Stuff affecting the point of simulation reaches the point of simulation at light-speed; e.g. the gravity waves of distant stars. So, at the time at which the gravity from a star would have reached the point of simulation, that star has to exist within the simulated area - i.e. a lightsphere surrounding the point of simulation. So, yeah, I guess if you like, the computer needs to be larger in physical size than a sphere of radius t x c where t is the expected duration of the simulation. That's still, uh, pretty big.

I think. I'm not an expert!