Depends what the forked code does. CPUs and memory as a rule are way faster than disk IO. If there's a lot of processing to be done, forking (or threading) means grabbing more processors. If you're IO bound, then it probably won't help much - any gains are offset against contention.