ENIAC : A computer is born
ENIAC--monster and marvel--debuted 60 years ago
By Michael Kanellos
Staff writer, CNET News.com
February 13, 2006, 4:00 AM PST
In February 1946, J. Presper Eckert and John Mauchly were about to unveil, for the first time, an electronic computer to the world. Their ENIAC, or Electronic Numerical Integrator and Computer, could churn 5,000 addition problems in one second, far faster than any device yet invented.
The scientists knew that they had, but they weren't sure how to convey their breakthrough to the public. So they painted numbers on some light bulbs and screwed the resulting "translucent spheres" into panels. Dynamic, flashy lights would thereafter be associated with the computer in the public mind.
That touch of showmanship would later prove fitting for the importance of the ENIAC, which celebrates its 60th anniversary this week at the University of Pennsylvania?s Moore School of Electrical of Electrical Engineering. Many historians acknowledge that other computers came earlier--the Z3 in Germany, England's , the Atanasoff-Berry Computer (ABC) at Iowa State. But ENIAC arguably accomplished something more important: It sparked the imagination of scientists and industrialists.
In a few years, computers would pop up at universities, government agencies, banks and insurance companies. A UNIVAC computer (with decorative lights, of course) from the company subsequently founded by Eckert and, predicted the outcome of the 1952 presidential election, while another appeared in a brassiere ad touting yet another advancement in science. The English code cracking machine, Colossus, became famous in military circles. But it was demolished after the World War II and remained shrouded in secrecy for decades.
Compared with other computers that performed such practical functions, ENIAC was an odd bird in technical terms. It relied on a 10-digit decimal system, rather than the binary systems of ones and zeros used by virtually all subsequent computers, even those developed by Eckert and Mauchly. Programs could not be stored on ENIAC. It didn't really employ conditional branching--the if/then statements that form the cornerstone of modern programming.
And only one ENIAC, in fact, was ever built.
"It was a monstrosity. It was rapidly overtaken by general purpose machines," thundered Jay Forrester, a Massachusetts Institute of Technology professor and one of the leading computer architects of the last century. "There wasn't anything in it that survived into modern machines, except maybe electricity."
Butwith an indisputable fact: It worked. Until it was immobilized by lightning in 1955, ENIAC performed computational problems relating to the development of the hydrogen bomb and other military projects. Penn professor Irving Brainerd once even speculated that during the 80,223 hours ENIAC operated, it crunched more calculations than had been performed by all humanity since time began.
"Some, judging from the tube replacement rate in home radios, said this monster could not run for five minutes! However, all tubes were 'burned in' (tested) for 100 hours, so this was not a problem," ENIAC engineer Harry Huskey, 90, said in an e-mail interview from his home in South Carolina.
Some of ENIAC's competitors, namely the ABC and Z3, were far slower and could tackle only small problems. A debate even exists about whether the ABC, which performed calculations only for demonstrations, was ever finished. Eventually, the experience of the inventors behind those computers became a cautionary tale of scientific egos.
Beating the Germans
The roots of ENIAC and its contemporaries can be traced to World War II. Artillery units employed tables to help them predict the trajectory of the shells they were firing, but calculating the variables--the angle of the gun, the condition of the terrain and other factors--was a mind-numbing task.
Hear ENIAC programmer Jean Bartik explain how the computer was tested.
Listen up... (2.3MB mp3)
Figuring a single trajectory (out of several hundred) with a hand calculator took around 40 hours, and even electromechanical devices like the Differential Analyzer designed bymight take 30 minutes. Ballistics tables had a limited tactical value at best, said Mitchell Marcus, a professor of computer science at Penn.
At the same time, academics were trying to come up with ways to accelerate the electromechanical machines and eliminate errors caused by stuck pins or poorly shifting gears. In 1937, Iowa State professor John Atanasoff sketched out an idea on a bar napkin for an electrically powered box that could solve equations through binary math.
ENIAC--monster and marvel--debuted 60 years ago
With help from grad student Clifford Berry and a few research grants, Atanasoff built a prototype of the ABC computer, which was demonstrated in October 1939. A more advanced version, which contained 300 tubes and took several seconds to complete a math problem, was erected in 1941. It worked, but the outbreak of war forced both Atanasoff and Berry to drop it and move to more urgent defense projects.
