ENIAC (1945): The First General-Purpose Electronic Computer

ENIAC, the Electronic Numerical Integrator and Computer, was the first programmable, general-purpose electronic digital computer. Built at the University of Pennsylvania in Philadelphia for the US Army during the Second World War, it was completed in late 1945 and formally unveiled on 15 February 1946. By showing that thousands of vacuum tubes could work together reliably at high speed, ENIAC opened the way to the modern computer industry.

Background: the firing-table problem

Every new artillery gun and shell needed a firing table, a printed guide telling gunners how to aim under different conditions of range, elevation, wind and air density. Each table required thousands of trajectory calculations. During the Second World War this work fell to the Army Ordnance Department’s Ballistic Research Laboratory (BRL) at Aberdeen Proving Ground in Maryland.

The BRL relied on a mechanical differential analyser and on teams of “computers”, the job title then used for people, mostly women with mathematics degrees, who calculated trajectories with desk calculators. One trajectory could take a skilled human computer a day or more. By 1942 the work was falling badly behind as new weapons poured out of American factories.

Part of the computing effort was based at the Moore School of Electrical Engineering at the University of Pennsylvania. There, the physicist John W. Mauchly had become interested in using vacuum tubes, the electronic switches of radios and radar sets, for high-speed calculation. In August 1942 he wrote a short memo, “The Use of High Speed Vacuum Tube Devices for Calculating”, proposing an electronic machine.

Building the machine

The memo reached Lieutenant Herman Goldstine, a mathematician serving as the Army’s liaison at the Moore School. Goldstine championed the idea, and on 5 June 1943 the Army signed a contract for what was code-named Project PX. The initial budget was about $61,700; by the time the machine was finished the cost had risen to roughly $487,000.

Mauchly provided much of the overall concept, while the young engineer J. Presper Eckert, then only 24, became chief engineer. Eckert’s great contribution was reliability. Vacuum tubes were notoriously prone to failure, and many experts doubted that a machine with thousands of them could run long enough to be useful. Eckert ran the tubes well below their rated power, tested components carefully and designed circuits with generous safety margins.

A team of around 50 engineers, technicians and assembly workers built the machine between 1943 and 1945. Key engineers included Arthur Burks, Kite Sharpless and Harry Huskey, among others. Workers hand-soldered about five million joints.

The finished ENIAC was enormous. It consisted of about 40 panels arranged in a U shape around a room, with roughly 17,500 vacuum tubes, 7,200 crystal diodes, 1,500 relays, 70,000 resistors and 10,000 capacitors. It weighed about 27 tonnes and drew some 150 kilowatts of power, much of it turned into heat that had to be removed by fans.

How ENIAC worked

ENIAC worked in decimal rather than binary. Its 20 “accumulators” could each store a ten-digit number and add or subtract it at electronic speed. Separate units handled multiplication, division and square roots. Three portable “function tables” held sets of numbers entered by hand with rotary switches, and IBM card readers and punches provided input and output.

The machine could perform about 5,000 additions per second, roughly a thousand times faster than electromechanical calculators. A trajectory that took a human computer a day or more could be calculated in about 30 seconds, faster than a real shell took to reach its target.

ENIAC had no stored program in the modern sense. To set it up for a new problem, operators connected the units with cables on plugboards and set thousands of switches. Each unit could be told what to do and when, and the machine could branch, choosing different steps depending on an intermediate result. This made it general-purpose: in principle it could tackle any calculation, but reconfiguring it for a new task could take days.

The ENIAC programmers

In 1945 six women were selected from the BRL’s human computers to become ENIAC’s first operators: Kathleen McNulty, Frances Bilas, Betty Jean Jennings, Ruth Lichterman, Elizabeth (Betty) Snyder and Marlyn Wescoff. No manuals or programming courses existed. They learned the machine from its block diagrams and wiring charts, and worked out how to break mathematical problems into sequences of operations that the hardware could carry out.

Their work amounted to what we would now call programming and debugging. Jennings and Snyder prepared the ballistic trajectory program shown at the public demonstration in February 1946. Snyder, later known as Betty Holberton, went on to contribute to early programming tools, and Jennings, later Jean Bartik, worked on the design of later computers.

For decades their role was overlooked. Press photographs of the 1946 launch often showed the women beside the machine without naming them, and some viewers assumed they were models. Historians and the programmers themselves restored the story from the 1980s onwards, and in 1997 all six were inducted into the Women in Technology International Hall of Fame.

