Everybody knows Intel's 4004, designed for a calculator, was the first CPU on a chip. Everybody is wrong.
For a long time, what is now considered to be a prime candidate for the title of the 'world's first microprocessor' was a very well-kept secret. The MP944 is the inauspicious name of the chip we want to highlight today. It was developed to be the brains behind U.S. Navy's F-14 Tomcat's Central Air Data Computer (CADC). Thus, it isn't surprising that the MP944 was a cut above the Intel 4004, the world's first commercial microprocessor, designed to power a desktop calculator.
The MP944 was designed by a team of engineers approximately 25-strong. Leading the two-year development of this microprocessor were Steve Geller and Ray Holt.
The processor began service, in the aforementioned F-14 flight / control computer in June 1970, over a year before Intel's 4004 would become available, in November 1971. An MP944 worked as part of a six-chip system for the real-time calculation of flight parameters such as altitude, airspeed, and Mach number – and was a key innovation to enable the Tomcat's articulated sweep-wing system.
By many accounts, the MP944 didn't just pre-date the 4004 by quite a margin, it was significantly more performant. The tweet, we embedded above, suggests Geller & Holt's design was "8x faster than the Intel 4004." Completing all the complicated polynomial calculations required by the CADC likely dictated this degree of performance it delivered.
[...] As well as offering amazing performance for the early 1970s, the MP944 had to satisfy some stringent military-minded specifications. For example, it has to remain operational in temperatures spanning -55 to +125 degrees Celsius.
Being an essential component of a flight system also meant the military pushed for safety and failsafe measures. That was tricky, with such a cutting-edge development in a new industry. What ended up being provided to the F-14 Tomcats was a system that could constantly self-diagnose issues while executing its flight computer duties. These MP944 systems could apparently switch to an identical backup unit, fitted as standard, within 1/18th of a second of a fault being flagged by the self-test system.
As mentioned above, this processor of many firsts seems to be of largely academic interest nowadays. However, if Holt's attempts to publish the research paper outlining the architecture of the F-14's MP944-powered CADC system had been cleared back in 1971, we'd surely now all be living in a different future.
(Score: 0) by Anonymous Coward on Thursday November 20, @01:49PM
There is zero chance this would have been sold commercially at a non-military $$$
There was also the whole Four Phase / TI microprocessor patent catfight.
(Score: 2, Informative) by Damp_Cuttlefish on Thursday November 20, @04:04PM
It's for sure an interesting bit of history, but not by most measures a microprocessor. No single one of the six chips really does anything like what a 4004 alone can manage.
You can find a much better summary than I could write on Ken Shirriff's blog [righto.com], and if that tickles your pickle you might also enjoy this [cpushack.com]wander through the halls of microprocessor history
(Score: 4, Interesting) by VLM on Thursday November 20, @04:54PM
Thats always been a moving target. Not a F-14 joke.
For example "now a days" you slap a crystal across two pins of a microcontroller or inject a bland TTL square wave into a clock chip, but good luck doing anything with a 8080 "cpu" with its little buddy the 8224. But there's a mythology that the 8080 is an entire processor on a chip. Uh huh keep thinking that. Thats nothing, wait until you're introduced to the TMS9900 (usagi electric has a series on that chip). You need a 4-phase clock generator for that beast, using a weird dedicated clock gen chip. But its a "single chip CPU" uh huh it just can't run without its custom support chip stable, its totally a "single chip" design LOL ha ha. A 6809E wants a weird dual phase quadrature clock, is that a single chip CPU?
Until remarkably recently, external floating point units were a commercial success, ya know. Is every cpu chip that can't do FDIV on the die not a "real" processor? Trivia people say no, but the market says yes.
This is not just a CPU problem. Does a "complete video generator on a single chip" need to include sprite support? If so, sucks to be a motorola 6847, spriteless. Even a "fancy" TI 9918 doesn't include 3-d rendering stuff that would be considered mandatory now a days.