5.2 John von Neumann and the “Report on the EDVAC”

J. Robert Oppenheimer (left) and John von Neumann at the October 1952 dedication of the computer built for the Institute for Advanced Study. Oppenheimer, who was head of the Los Alamos Laboratory during the war, became the institute's director in 1947.

In the summer of 1944, a few weeks before the EDVAC contract was let, a powerful new figure stepped into the history of computers – a figure who would exercise enormous influence on both the development of computers and on the historical record of that development. John von Neumann (Noy-man) was one of the world’s most talented and famous mathematicians, a dazzlingly productive and original intellect. He was a consultant to Los Alamos and the BRL and a professor at the Institute for Advanced Study in Princeton, New Jersey. Despite his connections with the BRL, he learned about ENIAC and EDVAC by accident, not through channels. One day in August, Goldstine was at the Aberdeen train station waiting for the train back to Philadelphia…

when along came von Neumann. Prior to that time I had never met this great mathematician, but I knew much about him of course and had heard him lecture on several occasions. It was therefore with considerable temerity that I approached this world-famous figure, introduced myself, and started talking. Fortunately for me von Neumann was a warm, friendly person who did his best to make people feel relaxed in his presence. The conversation soon turned to my work. When it became clear to von Neumann that I was concerned with the development of an electronic computer capable of 333 multiplications per second, the whole atmosphere of our conversation changed from one of relaxed good humor to one more like the oral examination for a doctor’s degree in mathematics.

Soon thereafter the two of us went to Philadelphia so that von Neumann could see the ENIAC. At this period the two accumulator tests were well underway. I recall with amusement Eckert’s reaction to the impending visit. He said that he could tell whether von Neumann was really a genius by his first question. If this was about the logical structure of the machine, he would believe in von Neumann, otherwise not. Of course this was von Neumann’s first query.

Von Neumann arrived on 7 September 1944. According to Mauchly, writing in 1979, this is what he and Eckert told him during their briefing:

We started with our basic ideas: there would be only one storage device (with addressable locations) for the entire EDVAC, and this would hold both data and instructions. All necessary arithmetic operations would be performed in just one arithmetic unit (unlike the ENIAC). Of course, there would be devices to handle input and output, and these would be subject to the control module just as the other modules were.

Von Neumann was fascinated by the Moore School’s work. “Like a child with a new toy,” wrote Mauchly, “he could not put it aside.” Flattered by his interest, the school invited von Neumann to join the staff as a consultant. Although he had arrived too late to influence ENIAC’s design – his chief contribution to ENIAC was the suggestion that it be used to solve the implosion problems of Los Alamos – von Neumann came in time to participate in the development of EDVAC.

As a highly respected scientist, von Neumann lent a badly needed air of legitimacy to the school’s pioneering endeavors. Many scientists – in particular, Bush at MIT, Aiken at Harvard, and Stibitz at Bell Labs – regarded ENIAC as a foolish endeavor, bound for failure, and a waste of government funds that would have been better spent on the proven technologies of relay calculators and differential analyzers. There was a lot of bad blood between the University of Pennsylvania and its detractors. (After visiting MIT, which built a two-thousand-tube electronic differential analyzer in 1941, Goldstine wrote: “It was, I think, a pretty sad spectacle of what the supermen at NDRC can do.” NDRC, or the National Defense Research Committee, was a major source of government research funds.) But if von Neumann thought the Moore School’s work was good enough to merit his attention, then perhaps there was something to it after all; at least the chiefs of the Ordnance Department probably slept better knowing that he was involved with EDVAC.

For von Neumann was one of the most respected scientists of his time. Born in Budapest, Hungary, in 1903, von Neumann had a genius’s ability to perform complicated calculations in his head. At eighteen, he published his first mathematical paper; at twenty-two, he earned his doctorate in mathematics from the University of Budapest; at twenty-four, he became a privatdozant (“lecturer”) at the University of Berlin, a rare honor for one so young. By that time he had published several papers on algebra, set theory, and quantum mechanics, the first installments of a creative output that filled six volumes by the time of his death in 1954. His most influential mathematical achievement was the invention of the theory of games; in a paper published in 1928, he showed how to find the best line of play, the one guaranteeing the smallest losses, in any game of strategy. Since a “game” in mathematics is more than a pair of dice and a brightly colored playing board, game theory has many important applications in economics, military strategy, and the social sciences.

In 1930, von Neumann immigrated to the United States, where the opportunities for academic advancement were greater than in Central Europe. (A little study of the odds of his becoming a professor in Europe as opposed to America had sent him packing.) He became a visiting lecturer at Princeton University and then, when the mighty Institute for Advanced Study (IAS) was established at Princeton in 1933, he received a permanent professorship on the IAS faculty. (Einstein also joined the institute that year.) A dapper, worldly, sophisticated man, Johnny, as his friends called him, was fluent in four languages and spoke English without an accent. He possessed an Old-World courteousness and a racy, Americanized sense of humor; he knew an endless string of jokes and anecdotes and his friends and colleagues delighted in telling him the latest ones. He had a substantial income – $10,000 a year from the IAS plus a large inheritance from his father, a successful banker – and moved easily in the highest academic and government circles, where he was a much sought-after consultant.

