I don't know anything about the Oracle module but using Pike on a SMP machine is useless since there is a global interpret lock in Pike so it can only use one CPU at a time.

Not quite true. Much threaded pike-code runs noticeably faster on a SMP machine.

The reason is somewhat obvious: The lock is released for most operations that take significant time (such as image operations, database queries, I/O etc)

Also, having an extra CPU for any databases and other external programs, and the system itself for that matter, really helps a lot.