[hakfoo]

Isn't this the classic Wheel-of-Reincarnation cycle?

The Amiga's dedicated coprocessors let it do things an IBM AT or Mac 512k couldn't do easily.

But the AT evolved into the 386 and 486, and suddenly you no longer needed coprocessors to do the heavy lifting.

So you end up with two issues:

1. Did the Amiga market keep pace CPU-wise? I can see the 3000 and 4000 shipped with better CPUs, but how much was the software ecosystem defined by the 500/600/1200? I wonder if this created a tarpit for developers: the Amiga might have been able to do the same tricks as a faster PC, but the development effort was much higher than just recompiling brute-force x86 code to 68000.

2. Where were the custom coprocessors going? The original Amiga chipset was good, and I'm sure the ECS/AGA stuff was even better, but it seems like it was close to "at par" to a PC with a SoundBlaster and a basic Super-VGA chipset by the early 1990s. What could they use to create a new defensible territory? I could imagine a 3-D coprocessor, but designing something that both impressed today and didn't immediately obsolesce is hard-- see the corpses of a dozen early accelerator vendors.

[icedchai]

The Amiga stagnated for years. The custom chips barely changed from 1985 until late 1992. ECS barely added anything over OCS (it had a couple new video modes nobody used.) As I mentioned in another thread, AGA was too little, too late. By the time AGA was out, 386 systems with SVGA and SoundBlasters were cheap and plentiful.

[DerekL]

Yes, if Commodore wanted to compete, the Amiga chips needed to keep up with Moore's Law. Every 18 to 24 months, the processing and display power should double or more. In 1988, the Amiga 500 and 2000 should have had something like AGA.

The article correctly points out that 68k chips fell behind the x86 series. Yes, chunky pixels are useful, and the display hardware should support them. But if the game requires the CPU to draw each pixel one by one, the hardware has already failed. Instead, there should be some specialized processors that do the work. In fact, that's what happened in PCs. Doing all of the drawing with the CPU lasted only about five to seven years, then everyone had GPUs.

[KerrAvon]

Yes! What killed the Amiga was that Commodore effectively stopped R&D on the chips in the mid-1980’s, and by the time they restarted they had already lost too much time.

(It’s possible to argue Steve Jobs was right when he dismissed the Amiga because it was too much hardware. He knew it would be difficult to keep evolving such an architecture. It’s also possible he was wrong because he didn’t account for Commodore’s chip design and manufacturing processes.)

In any case, by 1992, there were Macs capable of 24-bit color, and the 68040 was certainly capable of pushing enough pixels quickly to run Doom / Marathon / Duke 3D without hardware acceleration.

[flomo]

I remember someone saying "There was no market for 'Amiga games', there was a market for Amiga 500 games." If you look at it that way, it had a nice long run as a game machine, while failing in most other segments.

[icedchai]

That makes total sense. The A500 was the most popular machine, with a 68000, 512K RAM. If your game required 1 meg, you'd lose a decent number of potential customers. (My A500, circa 1989, was souped up! 3 megs of RAM and a hard drive!)

[flomo]

Meanwhile, there was a market for high-end PC games. If you bought a better PC, you could show it off. That was important to the type of person who owned an Amiga (they always wanted to show you).

[ekianjo]

Hardware designers had a better version than AGA in the works but Commodore scrapped it and instead went for the minor upgrade than AGA was

[zozbot234]

The early "3D coprocessors" weren't very 3D at all, they basically accelerated triangle rendering in screen pixel coordinates. So "2.5D" at most. I could definitely see some version of the Amiga shipping with something like that, leading to something not too unlike Sony-PS1 level graphics or so. But the real problem for the Amiga (and its nearest competitor, the Atari ST/TT) was that the 68k architecture was ultimately abandoned by Motorola, and at the time (with Moore's law in full swing, and thermal constraints not too important just yet) the PowerPC looked like the best alternative. Of course ARM was a thing already, and it even got used in a high-level game system (the 3DO). So you could surmise that we could've gotten an ARM based Amiga/Archimedes mix instead which would've kept some kind of "cheap home computer" market going for some time, trying to disrupt the costly PC and Apple Mac platforms at the low end.

[kazinator]

I would say that hardware which helps rasterize triangles in screen coordinates is squarely 3D if it does Z-buffering (or any other hidden pixel removal).

[KerrAvon]

The 68k architecture still had some runway by the time Commodore’s fate was sealed, though. Commodore really needed to be taping out the next generation chipset no later than 1990, and arguably 1988 would have been better.

[hakfoo]

I wonder if Motorola lost faith in the 68k because of a diminishing number of signature customers.

When Apple went for the PowerPC instead of continuing on to the 68060, the remaining audiences for high-end 68k were not going to move anywhere the same numbers.

If Commodore and Atari had remained competitive longer, there might have been more demand (and conversely, enough R&D effort to tide 68k over until we got to modern "everything is RISC after the decode stages" design paradigms.

[mardifoufs]

Stupid question: why didn't they replace the 6800 with the newer Intel chips? Was it "just" (huge understatement, I know) because they'd need to port or break software due to the new arch? I guess my question is if their multi coprocessor architecture could've worked with a better, stronger main processor like Intel's?

[zozbot234]

Commodore did have a line of IBM PC clones if that's what you're suggesting. Adopting x86 without the other elements of the PC architecture would've been pointless. The 386 would've made programming just about bearable for devs used to the m68k's flat address space and elegant from-scratch ISA design. It would not have been very successful.

[actionfromafar]

How can we know? There were many developers used to Amiga hardware. With an easy to way port games to an x86, who knows what could have happened?

[timbit42]

When Commodore died, Motorola's 680x0 series of CPUs were still competitive with Intel's x86, but Commodore would use the previous generation (or two) model to keep the price low. If you wanted the latest, fastest CPU, you had to buy an accelerator (eg. GVP). They should have had the top of the line processors available for those who needed or wanted them and could afford them.

A few years later, Motorola was failing to keep up, which is why Apple switched the Mac to PowerPC.

[rasz]

> Commodore would use the previous generation (or two)

Reason for picking 68000 was bargaining power you had when owning a chip Fab.

"Live with Dave Haynie - Commodore Business Machines C128, Plus4, Amiga" - BilHerd https://www.youtube.com/watch?v=6ZT209i-3Lo Dave reveals Commodore was paying $2.5 per 68000 Hitachi CPU compared to Apple $8 from Motorola.

_$2.5_ per CPU.