Digital computers are a fairly old technology at this point. The first ones date from the mid-1940s, which makes them older than nuclear reactors, integrated circuits, polypropylene, and orbital satellites.
What they aren’t is a durable technology. Computers age fast, both because they’re easily superseded by later models, and because their components wear out quickly. The early machines failed more than once an hour because of bad vacuum tubes, while current ones fail because of heating of their boards and aging in the wiring and transistors of their chips, and because of shifts in software environments that render old programs useless (“bit-rot”).
It’s not uncommon for jet aircraft to fly for 40 years, and there are still nuclear reactors running from the 1950s. Can any computers claim similar longevity? That is, are there any machines that are still performing something like their original function? Let’s exclude museum exhibits or hobbyist re-creations of old machines, because anything can be kept working if people love it enough. People love the USS Constitution enough to keep it an active warship of the US Navy at age 214, but it’s not keeping the British out of US waters any more. I’ll also exclude calculators because they’re not quite in the category even though the programmable ones might be Turing-complete.
I describe a few candidates below, and would love to hear of any other machines that people know of, but first let me describe a few machines that didn’t make the cut:
- The Space Shuttle flight computers – These were IBM AP-101 machines, which are compatible with the 360 series. The original machines in the 5 Shuttles were AP-101b’s, built in 1971, but they were replaced with the AP-101s line in 1992. The biggest change was to go from core memory to 64Kx1 SRAM chips. The Russians did use the Argon-16 flight computer in their Soyuz spacecraft from 1975 to 2010, but it wasn’t literally the same machine throughout.
- Air Traffic Control Machines – The automation of the US air traffic control system was started by Burroughs Corp in the 1960s, and originally ran on their machines. However, all the hardware was converted over to Motorola 68000 microprocessors in the 1980s, and to IBM and Motorola PowerPC microprocessors in the 1990s. The system is called Common ARTS, and is now in turn being replaced by one called STARs.
- BART Station Signage – The San Francisco subway system used DEC PDP-8e minicomputers to control the signs indicating when the next trains would arrive. They were installed in 1971, and there were rumors that they were still in use, but Gary Messenbrink (who wrote the PALBart PDP-8 assembler for use on those systems) reports that they were all replaced in the early 2000s by x86 systems. 30 years is a really good run for a computer, but not quite as good as the ones below.
So what are some genuinely old systems? I’ll describe a few in youngest to oldest order:
1978 – the Cadillac Seville Trip Computer
This was the first use of a microprocessor in a car. It used the Motorola 6802 microprocessor, a chip that used 8 bit data with a 16 bit address, and had 128 bytes of on-chip RAM. It was a $920 option, and could calculate miles per gallon, miles to empty, and time of arrival. About 57,000 of these cars were built, but it’s not clear how many had the option. The cars themselves are still readily available.
1977 Bridgeport Series 1 numerically-controlled milling machine
The Series 1 is a famous line of mills, with some 370,000 built over the last 70 years. They added computer control in the 70s, driven by a DEC LSI-11 board, which is part of the 16-bit PDP-11 family. The LSI-11 used a set of four VLSI chips from Western Digital called the MCP-1600, which was a microcoded general purpose CPU. The mill could be programmed to cut slots and circles using something called the BOSS software, which evolved through the late 70s and early 80s. The machines are still widely available, although the LSI-11s are often replaced by PCs.
1977 Voyager 1 flight computers
The two Voyager probes are still active after almost 35 years. They contain 3 computers each according to this FAQ:
Computer Command System (CCS) – 18-bit word, interrupt type processors (2) with 4096 words each of plated wire, non-volatile memory.
Flight Data System (FDS) – 16-bit word machine (2) with modular memories and 8198 words each
Attitude and Articulation Control System (AACS) – 18-bit word machines (2) with 4096 words each.
They are not microprocessors – they were built out of small-scale-integration (SSI) RCA CMOS chips by General Electric according to JPL’s specs. When the Voyagers were specified in the early 70s, there were no rad-hard micros available. The RCA 1802 came out in 1976, and would have served, but it was too late. Each machine had a backup for redundancy, and some of those have failed. A lot of the memory bits have gone bad as well, so JPL has programmed around them.
Being able to upload commands to the probes made a big difference. For instance, when Voyager 1 went past Saturn, it was able to rotate itself to keep the planet centered on the camera, and so avoid motion blur. It was a better probe then than when it went past Jupiter!
The probes are powered by radio-isotope generators, and they’re steadily losing power. JPL has already had to turn off most of the instruments. By about 2025 there won’t be enough power for even one instrument, and that’ll be it. At that point the probes will be almost 50 years old, and will almost certainly be the oldest operating computers. Given what a grand success the Voyagers have been, that will be a noble legacy.