Remember speed will depend on protein size and other factors, as well as on the final algorithm we choose. So lets not concern ourselves with slow computers just yet.
As Ive mentioned before, we have a good pile of P3-450 MHz's locally, so I can assure you the minimum requirements will not exceed this at least, in the foreseeable future.
It sounds like most people are OK but a few are having strange problems. For those with problems, please try to describe in as much detail as possible step-by-step what you did to encounter the error, and post the exact error message (from screen or error.log) if you haven't already.