Computational chemistry on RS/6000?

[Christopher Gunn]

We've just started playing with an RS/6000 server model and some
of our local computational chemistry codes and so far have been
pleased--we're getting benchmarks of 30-31 microVUPs, vs. ~24
'official' SPECmarks for the 320 (520?) model in question.

We're begun looking into acquisition of one or more larger
RS/6000 units and I was seeking experience and feedback concerning
system performance in computational chemistry applications.
Particularly interesting to me:

a.  What does performance look like across spectrum of xy0 models
where y = 2, 3, 4, 5?  Does real-world performance speedup look
like advertised Linpack ratios (550/520 = 2.7) or SPECmarks
(550/520 = 2.2)?

b.  What is implementation (and performance) like for any of the
larger commercial packages ported to RS/6000?  We're a Sybyl and
Macromodel shop, primarily, and apparently there aren't live RS/6000
versions of those yet, but I'm hoping some sort of cross-platform
comparisons are available:  Biosym, Polygen, whatever?

I hope this isn't a dreary or inappropriate topic for comp.unix.aix--
I just discovered a couple of days ago that such was the home of
the RS/6000ites, and don't have a good feel for the group yet.
Appreciate any help or guidance.

Christopher Gunn	Molecular Graphics and Modeling Lab
SPAN--KUPHSX::GUNN	Department of Medicinal Chemistry, Malott Hall
913-864-4428 or -4495	University of Kansas, Lawrence, KS  66045