> The major updates, like AGA, were too little, too late.
And AGA was a mixed bag. The extra bitplanes were really welcome, but not having chunky (1 byte per pixel) mode when all the 3d coming out really required it, and having to do an expensive operation to go from chunky to planar, did really hurt efficiency.
It was a great addition that extended the existing idea of bitplanes, which was a really good one in lots of ways though.
I disagree that it needed a chunky pixel mode most of all. What you're asking for is a machine that can draw a scene byte by byte with the CPU, then just display that. But if your hypothetical Amiga is doing most of its graphical manipulation with the CPU, then it has failed as a platform. The main idea of the Amiga is to handle media data using specialized chips that are much faster than the CPU for certain tasks.
What an upgraded Amiga really needed was two things. The first is a fast blitter that could also horizontally stretch or shrink a bitmap by some fractional amount. The second was some sort of “flipper” device (or new blitter feature) that could reflect a bitmap across a diagonal line (or rotate by 90 degrees).
Here's how you'd use these for a third-person shooter. Store the wall bitmaps flipped along the diagonal; each line of those bitmaps correspond to a vertical slice of the wall. For each vertical line in the scene, find the correct wall tile and row, and blit that line of pixels into a scratch space, squishing it and shifting it by the correct amount. Then use the flipper to copy that to the screen.
Yeah, as you mentioned before, the chipset just didn't keep up. I had an A3000 for a while (68030 25mhz) You could do faster blits with the CPU. I remember running a utility called "cpublit" that accelerated window scrolling. It was noticeably faster. This was 1991 or so. ECS was long in the tooth.
The problem with AGA is it arrived when 386 clones with SVGA were becoming incredibly affordable. If it had arrived in early 1990 instead of late 1992, the Amiga might've had more of a chance.
Basically, Commodore should've skipped ECS entirely. ECS was essentially useless to most consumers.
Actually, Commodore should have kept their chips up-to-date with process improvements. That means releasing a new chipset with four times the performance every three or four years. So the AGA should have arrived in 1988, and the blitter should have been four times faster as original, not just twice as fast.
The Amiga 2000 should have been delayed until 1998 to include the new AGA. Keeping the cheaper Amiga 500 on the OCS would have been fine.
And AGA was a mixed bag. The extra bitplanes were really welcome, but not having chunky (1 byte per pixel) mode when all the 3d coming out really required it, and having to do an expensive operation to go from chunky to planar, did really hurt efficiency.
It was a great addition that extended the existing idea of bitplanes, which was a really good one in lots of ways though.