The world’s oldest functioning computer is now the WITCH, aka the Harwell Dekatron Computer, recently resorted to operating condition in Britain. It is also, almost certainly, the world’s slowest functioning computer.
The WITCH, built at Britain’s Atomic Energy Research Establishment, was designed circa 1949 and completed in 1951. Circuits are comprised of dekatrons..a type of electron tube with an inherent count-to-ten capability..and ordinary telephone relays. The machine can store 40 eight-digit numbers, and while it can operate in stored-program mode, it was more normal to execute programs directly from punched paper tape. A “loop” when operating in this mode was an actual physical loop, with the ends of the tape glued together.
This computer was designed for reliability rather than blazing (by the standards of the time) speed, and it was so slow that a human could keep up with it for a limited period of time—but after half an hour or so the human would have to drop out, exhausted, while the machine plowed on, “utterly relentless.” I’d guess that the WITCH..taking into account its ability to operate unattended..could do as much computation as a human group of 5-10 people. Despite the machine’s slowness, the staff at Harwell found it useful, and it continued in use there until 1957, when it was given to Wolverhampton Polytechnic…which is where it got its current name, the Wolverhampton Instrument for Teaching Computer from Harwell. There it supported research and teaching until 1973. It was rescued from storage a few years ago and restored by volunteer efforts. It is now on display at Bletchley Park, the WWII center for codebreaking.
The restored WITCH displaces the previous oldest-working-computer champion, the Ferranti Pegasus.
WITCH video here.
Related post about the reconstruction of another early British computer, the EDSAC.
That is amazing….
About 1974 I worked with a Burroughs B-700 – an early minicomputer with performance that would be laughable today. But it was light years ahead of the Harwell Dekatron
All histories of computing written before the Ultra revelations of the mid 70s must have been unavoidably wrong. Do any of their errors survive in modern accounts? I’m thinking more of popular accounts than scholarly ones.
An interesting story. In 1959, we used to do the calculations on a Marchant calculator before the program went into the IBM 650. The Marchant was probably faster than the Dekatron.