ENIAC

• 1. Introduction

  ›

  ENIAC, the Electronic Numerical Integrator and Computer, was the first programmable, electronic, general-purpose digital computer ever built. When the press got their first look at it in early 1946, they called it a "Giant Brain." That phrase stuck. Here was a machine that occupied 1,800 square feet, weighed more than 30 tons, and consumed 150 kilowatts of electricity. It had roughly 18,000 vacuum tubes, and several of those tubes burned out almost every single day.

  The machine had been conceived during World War II to solve a pressing military problem: how to calculate artillery firing tables fast enough to be useful on the battlefield. By the time it was completed in 1945, its first real program was not about artillery at all. It was a study of whether a thermonuclear weapon was even feasible.

  Behind those towering panels of vacuum tubes and wire was a team of mathematicians whose contributions would be overlooked for decades. Six women did the bulk of the programming. They were classified as subprofessionals, paid less than the men around them, and were not even invited to the formal dedication ceremony. Their story, and the machine's, raises questions that still echo through computing today: who builds a technology, who gets the credit, and what does a machine do to the world once it is turned loose?

• 2. A Wartime Calculation

  ›

  John Mauchly was a physics professor at Ursinus College when the problem first caught his attention. In June 1941, the United States Army Ordnance Department was relying on Friden calculators and mechanical differential analyzers to compute firing tables for artillery. Graduate students did the work under Mauchly's supervision. Mauchly began to wonder whether electronics could do the same calculations far more quickly.

  He was not an electronics expert, so he partnered with J. Presper Eckert of the University of Pennsylvania to draft a proposal. In August 1942, Mauchly formally proposed an all-electronic calculating machine to the Army. The Army Ordnance Corps accepted the plan and handed the University of Pennsylvania an initial six-month research contract worth $61,700. The construction contract was signed on the 5th of June, 1943.

  Work began in secret the following month at Penn's Moore School of Electrical Engineering under the code name Project PX. John Grist Brainerd served as principal investigator. Herman Goldstine, who had persuaded the Army to fund the project, was placed in charge of overseeing it on the Army's behalf. Assembly of the actual hardware did not begin until June 1944. By September of that year, Eckert and Mauchly had completed their conceptual design. Construction finished in May 1945, and testing began at the Moore School.

  The project was financed by the Army Ordnance Corps Research and Development Command, led by Major General Gladeon M. Barnes. Final costs came in at roughly $487,000, far beyond the original $61,700 contract. That first program, the thermonuclear feasibility study, ran in December 1945 using equations to calculate nuclear reactions. The data supported research into building a hydrogen bomb.

• 3. The Machine Itself

  ›

  Physically, ENIAC was unlike anything the public had seen. It stood 10 feet tall, stretched 100 feet long, and sat 3 feet deep. Its guts included 18,000 vacuum tubes, 7,200 crystal diodes, 6,000 relays, 70,000 resistors, 10,000 capacitors, and approximately five million hand-soldered joints.

  Twenty of its modules were accumulators, each capable of holding a ten-digit decimal number in memory and performing both addition and subtraction. Numbers moved between units across several general-purpose buses, which engineers at the time called trays. To achieve its remarkable speed, the panels had to send numbers, compute, save results, and trigger the next operation, all without any moving parts.

  Speed was the machine's headline feature. ENIAC ran at roughly one thousand times faster than the electro-mechanical machines it was meant to replace. In practical terms, it could perform 5,000 addition or subtraction operations per second. Multiplication of two ten-digit numbers took 14 machine cycles, or 2,800 microseconds, for a rate of 357 per second. Division and square roots were slower, at up to 143 cycles each.

  Input came from an IBM card reader. Output went to an IBM card punch, and those cards could then feed into an IBM accounting machine such as the IBM 405 for printed results. ENIAC originally had no built-in memory storage; punch cards served that function externally. In 1953, a 100-word magnetic-core memory built by the Burroughs Corporation was added to the system.

  The reliability problem was constant. Several tubes burned out almost every day, leaving the machine nonfunctional roughly half the time. High-reliability tubes were not available until 1948. Engineers focused their attention on the warm-up and cool-down periods, when tube heaters and cathodes faced the greatest thermal stress. By managing those periods carefully, they reduced failures to about one tube every two days. In a 1989 interview, Eckert recalled: "We had a tube fail about every two days and we could locate the problem within 15 minutes." The longest uninterrupted run on record lasted 116 hours, reached in 1954.

• 4. The Women Who Programmed It

  ›

  At least 200 women were hired by the Moore School of Engineering during World War II to work as human computers, producing the numeric results of mathematical formulas by hand with mechanical calculators. Herman Goldstine selected six of them to become the programmers of ENIAC: Betty Holberton (then Betty Snyder), Kay McNulty, Marlyn Wescoff, Ruth Lichterman, Betty Jean Jennings, and Fran Bilas.

  These women held professional degrees in mathematics. Yet while men with equivalent education and experience were classified as professionals, these six were classified as subprofessionals. The National Advisory Committee for Aeronautics stated in 1942 that engineers admitted the women computers did the work "more rapidly and accurately" than the engineers themselves would. The justification given for bringing women into computing was that it freed male engineers from repetitive calculation, not that the women were especially capable, even though they were.

