ENIAC

Scale and purpose

The scale of the machine runs to 17,468 vacuum tubes, 7,200 crystal diodes, 1,500 relays, 70,000 resistors, 10,000 capacitors, and some five million hand-soldered joints. Keeping tubes on this scale running at all times meant enormous power consumption and heat. Contrary to the expectation that more tubes means more failures, the actual rate of operation seems not to have been low; Eckert recalled in later years that a tube failed about every two days and that the location of the problem could be found within fifteen minutes.

The purpose was artillery firing tables, that is, tables of trajectories according to the angle and range of a gun. At the time this calculation was done by hand and with mechanical calculators, and finishing a single table took an enormous amount of time. The work it was actually put to first, however, is said to have been calculation related to nuclear weapons rather than firing tables. There is an account that practical operation began on 10 December 1945 and that the first problem was an examination of the feasibility of a thermonuclear weapon.

Numbers are handled not in binary but in decimal. Each digit is held in a ten-position ring counter built from vacuum-tube flip-flops, and there were twenty accumulators, each holding a signed ten-digit number in ten's complement form. Where later computers converged on binary, ENIAC carried the format humans read and write straight into the machine.

How it was run

The greatest constraint lay in how a program was given. ENIAC made the flow of processing with cables and switch settings on plugboards, and having it do a different calculation required physically rewiring it. This work took time on the order of days, and the structure was one in which the speed of the calculation itself was cancelled out by the slowness of setting up. In other words, ENIAC was a machine fast electronically but mechanically slow to change what it did.

This wiring work and the formulation of the problems were carried out by six women hired as computing staff: Kathleen McNulty, Jean Jennings (later Bartik), Frances Snyder (later Holberton), Marlyn Wescoff, Frances Bilas and Ruth Lichterman^※1. With neither wiring diagrams nor instruction tables, the work was to work out how to assemble the processing from the machine's circuit diagrams, which amounts to programming in today's sense. It long went unmentioned in public and has been reassessed in later years.

In 1948 it was modified so that the sequence of processing was placed in function-table switches, making it possible to change programs without rewiring^※2. It was a retrofitting of the idea of the stored-program method onto an existing machine, a modification that partly removed the constraint of slow setup. The group led by Bartik and others is credited with leading it.

The dispute over "the first"

Electronic computers were running before ENIAC. Britain's Colossus had been running since 1943 for codebreaking, but its use was specific and it remained classified for long after the war. The ABC, by John Atanasoff and Clifford Berry at Iowa State University, was running around 1942, but it too was not a programmable general-purpose machine. Zuse's Z3 in Germany dates from 1941 and was relay-based rather than electronic.

This "first" became a point of contention in court as well. In 1967 Honeywell brought suit against Sperry Rand, and on 19 October 1973, in the federal district court of Minnesota, Judge Earl R. Larson handed down a decision setting aside the ENIAC patent granted in 1964. The ABC was held to constitute prior art, and the ENIAC inventors were found to have derived the subject matter of the electronic digital computer from the ABC. Accounts differ, however, on the legal ground for the decision, some giving unreasonable delay in filing and some giving inequitable conduct in concealing knowledge of prior art^※3.

The end, and what remained

ENIAC was moved in 1947 to Aberdeen, where the Ballistic Research Laboratory was, and its operation ended when the power was switched off on 2 October 1955. That makes it a computer used in practical work for about ten years. Mauchly and Eckert resigned a month after the unveiling, refusing to turn their patent rights over to the university, and founded their own company. This course of events is a distant cause of the later patent litigation.

Seen in terms of technical lineage, the problems in ENIAC's design became the starting point of the next design as they stood. The inconvenience of giving a program by wiring leads to the method of placing instructions in the same storage as data, that is, to the design of EDVAC. What ENIAC showed was not a finished form but the fact that an electronic, general-purpose computer could actually be built and used in practical work — and also a list of what to fix next.