Home > V5I2 > Commemorating the Platinum Jubilee of ENIAC

## Introduction

On February 15, 1946, ENIAC (Electronic Numerical Integrator and Computer) was formally unveiled at the Moore School of Electrical Engineering of the University of Pennsylvania at Philadelphia in the United States of America. This year marks its 75th anniversary. ENIAC was a landmark in the history of computing as it was the first electronic computer that worked reliably. It was used at first to tabulate artillery firing tables and later a refurbished model was used to solve several important problems including Monte Carlo simulation, weather prediction, and fluid dynamics problems. All earlier computers such as Complex Number computer built by George Stibitz at Bell Telephone Laboratories in 1940, Z3 by Konrad Zuse in 1941, and Harvard Mark1 built by Howard Aiken in 1944 used electromechanical relays. Atanasoff and Berry built an electronic computer in 1942. It was, however, not general- purpose and was incomplete. In this article, I will recount the history of the design and development of ENIAC.

## Genesis of ENIAC

The United States of America did not participate in the second world war when it began in 1939. It was, however, helping the United Kingdom and its allies with armaments and logistics. It entered the war when Japan bombed Pearl Harbor in 1941. One of the important wartime efforts of the US army was to calculate artillery firing tables using trajectories of shells and bombs. There were thousands of artillery guns and the task was arduous. Tables for artillery were prepared at the US army’s Ballistics Research Laboratory (BRL) at Aberdeen in Maryland. A mechanical differential analyser designed by Vannevar Bush of the Massachusetts Institute of Technology was used in preparing the tables. The Moore School of Electrical Engineering at the University of Pennsylvania at Philadelphia owned a similar, more powerful, differential analyser. The university was a few hours’ drive from Aberdeen. The army decided to contract out the preparation of some artillery firing tables to the university. It appointed Lieutenant Herman Goldstine, who had a doctorate in mathematics from the University of Chicago and had been inducted into the army, to liaise with Moore School and oversee the progress of the project. As there was considerable work, the university also employed around 200 ladies who were trained to use mechanical desk calculators to prepare the tables. Adele Goldstine, Herman’s wife was a qualified mathematician, having obtained a Master’s degree in mathematics from the University of Michigan, and assisted in selecting and training the ladies. These ladies were called “computers”! The calculation of the tables was laborious. It took 20 hours of manual calculation to tabulate a 60-second shell trajectory. The differential analyser was faster and took 15 minutes for the same task. As the number of tables required was huge the project was behind schedule. A faster device for calculation was required.

