> The II was a dead end. The 65816 was not a nice ISA to program for.

I mean you make this statement like its some kind of technical roadblock. Which might have seemed appropriate in the early 1980's but with a bit of digging would have been understood to be false.

There have been various proposals for 32-bit 65816's, and it wouldn't have been any different than any of the other architectural extensions done to other processors (never mind the conversion from 6502->65816). Modal decoders (ex: Arm), prefix bytes (ex: amd64), virtual execution modes (ex: vm86/i386), etc all provide a method for building a new architecture while maintaining backwards compatibility with the old.

This stuff is an economic or business decision. At Apple apparently it was a political/business decision, rather than anything technical. They could have (as the IIGS proved) just have created an mac like experience on the apple ][ line, while maintaining software compatibility. None of it would have been "pretty", but that is almost always the costs associated with evolving a technology stack. It took roughly a decade to evolve linux from a UP kernel to a SMP one without the BKL. In that time, and some of the results weren't what anyone would call pretty.

> There have been various proposals for 32-bit 65816's

The 68000 was already there - there was nothing to do except waiting for Motorola to turn out an even faster and compatible chip for next year's model. Motorola was also a design powerhouse back then and would continue to be one for the foreseeable future and, in many ways, lead the way ahead of Intel (who had a design that shares the same drawbacks the 65816 had).

> This stuff is an economic or business decision.

That's true. They could stretch the II family further (I would love to see what an Apple IV could have been), but then they'd need to make WDC design chips that were not on the roadmap (or acquire WDC and put those on the roadmap). This would be a higher technological risk than continuing with the Mac and 68K, which had a growing user base and a clear evolutionary path (because technical workstations were showing what you could do with a 680x0 paired with high-res screens and lots of memory). The risk, with the 68K was of being able to differentiate, both from lower end Windows boxes up to lower end UNIX workstations and the Mac did fine in that space for a long time.

> Motorola was also a design powerhouse back then

Was it though?

The 6800 was overcomplex and over-expensive as a result. The competition ate it in pretty much all markets. The nmos chips were either dramatically cheaper like the 6502, or dramatically more powerful, like the Z80. Some of that was a victim of the design era, right on the cusp of moving to depletion mode, but some of that was design. And other PMOS designs outcompeted it right up, like the Intel and even the F8.

The 6809 was a joke. Sure it was more powerful than the 6800, but it was going up against things like the 8086 and Z8000 which were far more powerful. And the price/performance was absolutely horrible.

The 68000 was a success, but it's difficult to ascribe that to positive aspects of the design. Like the 6800 it was overdesigned, spending a full 3rd of its gate count on microcode. And then there's the separate data and address registers... urg. I think it's safe to say it was a success to no small degree because of NatSemi's failure, everyone that compared the 68k to the 32k preferred the later. And while it certainly held an edge over the 8088/8086, that edge was only significant until the 386, after which it was all downhill.

Which brings us to the 88000. This design was so overdesigned the company seriously considered just dropping it and building someone else's design (can't recall, Alpha/PRISM perhaps?). The only reason they continued was Data General had a freak-out because they based their entire AViiON line on it. And once it became clear they were going nowhere, Motorola dropped it and went to IBM instead, and DG went Intel.

So over a two decade period with four separate designs, they only really had one outright success. This does not say "powerhouse" to me. Yes, I'm aware there are lots of reasons one design is successful and the next isn't, but it's hard to point to any of their designs and see anything particularly impressive in design or price/performance terms. The 68k was in the right place at the right time, and deserved to be picked up for that reason, but that was short lived.

> The first benchmarks indicated the plain Archimedes had higher FP throughput than a 16MHz 386 with a 387.

After some poking about, it is clear this statement is not true for any definition of "first" or "benchmark" or even considering confusion between 386 and 387.

In FP terms, the A3000 got 76 kWIPS, compared to a 387/40's 5.7 MWIPS. Scaling to earlier 16 MHz parts would suggest ~1.5 MWIPS, still almost 20 times as fast as the ARM. Even a bare 386DX40 gets 316 kWIPS, scaling suggests very similar performance overall.

It's plausible that an ARM2 would outperform a 386*SX*/16, but that's without the 387SX. No other combination of 38x parts would not outperform it.

> The first benchmarks indicated the plain Archimedes had higher FP throughput than a 16MHz 386 with a 387.

I could imagine a plain Archimedes being competitive with a 386 in integer throughput, but an integer-only machine outdoing an FPU in floating-point performance would be very surprising. I don't suppose you remember which benchmarks they were?

I still wish VMS had managed to survive the 90s. I still think its CLI has not been excelled (and there’s something to be said for requiring commands to be registered rather than just treating any executable on the path as a command).

And yes, I know about OpenVMS, but it’s not really a player in any meaningful way.

And a bit of VMS lives on in its bastard child, Windows NT.