On this page
During the Second World War, the "computers" at the University of Pennsylvania were people: teams of women working through ballistics calculations by hand. The US Army needed more of those calculations than they could finish. The machine built to solve that problem was ENIAC. This post covers who built it, how it worked, two worked calculations about its speed and reliability, and what students can take from the team behind it.
Why was ENIAC built?
Penn Engineering's ENIAC history explains that the Army needed firing tables, which told gunners how to aim their weapons to hit a target. Teams of women at Penn's Moore School of Electrical Engineering computed these by hand and with a mechanical differential analyser, yet a backlog kept growing.
John Mauchly was a professor at Ursinus College when he enrolled in a wartime electronics course at the Moore School. Penn later hired him to teach. There he met J. Presper Eckert, a Penn graduate who ran the course's laboratory. Both wanted to speed up computation. After they secured a contract from the Army's Ballistic Research Laboratory, work on ENIAC began. The Computer History Museum dates the build to 1943 to 1945.
How did Mauchly and Eckert divide the work?
Penn describes Mauchly as the visionary behind the machine's use of vacuum tube technology, and Eckert as the engineer who solved its technical problems. The hardest problem was reliability. Eckert got good results by running the tubes at one-quarter power.
ENIAC was huge. Penn lists a room 30 by 50 feet, a weight of 30 tons, 40 panels, about 18,000 vacuum tubes and a power draw of 150 kilowatts. Twenty accumulators handled addition and subtraction, with separate units for multiplication, division and square roots. It was announced on 14 February 1946.
Who programmed ENIAC?
Six women did: Kathleen McNulty Mauchly Antonelli, Jean Jennings Bartik, Frances (Betty) Snyder Holberton, Marlyn Wescoff Meltzer, Frances Bilas Spence and Ruth Lichterman Teitelbaum. Penn notes they programmed each job by plugging in cables and setting switches. Meltzer and Teitelbaum ran the February 1946 public demonstration, yet the press coverage of the time featured only men. The six were inducted into the Women in Technology International Hall of Fame in 1997.
Worked example 1: how much faster was ENIAC?
Penn says a ballistics calculation that took 12 hours on a hand calculator took ENIAC 30 seconds.
Convert to the same unit: 12 hours = 12 × 60 = 720 minutes = 720 × 60 = 43,200 seconds.
Divide: 43,200 ÷ 30 = 1,440.
So ENIAC was 1,440 times faster on that job, the same factor Penn gives.
Worked example 2: why did tube reliability matter?
The museum notes that ENIAC lost a vacuum tube roughly every day or two, out of almost 18,000. Take one failure every 1.5 days as a rough middle value.
If 18,000 tubes together give one failure every 1.5 days, an average tube runs about 18,000 × 1.5 = 27,000 days between failures.
27,000 ÷ 365 ≈ 74 years.
Each tube was very reliable on average, yet the machine still stopped every day or two. With thousands of parts, rare failures add up. That is why Eckert's choice to run tubes gently mattered, and why the team learned to find and replace a failed tube in about 15 minutes.
Common mistakes: dividing hours by seconds without converting units; reading "one failure a day" as "tubes last a day". The second worked example shows the opposite.
What happened to ENIAC?
After construction it moved to the Army's Aberdeen Proving Ground in Maryland. Penn says it served as the Army's main computing engine for the next nine years, working on firing tables, weather prediction, atomic-energy calculations and other problems. The museum adds that it ran for a decade, until a 1955 lightning strike.
Mauchly and Eckert went on to build UNIVAC I. The US Census Bureau commissioned it in the late 1940s and first used it on the 1950 census, the same job that Herman Hollerith's punched cards had transformed in 1890. For the stored-program idea that followed ENIAC, read about John von Neumann.
What can a student take from the ENIAC team?
Pair vision with engineering. Mauchly saw what was possible; Eckert made it work. Most hard projects need both.
Design for failure. With 18,000 parts, Eckert planned for reliability from the start.
Do the unit conversion first. The speed-up only makes sense once hours and seconds are the same unit.
Credit the whole team. The programmers were left out of the story for decades.
Rough-estimate problems like these appear in physics too. If you get stuck on how to set one up, the Fermi Problem Library has worked problems with a step-by-step solver that explains the reasoning rather than printing an answer.
Try it now: if ENIAC drew 150 kilowatts for an 8-hour shift, how many kilowatt-hours did it use? (Answer: 150 × 8 = 1,200 kWh.)
