It all depends on what kinds of sound you produce. I regularly overload my dual 2 GHz athlon box during performances making only sounds. But those are all synthesized on the fly. For sample playback you need a lot less CPU so if you want a lot of simple voices more or less out of sync a couple a slow computers will do the trick.
Yikes, with my limited knowledge of sythesis, I don't think I've ever come close to overloading my single Athlon 1.7ghz. This may be a consideration to take into account as the performance may use more complex synthesis than I have been using. But even a dual Athlon like yours is out of range for this project. I'm sure, however, that I'll be able to find some kind of work around, if only by avoiding complex sythesis.
In a studio environment yes. A reasonably quiet of-the-shelf computer (not the cheapest allready) I have found not to interfere in a theater setting, with all the lights and aircondition etc. But in a recording studio with nice mics you really notice them. It all depends again on the kinds of sound manipulation you do and in what kind of environment. A noisy PA for instance will mask more than a high end one. I you have the good mic and you want to have it take samples, so you are gating the input, you will notice noisy input a lot sooner than if you only amplify it since noise becomes more noticeable when it is interupted. That can be nice or not depending on your esthetics.
This is what I expected. In this case I don't think I will worry too much about quietness, as the audience's distance from the machinery, the ambient noise, and the noise of the performance itself will drown out computer noise. Also I probably won't be taking straight samples and rebroadcasting them, so ambient noise on the mics shouldn't be too big a problem. Rather, I intended to use mics as a poor man's sensor, whereby if the analyzed input to the mic has certain characteristics, another event will be triggered. Noise could easily be canceled out by calibrating the analysis correctly.
Thanks.
-ian