Moore School of Electrical Engineering at the University of Pennsylvania had a group of engineers who were teaching electronics. Among them was John Mauchly, who was interested in developing electronic computers. He had obtained a Ph.D. in physics from Johns Hopkins University in 1932. He taught physics at Ursinus College, a small liberal arts college in Pennsylvania. He was an extrovert and liked to learn about new developments. In June 1941, he had visited John Atanasoff who was building a digital computer with vacuum tubes at the University of Iowa to solve simultaneous algebraic equations. Atanasoff and his student Clifford Berry had used binary numbers and Boolean logic in designing their machine. The computer called Atanasoff- Berry Computer (ABC) was not fully functional as war intervened and the navy drafted Atanasoff. Mauchly saw how vacuum tubes were used to perform arithmetic operations and took copious notes about the design of the machine [1]. When he returned from Iowa, he was pleasantly surprised that he had been admitted to a wartime training course in electronics at the Moore School. The course was about the design of vacuum tube circuits and J. Presper Eckert was an instructor. After the course, Mauchly was offered a position of an instructor in electronics at Moore School.  He accepted the offer as he was interested in designing electronic circuits. He became friendly with Eckert who he knew was an excellent designer of electronic circuits. Mauchly was familiar with the differential analyser and its use in calculating artillery tables. He was also aware of its slow progress. He discussed with Eckert his idea of designing an electronic computer to speed up the calculation. They teamed up and proposed to John Brainerd, the head of Moore School, that an electronic computer be built to calculate the artillery tables. Brainerd agreed.  Eckert and Mauchly wrote a proposal in early 1943 requesting funding of $61,700 from the US army to build Electronic Numerical Integrator and Computer (ENIAC). Goldstine who was liaising with the group at Moore School supported the proposal. The US army agreed to fund the project and released funding in June 1943 and classified the project as top secret. ## Design of ENIAC Eckert and Mauchly started constructing ENIAC as soon as the money was released. Mauchly was the designer of the logic of the system. Eckert designed the electronic circuits and built the computer. The computer stored decimal digits in accumulators. Each accumulator was 10 digits long and there were 20 accumulators. Counters were used for arithmetic operations and accumulators stored the results. ENIAC could add, subtract, multiply, divide, and find the square root of a number. It had three function tables to store values of arbitrary functions. The circuits were driven by a clock that gave a pulse once every 10 microseconds. It could add 5000 and multiply 333 ten-digit numbers in one second. Data were input to the computer using an IBM card reader and the results of the computation were punched on an IBM card punch. An offline printer printed the punched results. ENIAC did not have a memory. It was programmed by interconnecting its units by plugging wires on a large 1 square metre plugboard. It was tedious and required a good knowledge of the logical structure of the computer. It took weeks to set up a program to solve a problem and debugging was difficult. As it was primarily used in its early days to tabulate artillery tables this was not a serious disadvantage. It could be programmed to perform looping, conditional branches, and subroutine calls. Six mathematically savvy ladies were selected from among the ladies operating the desk calculators and trained to program the computer. Their contribution was invaluable in using ENIAC to solve problems. Unfortunately, at that time, their importance was not recognized and they were not even invited to the dinner that was given to celebrate the successful completion of ENIAC [2]. Belatedly, in 1997 they were inducted into the Women in Technology International Hall of Fame. (Left) J. Presper Eckert (Photo courtesy nationalmedals.org) & John Mauchly (Photo courtesy ethw.org) ENIAC was constructed one unit at a time. As the construction progressed, the project was frequently running out of funds. On the recommendation of Goldstine, the project was periodically renewed and funds released. When it was completed, it cost a little over$ 485,000 and used around 18,000 vacuum tubes, 7200 diodes, 1500 relays, 70,000 resistors, and 10,000 capacitors.  It weighed 27 tons, occupied 1800 sq. ft. room, and consumed 150 kW power. It required a massive air circulation system to cool it.

One of the major problems encountered during its use was the failure of vacuum tubes. The average life of a vacuum tube was 3000 hours. With 18,000 tubes the probability of failure of one of the tubes and consequently the computer was around 10 minutes.  Eckert increased the life of vacuum tubes a hundred-fold by operating them at two-thirds rated voltage. The filaments of the tubes fail if they are turned on and off frequently. Eckert decided to keep the computer always on [3]. With these changes, the computer stabilized and used to fail only once a day.  Procedures were found to quickly locate the bad vacuum tube and replace it. ENIAC could tabulate the 60-second trajectory of an artillery shell in 30 seconds, 30 times faster than the differential analyser.

A view of ENIAC (Photo courtesy Wikipedia.org)

The machine was fully functional in late 1945 and was publicly unveiled on February 15, 1946. By that time the second world war was over.

(Left) John VonNeumann & Herman Goldstine
(Photos courtesy Wikipedia.org)

## Von Neumann’s involvement with ENIAC

John Von Neumann was a famous mathematician who was a professor at the Institute for Advanced Studies at Princeton. He was a consultant to the US army and was a frequent visitor to BRL at Aberdeen. He was interested in using computers to solve mathematical problems related to the design of the atomic bomb in which he was involved. In this connection, he had visited the computing laboratory at Harvard University and  the Bell Telephone Laboratories. Both these computers were too slow to solve his problem. Goldstine met him by accident at the Aberdeen railway station in August 1944. He was in awe of the great mathematician and built up the courage to talk to him. He was pleasantly surprised when Von Neumann evinced a keen interest in computing. Goldstine briefly described ENIAC and mentioned that it used vacuum tubes and multiplied 333 ten-digit numbers per second. Von Neumann who had seen slow relay computers was  interested in knowing about the fast electronic computer. A visit was arranged for him to see ENIAC. He went to Moore School later in August and was requested to be a consultant to the ENIAC group. In 1944, Eckert and Mauchly were exploring how to design a memory for ENIAC and use it to ease programming. When Von Neumann joined the group, the discussion was steered towards programming ENIAC. During several brainstorming sessions, the idea of storing program and data to be processed by the program in the main memory of a computer emerged. The construction of ENIAC had progressed far and it was too late to incorporate in it a memory to store programs. Even though Eckert had an idea of designing a memory using mercury delay lines it would have taken too long to realise it. It was decided to incorporate this idea into a successor computer EDVAC (Electronic Discrete Variable Automatic Computer).

