The Archimedes was so far ahead of its time that no-one really knew what to do with it. They ended up in schools, especially in commonwealth countries, mainly because Acorn's previous big success, the BBC Micro, had been specifically targeted at schools.
But in 1987 the ARM was so astoundingly powerful that it should have become the king of the workstation market. A failure of marketing, perhaps.
The failure of archimedes was it not being a PC clone, at which all "big boys" were putting bets on, and secondary to that was Taiwanese not buying into it.
And remember, back then the whole notion of a _personal_ PC was very new, and thus companies with first mover advantage got huge leverage by instilling the idea into common people that AT/XT "common phenotype clone" is the computer.
Lots of people then got an idea that a computer must be a kind of at/xt derivative and nothing else:
My parents brought a second hand CZ310 from Japan in late 1994. A super expensive machine even for them (they were possibly within the top 2000 at that time in Russia.) I remember, mom saying about those times affectionately that she asked for "the best computer the money can buy." And that when it gave up the ghost in 1997 or 1996, the repairman simply did not believe that it was a computer and not a some kind of a gaming console.
It doesn't help that they didn't ship it with Unix, which is what workstations ran back then. Porting System V to it and selling it with it may have helped.
Then again, both Atari and Amiga tried that with their 030 machines and had no success. They were not taken seriously as workstation vendors (in the then lucrative workstation market) in many cases because of the brand on the box. I distinctly remember a UnixWorld article about the Atari TT030 with the headline "Up from toyland" (above an otherwise positive review)
"What I can say with certainty is that the Archimedes running C programs without a math coprocessor rivals the Compaq Deskpro 386, a 16-MHz 80386 machine with an 8-MHz 80287, and comfortably outpaces a Macintosh SE with a Hyper- Charger, a 15.67-MHz 68020 with a 7.83-MHz 68881 , on all but the floating- point-intensive Savage benchmark (the Compaq also beats the Archimedes on the Sort). Even more remarkably, the Savage benchmark in interpreted BASIC V in RAM on the Archimedes takes only half again as much time as it takes in compiled C on the Deskpro 386 with a math coprocessor.
Benchmarks are not everything, but the experience of using the Archimedes tells me that on many untested tasks, like writing to the screen, it is far faster than anything else I've seen. If I had to take a stand on benchmark figures alone, I would look at the Dhrystone, which is the most general-purpose test (even though it doesn't test floating point). The Archi- medes runs 3 1 percent more Dhrystones."
I had one from 1988. It was totally awesome. My first experience with a Windows PC 5 years later was a like being slapped with a wet fish.
Also, the 8MHz ARM out performed the 486 100MHz by a long way
But in 1987 the ARM was so astoundingly powerful that it should have become the king of the workstation market. A failure of marketing, perhaps.