[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.