my recommended configuration for >= 8-cores

[alpha754293]

So, recapitulating:

if you want the fastest WU speed, go with a single "-smp 8" client. You may periodically be using half the system for a few days at a time (if you don't want to babysit it), otherwise, you'd have to manually start a second "-smp 4" client if it picks up an a1 WU.

For the most PPD, and most WU done within a given time period (depending on your processor speed, for me; because my procs were 2.4 GHz, so I use a 3-4 day window); go with two "-smp 4" clients. The WU speed will be slower, but you'd be able to do more "work" in the same period of time, and it's a much more "hands-off" approach for managing the system so you can just periodically check in on it using FahMon or something to make sure that the system is still running.