kovacm, I think it's positive that you ask about what makes certain effect possible/fast/slow on A1200/Falcon. When it comes to demos, this is what most coders care about - regardless of platform they put their reputation on the line with each release - is the quality high enough? 3 months old code? baaah, I can make it faster now! etc
So for the optimizations, we're talking hard and intelligent software work and time put in by coders focused on a single task - to impress other coders. At least in the early 90s which is the era for the vanilla machines, but it's still partially true today, even if many just appreciate "a good show".
Just make sure you only post videos (or better, binaries for real vanilla Falcon030 or any cycle-exact emulator) if you intend to make comparisons - just so that others can validate any claims about how fast it runs. You absolutely have to know that it runs at this speed on real hardware. Any extra memory also gives an advantage, so you would have to know that.
That said, the A1200 was a 32-bit computer hampered by bitplane graphics, and the Falcon was still on 16/32-bit but with 16-bit chunky mode which lends itself to exactly the effects that were popular then, and it had a DSP. I don't think many would think it strange that this hardware advantage resulted in better demos, if that's the case. It has certainly been the story both before and since.