> Also, the much-beloved 68000 series processors at the heart of the Amiga and Atari ST were ultimately doomed by the combination of being made by Motorola and being, perhaps, the ultimate expression of CISC architecture (which made them fun to program by hand in assembly language).

I agree that the issues w/ m68k-series processors were an underappreciated factor behind the demise of the Amiga and Atari ST computers. The Macintosh line managed to transition to PowerPC, but they did so via software emulation. That would've been highly problematic for the kind of software (games and early multimedia - not DTP, which was the Macintosh domain) where the Amiga and Atari ST were at their strongest. Other m68k machines in common use included, e.g. Sun and NeXT workstations which would've been running highly portable code without 'tight' performance requirements.

> The Macintosh line managed to transition to PowerPC, but they did so via software emulation. That would've been highly problematic for the kind of software (games and early multimedia - not DTP, which was the Macintosh domain) where the Amiga and Atari ST were at their strongest.

This is an excellent point and one I hadn't fully appreciated until reading your comment. Most of the popular 68K software on Macs (such as DTP) was more amenable, or at least tolerant, of running under emulation. Even the popular games on Mac like Myst weren't as real-time critical as popular Amiga and Atari ST games which tended more toward arcade style and sometimes even accurate arcade ports. While I'm sure there were arcade style games for color 68K Macs, they weren't the majority. Also, because the Mac didn't have so many tightly integrated custom co-processors, my sense is that Mac 68K software wasn't as tightly hardware coupled and counting on specific timing interactions. A fair amount of Amiga software would read and write directly to hardware registers instead of using OS calls and even if it only used OS routines, it could still be highly dependent on precise behavior. Once again, we see that aspects which had made the Amiga and Atari ST great in the 80s, made it harder to navigate the transitions necessary to survive the 90s.

It would be interesting to have modern emulator authors compare notes about the software libraries between these platforms. While I'm sure Mac emulator authors still found a lot of 'misbehaving' Mac apps to deal with, my sense is the Amiga software library was bonkers to support in emulation. The WinUAE Amiga emulator has long had a precisely cycle-accurate mode which is less performant but simply necessary in many cases. And as mature as WinUAE is, the team is still discovering edge cases where 40 year-old games and graphics apps have never been emulated correctly.

Conversely, I remember around 1992 I bought a Macintosh emulator for my 68020 Amiga and it performed quite well. I used it for work to run Mac DTP applications. The emulator ran in software but used a small hardware dongle on the Amiga's parallel port to import original Mac ROMs which you needed to buy separately. Of course, both the emulation source and target were 68K-based but it indicates that most Mac software was reasonably well-behaved in terms of hardware dependence. If a little Amiga startup was able to write a pretty good Mac emulator, it was certainly possible for Apple themselves to it better as few years later with a much faster PowerPC CPU.

Finally, it's clear that post 1990 both Atari and Commodore were in increasingly weaker positions, not only financially but in terms of staff depth. While both still had some remarkably talented engineers, the bench wasn't deep. I know that at least at Commodore, toward the end they'd canceled their much improved, new Amiga chipset project (AAA). Even though it was almost complete with (mostly) working test silicon on prototype boards, they canceled it because it had become obvious future Pentium and RISC CPUs would outperform even the 68060 and AAA custom chips. At the time Commodore folded engineering was working on the 'Hombre', an entirely new design which would have been based on an HP RISC CPU. For graphics the main thrust would have been new retargetable graphics modes for hi-res, high-frequency monitors (1280 x 1024). https://en.wikipedia.org/wiki/Amiga_Hombre_chipset

The plan was to support legacy Amiga software with a 68K emulator on the RISC CPU driving a new chip created specifically to support legacy Amiga graphics modes. When I later read this, I remember being quite skeptical that hybrid software/hardware emulation would have worked very well for the eclectic Amiga library of legacy software. As much as I loved the Amiga, the OS stack could then only be described as 'crufty'. It had been upgraded a little over the years but still contained major legacy components from different eras and many of the people involved were no longer at Commodore. Given that reality, the plan had been to base the new Amiga on Windows NT.

But - even if Commodore somehow overcame the myriad technical challenges, lack of resources and depleted talent bench, once a next-gen Amiga isn't based on the 68K, AmigaOS or the custom chips and boots Windows NT in XGA mode - is it still really an Amiga? Certainly, at least some of my software wouldn't have worked anymore so, facing the decision to buy a new, quite different computer, why wouldn't users also look at the, probably, cheaper Packard Bell Pentium running Windows 95 down at Costco? After all, with the Pentium and Windows 95, the PC juggernaut had finally coalesced into a coherent whole that could be compelling to both home users and graphics, gaming, multimedia obsessed hobbyists. And new Doom/Quake quality games were coming out almost weekly. That's when even I bought a PC and started using it as my main daily driver. Of course, I kept my awesome, fully loaded, tweaked out, much beloved Amiga system on the desk alongside it for a couple years. But web pages never quite looked right on the Amiga and sharing files on the network was hardly seamless. Sadly, it was increasingly clear the world had moved on. In many ways the Amiga (and other notable platforms of the 80s) had blazed the trail showing the way to the future - but it was a future they would not be a part of.