Extensive notes were taken during these discussions. Von Neumann used these notes and during a long train journey to Los Alamos wrote a document describing the logical structure of the proposed computer. It was titled “First draft of a Report on the EDVAC”. He sent it to Goldstine.  Goldstine got it typed and attributed the document solely to Von Neumann, even though it was the result of a group effort. Several copies were made and distributed in June 1945. This document described in detail the architecture of a stored-program computer as one consisting five units; an input unit to read program and data, the main memory in which the program and data are stored, a control unit that interpreted instructions of a program and ordered the other units to carry them out, an arithmetic unit to carry out arithmetic operations,  and an output unit to print the results. This was one of the most important documents in the history of computing as it steered the development of all future computers to this day. The architecture came to be known as the “Von Neumann architecture” even though  the idea was the result of  brainstorming that took place with several participants including Eckert and Mauchly.

## ENIAC – Second Avatar

A dispute arose between Eckert – Mauchly and the University of Pennsylvania on the patent rights of ENIAC. It led to Eckert and Mauchly leaving the university in March 1946. They started a company named Electronic Control Company, that was renamed Eckert-Mauchly Corporation, to manufacture computers. They also took along with them many engineers and programmers from the ENIAC team. Those who were left continued to use ENIAC.

In December 1946, the US army that had sponsored the design and construction of ENIAC decided to shift it from Moore School to BRL at Aberdeen. ENIAC weighing 27 tons and with 18,000 delicate vacuum tubes was disassembled, packed, and shipped to Aberdeen; an arduous task! A group of engineers at BRL reassembled it which was difficult and took long. It was quite a challenge to get it back to a working state. After seven months, in late July 1947 ENIAC was switched on and started working. Between December 1946 and July 1947, Von Neumann and Adele Goldstine started exploring methods for simplifying programming ENIAC as wiring the plugboard for each program was inflexible, tedious, and took weeks. ENIAC had three function tables each storing 1200 digits. They decided to re-purpose these function tables  and use them as a read-only memory in which programs  can be stored. Pulses generated from these tables were to be used to control the sequencing of calculations. A set of around 80, two-digit instructions (called order code) was designed by the group that included  Richard Clippinger of BRL. Von Neumann suggested how to re-purpose some accumulators to work as program counter and as a single address computer. The hardware of  ENIAC was modified by adding a unit to facilitate the use of function tables as program memory. The modified ENIAC initially failed to execute programs. The engineers slowed down the computer by reducing the clock speed and it worked [4]. ENIAC started working almost like a stored program computer in March 1948. It was used to successfully run simulation programs (using what is known as Monte Carlo method) to aid in the development of the hydrogen bomb. With this success, ENIAC was used for solving numerous problems such as weather prediction and wind tunnel design. Programming a new problem took days rather than weeks. ENIAC was the fastest and most easily programmable computer in the world between 1948 and 1950.

ENIAC was the workhorse for the US defence department during the late 1940s. It was improved with additional hardware, including a high-speed shifter in 1952 and a 100-word random access magnetic core memory built by Burroughs Corporation in 1953. Meanwhile, faster and more versatile commercial computers were entering the market. ENIAC was shut down on October 2, 1955. It is now in the Smithsonian museum in Washington D.C. The US post office released a stamp with the picture of ENIAC in 1996 on its golden jubilee.

Acknowledgment