(By comparison, look at modern graphics cards. You can buy the super-duper latest card for £600, or the 'ultra cut down version, don't bother cos it'll be slower than what you already have' for £100 - the only problem with AGA is that we were only given a cost-saving version, not the super duper version.)
Oh, don't worry about this: when AGA came out, game publishers still targeted to 1 MB expanded A500 models, thinking this way they could reach the widest possible audience for the Amiga version of their games. So we may have gotten two version of the AGA chips, but games would have barely used them, throwing away any need for an upgrade.
Then, at the end of the 80s and beginning of 90s the "power of the GPU" was a total nonsense: it didn't exist, since you measured the power of a graphic chip by the widest available video mode. You could do 1600x1200? Well! You were more powerful than the one that did 1280x1024 only! And don't forget the Amiga architecture in general was years above the PC one, that started improving only when it found a way to keep backward compatibility with older software, virtualizing where virtualizable (and this process started with Windows 95, much more before than actual virtual machines existed). A faster AGA wuould have been just a faster blitter/copper thing, but nothing revolutionary as, for instance, the first 3dfx VooDoo chip was.
After all these years, I am quite disappointed by this AGA lovers/haters debate, since at the time we were all of the idea that hardware should have been used at 100% of their potential or even more, like some magic demo coders did on the C64 and the original Amiga models. It's good someone noticed we were wrong but please, stop this nonsense because it's ridiculous.
