We migrated a single thread, batch application from an RS6000 R21
(Uniprocessor) to an RS6000 R50 (4-way processor).  The uni-processor R21
had more power than the sum of the R50 multi-processors.  But we found that
the R50 ran the same batch application in 53% of the time.  We assumed that
the increase in performance, even with a single thread application, was due
to the operating system overhead being spread out over the R50's smaller
processors.  Whatever the reason, the performance gain was real.

In light of this, I would assume that ADSM would have at least the same
performance gain, assuming that an I/O bottle neck does not exist.