His keen interest in ENIAC and EDVAC was more than a mathematician’s natural curiosity in calculating machines. As a consultant to Los Alamos, he played a central role in the development of the atomic bomb; he, Edward Teller, and other scientists on the Manhattan Project devised the all-important implosive lens of the first bombs. (Using conventional explosives, the lens generated a powerful spherical shock wave that imploded, or compressed, a ball of plutonium or uranium isotope to an atomically critical point, thus setting off the chain reaction.) He also showed the scientists at Los Alamos how to model an implosion mathematically and how to solve the resulting equations numerically, with the help of IBM card punchers and sorters. Los Alamos set up one of the largest punch card installations in the world, but the going was slow and von Neumann was on the lookout for faster computational methods when he met Goldstine at Aberdeen in August.

Eight months earlier, in January 1944, von Neumann had written the Office of Scientific Research and Development (OSRD), a governmental clearing house for scientific research, for information on the country’s computational resources. OSRD was headed by Vannevar Bush, who knew all about ENIAC, but the agency held to the conventional wisdom about the ineluctable fallibility of tubes and didn’t have much faith in the Moore School’s effort. As a result, OSRD didn’t tell von Neumann about ENIAC and instead referred him to Aiken at Harvard, Stibitz at Bell Labs, and Wallace Eckert – no relation to the Moore School ‘s Eckertat the IBM computing center at Columbia. But the Columbia operation relied on punch card tabulators, Aiken’s machine was still under construction, and Stibitz’s Complex Number Computer could only process imaginary numbers and was unprogrammable.

Some of the EDVAC technicians gather before a completed section of the machine.

At the Moore School, von Neumann helped Eckert and Mauchly and the other EDVAC engineers refine their ideas. He was particularly influential on the subject of the machine’s internal logic; that is, its organization from the point of view of the efficient processing of information. In a team effort where ideas are batted about in informal discussion, it is difficult if not impossible to pinpoint the originator of one or another notion, and von Neumann certainly contributed many good ideas. But there is no question that Eckert and Mauchly had conceived of the all-important stored program long before von Neumann joined the effort. However – and this is a very important point – Eckert and Mauchly had not gotten around to outlining a design for a stored program computer when von Neumann appeared.

In the spring of 1945, von Neumann offered to write an analysis of EDVAC’s logical design, summarizing the staff’s thinking and expanding and developing it according to his own lights. Eckert, Mauchly, and the rest of the EDVAC staff agreed to the idea – in fact, they welcomed it. Working mostly at Los Alamos, where the first bombs were being readied for Japan and the demand on his time was less pressing than it used to be, von Neumann roughed out a 101-page manuscript on EDVAC and mailed it to Goldstine in June. It was a preliminary report, containing numerous blank spaces for names, cross-references, and other information that von Neumann intended to insert after his colleagues on the EDVAC project had read the paper and commented on it. The final draft would give credit where credit was due, identifying the originators of the more important ideas.

Commercially available digital test equipment didn't exist in the late 1940s, so the EDVAC team had to make their own. The gadget shown here is a word generator, which produced a continuous stream of zeros and ones; among other things, it enabled engineers to verify the reliability of data transmission lines.

Although the report was grounded in the work of others – notably Eckert and Mauchly – it was von Neumann’s through and through. Not surprisingly, it was a lucid and masterful analysis of the structure and operation of a computer, full of interesting ideas and written with an overarching concern for “logical” control. In brief, von Neumann recommended the construction of a computer based on a central control unit that would orchestrate all operations; a central processor unit that would carry out all arithmetical and logical operations; and a random-access read/write memory (this is a contemporary term) that would store programs and data in such a way that any piece of information could be entered or retrieved directly (rather than sequentially). He also recommended the use of binary math and Boolean algebra and, furthermore, the processing of all binary words in series rather than in parallel. In other words, instead of operating on every bit in a word at the same time, as ENIAC did, EDVAC would process every bit one at a time. All things being equal, a parallel computer is faster than a serial one, but it is more difficult to build – thus the reason for von Neumann’s suggestion.

Without Eckert and Mauchly’s knowledge, Goldstine put a cover on von Neumann’s report, listed him and him alone as the author, and distributed it under the title “First Draft of a Report on the EDVAC.” Thirty-two people in and out of the Moore School were on the original mailing list and many others received copies later on. “Report on the EDVAC” was not only the first paper on the design of a general-purpose digital electronic computer, it was also a work from the hand and mind of the great von Neumann, and it had a strong impact on everyone who read it. He, not Eckert and Mauchly, was regarded as the inventor of the stored-program computer. Goldstine’s actions were presumptuous but understandable, since he had chiefly wished to enhance the Moore School’s reputation and let other people in on the latest thinking on computers; it probably never occurred to him that the premature distribution of “Report on the EDVAC” would spawn years of misunderstanding and jeopardize Eckert and Mauchly’s patent rights to EDVAC.

As it turned out, EDVAC wasn’t completed until 1952. By that time, the distinction of developing the first stored-program computer had fallen to another country – Babbage’s homeland. Seeking to stimulate the development of computers, the Army Ordnance Department and the Office of Naval Research sponsored a summer course on computers at the Moore School in 1946. Twenty institutions – chiefly American companies, universities, and government agencies – sent representatives. Alone among the Allies, Great Britain was invited to participate, and the information it picked up there, coupled with its substantial engineering and theoretical knowledge, enabled it to jump ahead. So before we examine the first American computing projects – in particular, Eckert and Mauchly’s private venture and von Neumann’s effort at the Institute for Advanced Study – we’ll take a look at the valiant computing developments in postwar Britain.

Back                                                                                                                                   Continue