Mathematica integer vs floating point characteristics
- To: mathgroup at smc.vnet.net
- Subject: [mg4075] Mathematica integer vs floating point characteristics
- From: Phil McDonald <mcdonald at aix1.uottawa.ca>
- Date: Fri, 31 May 1996 01:34:25 -0400
- Organization: University of Ottawa
- Sender: owner-wri-mathgroup at wolfram.com
Hello,
We are currently operating an IBM RS/6000 model 390 that is
shared by researchers in our Math and Physics departments. This
system is running at virtually 100% cpu busy so we are now
planning to split the workload by acquiring a second system.
Since Mathematica is a significant part of this workload, how it
performs on different systems is of great interest to us.
Specifically, does Mathematica use integer or floating point
math to do its calculations. This is relevant because the
PowerPC line of RS/6000s gives very good integer performance but
lags the Power2 line in floating point performance, and the
PowerPC units typically have a lower price attached to them.
(eg: the PowerPC 604 based model C20 is rated at SPECint92 = 155
and SPECfp92 = 150.2, while the Power2 based model 390 is rated
at SPECint92 = 114.3 and SPECfp92 = 205.3. The base prices for
these systems in Canada are: C20 = $16,450, 390 = $32,760.)
Does anyone have any experience in this area? Has anyone
benchmarked Mathematica on different systems? I would also be
interested in info on other manufacturer's equipment, but it is
very unlikely that we will acquire something other than an IBM
RS/6000 unless there are compelling reasons to do so.
Any help that people can provide would be greatly appreciated.
P.S.: I do not use Mathematica myself, I just run the systems
that it is used on.
Regards,
Phil McDonald
Manager, Academic Informatics Systems
University of Ottawa
mcdonald at uottawa.ca
==== [MESSAGE SEPARATOR] ====