6502.org

Well, hence the ARM. For a fully-featured 32-bit CPU, it was very small and frugal, and could thus be built into a relatively inexpensive home computer (ie. the Archimedes). For the most part, the transition to its current life as an embedded CPU happened after the Archimedes stagnated.

This is apposite too: the ARM ran a very efficient 6502 emulator, allowing for Acorn's Archimedes range to have some binary compatibility with applications built for the 6502-based BBC Micro. Indeed, one of the Archimedes range was branded as a BBC machine, somewhat later. (There was also a degree of compatibility in the Archimedes running BBC Basic, in a version fairly compatible with previous versions running on 6502 (and on other platforms.))

Today, we can run an emulated 6502 on ARM, specifically on Raspberry Pi, which does a very good job of acting as a minuscule but very fast second processor to Acorn's original 6502 line.

Don’t forget that the Intel line went:
8008 -> 8080 -> 8085 but the Z80 from Zilog became much more popular than both the 8080 and the 8085.
The 8080 was NOT binary compatible with the 8008 but was source-compatible.
The 8085 was binary compatible with the 8080.

The 8086 was in part developed to counter Zilog and Motorola. Intel designed the 8086 so that conversion of existing 8080 and 8085 code could easily be done and run on the 8086 (source-compatible).

Of course, the 8085 is rather forgotten now, because the Zilog Z80 when compared to the 8080 was easier to use (hardware wise), had an extended instruction set and was binary compatible with the 8080. Hence by the time the 8085 came to market, the Z80 was the go - to choice if you wanted to run CP/M or 8080 based software.

If intel had not devoted a lot of effort into the 8086, it would have been interesting to see what would have happened. Would 8 bit Z80, 6502 and 6809 microprocessors continued for a bit longer until the price of the Motorola 68000 fell to a more affordable level? Would a later (non-binary compatible) Zilog microprocessor got a bigger slice of the market? If Zilog had introduced a binary compatible 16 bit upgraded Z80 at the time, would that have made a difference? One can only speculate.

Unfortunately for the 6502 side of the market, the 65C816 did not come to market until 1984. By then, the Motorola 68000 was seen as the future in home computers and Intel were dominating the PC market. Sinclair announced the 68008 based QL in 1984 and the Apple Macintosh was launched in 1984. Atari went the 68000 route with the ST range in 1985, Commodore went the 68000 route with the Amiga range and although Apple launched the Apple IIGS which used the 65C816, Apple then went the 68000 route for later machines.

The other thing worth talking about is software. With microprocessors, outside the embedded and specialist markets, they don’t sell well unless software that people want to use becomes available. But often, unless there is seen to be a market, software that people want to use will not be written for new microprocessors.

With some elements of backwards compatibility, this chicken and egg / catch 22 problem can be mitigated.

Of course if the new system is significantly faster, older software can now run under emulation...

Mark

The efficiency of that software matters somewhat, too. Okay, today we are inundated with "managed" and "script" languages that run quite a lot slower than natively compiled software ought to, and eat RAM for breakfast and then ask for more - but believe it or not, it is quite possible to do worse. Indeed, the Archimedes - and the ARM with it - was almost rendered irrelevant by bad software before a single unit was sold.

Anecdotally, the all-singing all-dancing UNIX-like ARX was largely written in Modula-2, the ARM compiler for which would regularly emit 20 instructions plus a subroutine call for an operation the CPU could actually handle in a single instruction. Memory at the time was still the dominating cost factor for the machine, so the fact that ARX required 4MB to run at all was a big problem. Arthur, which was almost a direct port of MOS from the BBC Micro with a GUI slapped on top, ran quite happily in 256KB. Acorn management were reportedly very troubled by the situation, but for fiscal prudence needed a cast-iron reason to scrap ARX. So a comparative demo was organised - and the details of what the ARX team planned to demo was leaked to the Arthur team.

So a top-of-the-line 4MB Archimedes was set up for the ARX demo, which proceeded to display sixteen clocks, thrashing madly at the swap and singularly failing to update every clock every second. Then the Arthur team unveiled a bottom-of-the-range 256KB Archimedes running what appeared to be the same sixteen clocks, but ticking smoothly every second, in absolute silence.

ARX was cancelled the very next day. Arthur became RiscOS.

Don't have anything to add other than to say I've certainly enjoyed reading this thread. Perhaps it's not gone the way that the thread originator wanted, but I think it has generated a lot of interesting information and trivia. Thanks.

Yes, it was gone too far i think it would,