwell) it ran from December 1957 to March 1973. Rediscovered in storage in 2009 by Kevin Murrell, it was restored at the National Museum of Computing at Bletchley Park by November 2012 and entered the Guinness Book of Records in 2013 as the oldest working electronic digital computer.