From the 1946 unveiling to Aberdeen

The war ended before ENIAC was finished, so it never calculated wartime firing tables. Its first major job, run from December 1945, was a set of calculations for Los Alamos scientists, including Nicholas Metropolis and Stanley Frankel, who were studying the feasibility of a hydrogen bomb. The problem was so large that it used a vast number of punched cards.

The machine was revealed to the press on 14 February 1946 and formally dedicated at the University of Pennsylvania on 15 February. Newspapers hailed an “electronic brain”. Demonstrations included adding 5,000 numbers in a second and computing a shell trajectory faster than the shell’s real flight.

ENIAC was switched off on 9 November 1946, dismantled and moved to Aberdeen Proving Ground, where it was restarted on 29 July 1947. It remained in service there for eight more years, working on ballistics, weather prediction, atomic energy, cosmic-ray studies, wind-tunnel design and random-number studies.

In 1948, following ideas developed with the mathematician John von Neumann, ENIAC was reconfigured so that it could run programs held as coded instructions in its function tables rather than by rewiring. Adele Goldstine and Klara von Neumann were among those who contributed to this work. The change made the machine much quicker to set up and is seen by many historians as an early form of stored-program operation. Among its notable runs were Monte Carlo simulations for nuclear physics in 1948 and the first computerised numerical weather forecasts, carried out in 1950 by a team including Jule Charney.

ENIAC was finally switched off on 2 October 1955, by which time faster and more compact machines had overtaken it.

Who was first? Disputes and rivals

ENIAC’s status has been argued over for decades. In Britain, the Colossus machines designed by Tommy Flowers were operating at Bletchley Park from early 1944, helping to break German teleprinter ciphers. They were electronic but built for a specific code-breaking task, and their existence remained secret until the 1970s. In Germany, Konrad Zuse’s Z3 of 1941 was programmable but used electromechanical relays.

In the United States, John Vincent Atanasoff and Clifford Berry built the Atanasoff-Berry Computer at Iowa State College between 1939 and 1942, an electronic machine designed to solve systems of linear equations. Mauchly had visited Atanasoff in 1941. In 1973, in the case Honeywell v. Sperry Rand, US District Judge Earl Larson ruled the ENIAC patent invalid, finding among other things that Mauchly had derived ideas from Atanasoff. The ruling remains debated by historians, but it put the key computing ideas into the public domain.

Myths and facts

• Myth: ENIAC dimmed the lights of Philadelphia when it was switched on. There is no reliable evidence for this popular story; the machine drew a lot of power, but not enough to affect a city’s supply.
• Myth: ENIAC’s tubes failed constantly, so it hardly worked. Tubes did fail, especially early on, but Eckert’s conservative design and later improvements meant that the machine ran productively for years. Most failures happened when it was switched on or off, so it was often left running.
• Myth: The women in the photographs were only demonstrators. The six original programmers did essential technical work that made the machine usable.

ENIAC by the numbers

Legacy

ENIAC’s most important legacy was proof that electronic computing worked. Even before it was finished, its builders and von Neumann had begun planning a successor, EDVAC, and von Neumann’s “First Draft of a Report on the EDVAC” (1945) set out the stored-program design that most later computers followed. In the summer of 1946 the Moore School Lectures spread these ideas to engineers from Britain and across the United States.

Eckert and Mauchly left the university in 1946 to found their own company, which built the UNIVAC I, delivered in 1951 as one of the first commercial computers in the United States. Their work helped create an industry that later moved from vacuum tubes to the transistor and eventually to the personal computer. The theoretical foundations had been laid earlier by thinkers such as Alan Turing in his 1936 paper on computable numbers.

Timeline

• 1942, August: John Mauchly writes his memo proposing an electronic calculating machine.
• 1943, 5 June: The US Army signs the contract for Project PX at the Moore School.
• 1945, December: ENIAC runs its first major problem, calculations for Los Alamos.
• 1946, 14–15 February: ENIAC is demonstrated to the press and formally dedicated.
• 1946, 9 November: The machine is switched off for its move to Aberdeen Proving Ground.
• 1947, 29 July: ENIAC restarts at the Ballistic Research Laboratory.
• 1948, April: First runs with the new coded-instruction setup.
• 1955, 2 October: ENIAC is switched off for the last time.
• 1973, 19 October: A US federal court invalidates the ENIAC patent in Honeywell v. Sperry Rand.

Why it matters

ENIAC turned electronic computing from a doubtful idea into a working reality. Its engineers showed that large electronic machines could be reliable, its programmers invented the practice of programming almost from scratch, and the ideas it inspired shaped every computer that followed.