  Programming ENIAC was not a clerical job. It required understanding the machine's logic, physical layout, circuitry, and operation from the ground up, because no programming languages existed yet. Under the direction of Herman and Adele Goldstine, the women studied ENIAC's blueprints and physical structure to figure out how to manipulate its switches and cables. The programmers were often able to narrow bugs down to an individual failed vacuum tube, which they could then point to for replacement by a technician.

  None of the six women were invited to ENIAC's formal dedication on the 15th of February, 1946, nor to the celebratory dinner that followed. Elizabeth Snyder and Betty Jean Jennings had developed the demonstration trajectory program shown to the press that evening, a demo that calculated a missile's path in 15 seconds, a task that would have taken a human computer several weeks. Herman and Adele Goldstine took credit for it.

  Decades passed before the women's contributions gained wide recognition. In the 1990s, researcher Kathryn Kleiman discovered that most of the ENIAC programmers had not been invited to the machine's 50th anniversary event. She set out to find them and record their oral histories. "They were shocked to be discovered," Kleiman later said. "They were thrilled to be recognized, but had mixed impressions about how they felt about being ignored for so long." In 1997, the six women were inducted into the Women in Technology International Hall of Fame. Kleiman published a book on them in 2022 through Grand Central Publishing, titled Proving Ground. Three Army supercomputers, as of 2020, are named Jean, Kay, and Betty, after Jean Bartik, Kay McNulty, and Betty Snyder respectively.

• 5. Programming and Its Limits

  ›

  Getting a problem into ENIAC was its own ordeal. The machine had no stored-program capability in its original form. Instead, programs were set up through a combination of plugboard wiring and three portable function tables, each containing 1,200 ten-way switches. Mapping a problem onto the machine could take weeks. Once the program was worked out on paper, physically configuring the switches and cables could take days more. Then came verification and debugging, which the machine supported by allowing step-by-step execution.

  Programs were changed only after massive numbers of tests of the current configuration, because the cost of reprogramming was so high. This constraint shaped what ENIAC was used for. After the hydrogen bomb feasibility study, ENIAC's versatility was demonstrated in other domains, notably in helping establish the Monte Carlo method as a tool in scientific computation. John von Neumann and Stanislaw Ulam had realized that ENIAC's speed could replicate calculations that previously required large groups of human computers modeling neutron travel through materials. The success of that project showed the broader value of the Monte Carlo technique.

  In April 1948, ENIAC was modified to support a primitive stored-program approach, using the function tables as a read-only program memory. The modification was developed by Richard Clippinger with input from von Neumann, who proposed a one-address architecture because it was simpler to implement. Betty Jennings, Clippinger, and Adele Goldstine carried out the programming of the stored-program system. When first demonstrated in April 1948, it ran a program written by Adele Goldstine for John von Neumann.

  The modification cut ENIAC's speed by a factor of six and eliminated its parallel computation capability. Engineers accepted this trade-off because analysis showed that almost any real-world problem was already input/output bound, meaning the machine spent most of its time waiting on card readers and punches rather than computing. Reprogramming time shrank from days to hours, which mattered far more in practice. A high-speed shifter added in early 1952 improved shifting speed by a factor of five, and a 100-word core memory expansion was installed in July 1953.

• 6. Legacy and Legal Battle

  ›

  ENIAC was formally dedicated at the University of Pennsylvania on the 15th of February, 1946, and formally accepted by the U.S. Army Ordnance Corps in July 1946. It was shut down on the 9th of November, 1946, for refurbishment and a memory upgrade, then transferred to Aberdeen Proving Ground in Aberdeen, Maryland in 1947. It ran in continuous operation there until 11:45 p.m. on the 2nd of October, 1955, when it was retired in favor of the EDVAC and ORDVAC computers.

  The patent for ENIAC was applied for in 1947 and granted in 1964. It did not survive legal challenge. In the 1973 federal court case Honeywell, Inc. v. Sperry Rand Corp., a judge voided the patent. The decision found that the ENIAC inventors had derived their electronic digital computer concepts from the work of John Atanasoff, who had prototyped the Atanasoff-Berry Computer with Clifford Berry in 1939. Mauchly had examined that machine in June 1941. The ruling placed the invention of the electronic digital computer in the public domain and gave legal recognition to Atanasoff as the first inventor.

  The machine's influence on what came after was direct and visible. In the summer of 1946, the Pentagon invited leading scientists from the United States and Great Britain to a series of 48 lectures in Philadelphia covering the design of digital computers, known as the Moore School Lectures. Half of those lectures were delivered by the inventors of ENIAC. John von Neumann's notes from the related EDVAC meetings, the First Draft of a Report on the EDVAC, were distributed by Herman Goldstine to government and academic institutions, sparking interest in stored-program machines. That document influenced the development of the EDSAC at Cambridge University in England and the SEAC at the U.S. Bureau of Standards.

  Pieces of ENIAC ended up scattered across institutions after its retirement. The Smithsonian holds several panels in the National Museum of American History in Washington, D.C. The Computer History Museum in Mountain View, California displays three panels and a portable function table on loan from the Smithsonian. In 1996, to mark the 50th anniversary of ENIAC's unveiling, the University of Pennsylvania built ENIAC-on-a-Chip, a silicon chip measuring 7.44 millimeters by 5.29 millimeters that replicated the machine's functionality. ENIAC was named an IEEE Milestone in 1987, and in 2011 the city of Philadelphia declared the 15th of February as ENIAC Day.