> In the past I've found Map24 acts very differently on different machines. The
> JVM can change things as well, e.g. the SUN JVM ran it mind-bogglinginly slowly
> but the MSJVM ran it very fast (60fps almost).
>
> On some machines neither runs it fast.
>
> Related maybe?
>

Maybe.
My experience is that nowadays the Sun JVM is almost always faster than the MSVM (especially the server VM), but I've come around a few cases in which the Sun JVM can perform really, really slow where the MSVM performs a lot faster in comparison. In those cases I tried to do some optimizations that are just *not done* in java, like (manually) inlining code or other micro optimizations I would do in C. The Sun JVM doesn't like that and can end up interpreting all code that should be jitted. You just have to depend that the Sun VM will do all those optimizations for you, and probably (hopefully) better than you could do yourself. If you write clean, well encapsulated code, the Sun JVM should perform very well in almost all cases, up to speeds comparable to C++. Which is kinda nice :-).
But of course there are still other cases the Sun VM is not very good at (like fast trigonometry for example).
In this case however, I doubt it. It's all just simple stuff going on and that should work reasonably well, even on old CPU's. I still think that it might be an openGL issue. Maybe for some reason it selects a software openGL mode for example. I've seen CottAGE (which is still the core of this emu) perform pretty well on a K6-III.
Using the server VM might make a big difference though (lot's of array access and the default client VM doesn't do bounds check elimination).

BTW, what's Map24?