Hear ENIAC programmer Jean Bartik describe the mammoth computer.
Listen now... (1.2 MB mp3)
Mauchly, then a professor of physics at Ursinius College, was working on an entirely different kind of science, researching ways to better predict weather with an analog device called a harmonic analyzer. Atanasoff attended a lecture by Mauchly in December 1940 and afterward the two began to correspond and discuss potential for electronic computers.
Mauchly soon had another life-changing encounter. Obsessed with electronics, he enrolled in a course at the Moore School being taught by Eckert. By the end of 1941, Mauchly was teaching at Penn and discussing computing ideas with Eckert.
The two had complementary skills. Mauchly was an expert in physics and math, but early in his career had disdained engineering. Eckert, the only son from a wealthy, globe-trotting Philadelphia family, was a natural tinkerer. At age 14, he rigged up an intercom system in his dad's building, and the Connecticut Telephone and Telegraph later bought it from him.
"Eckert was an absolute genius as an electrical engineer. He was one of the best designers of the 20th century," said Michael Williams, professor emeritus of history at University of Calgary and president-elect of The Institute of Electrical and Electronics Engineers. Mauchly, he added, was able to envision how such a machine might work.
Ideas on paper come to life
Mauchly exhibited that vision in a five-page memo titled "The Use of Vacuum Tube Devices in Calculating." It presented ideas that wove their way through the university-government bureaucracy and ultimately resulted in contract W-670-ORD-4926. Under that agreement, signed June 5, 1943, Penn would research the possibility of an electronic differential analyzer for the U.S. Army Ordnance Department.
The proposed work was to last six months and cost $61,700--a vast underestimation, it would turn out, of both time and money. ENIAC wouldn't be tested internally for two and a half years, in November 1945, at a final cost of $487,000. Despite the overruns, however, it was an engineering marvel.
How it worked
At the heart of ENIAC was a device called a ring counter, which consisted of 10 vacuum tubes in a circle. A "5" would be represented by a pulse at the fifth tube. If a person added 9 to that, the pulse would shift to the fourth tube, while the first tube on a second ring--representing the 10--would receive a pulse.
Ten ring counters were placed in each accumulator, which could store numbers ranging up to 10 billion minus one (9,999,999,999) or down to negative 10 billion plus one. When a single accumulator hit its maximum, a pulse could be sent via wire to a second one, continuing the process. In all, ENIAC contained 20 accumulators spread over 40 racks networked together through plug boards. Data was stored in pulses in 5-foot mercury tubes.
Six technicians were largely responsible for working the mathematical equations and programming functions. Because these jobs were considered an extension of clerical work, they were filled by women, as was the practice of the day.
"They were the first programmers and they didn't get the credit they deserved," said Kathryn Kleiman, an internet privacy lawyer with McLeod, Watkinson and Miller and a member of the Association for Computing Machinery's committee on women in computing.
One of the biggest challenges was preventing vacuum tubes from blowing out. Because the tubes would be required to pulse 100,000 times a second and the machine had so many of them, the threat of a blowout was constant. Eckert solved the problem by running below their threshold and designing the system to operate under the "worst worst scenarios," Williams said.
The scientists faced another concern that was decidedly low-tech but equally important: rodent control. "We knew mice would eat the insulation off the wires, so we got samples of all the wires that were available in a cage with a bunch of mice to see which insulation they did not like. We used that wire," Eckert said in a 1989 interview with Alex Randall, professor of electrical engineering at the University of the Virgin Islands and friend of the Eckert family.
Finally, ENIAC underwent its first full test in November 1945, when computational problems from the H-bomb product were fed into it. On Feb. 14, 1946, the Moore School invited Army officials, Penn professors and select scientists from around the country for a demonstration. Contrary to popular myth, the lights in Philadelphia did not dim and soldiers did not salute the machine.
Also contrary to popular myth, most people didn't care. Although the Moore school immediately began to get inquiries from other universities and researchers, the public mostly ignored it, despite front-page news articles. That, of course, would change with time.
"I don't think one could say there was much fanfare just after the unveiling," wrote Arthur Burks, one of the original engineers. "Those in the know had great expectations, but even these (expectations) would not have even approached what actually occurred over the years."