VMWare overhead was about 30% when I tested it.
The spooling process is almost 100% CPU.
Typical spool job goes for 1 to 5 minutes.
I run up to 3 at the same time, and I might have hundreds in the queue.
Right now I am running on a dual opteron, but can put another 2 CPUs if I need to.
I am unwilling to sacrifice the CPU unless I REALLY have to.