[Date Index] [Thread Index] [Author Index]

Re: MacOS vs. Windows PC for Mathematica?

• To: mathgroup at smc.vnet.net
• Subject: [mg13930] Re: MacOS vs. Windows PC for Mathematica?
• From: Harald Giese <giese at dkrz.de>
• Date: Tue, 8 Sep 1998 02:52:29 -0400
• Organization: Institut fuer Meereskunde, Universitaet Hamburg
• References: <6s5l6q$c84@smc.vnet.net> <6s8ed9$ijr@smc.vnet.net> <6ssus0$m83@smc.vnet.net>
• Sender: owner-wri-mathgroup at wolfram.com

Josip Loncaric wrote:
> ...
> Why this inconsistency?  The benchmarks are rather different, and
> Stefan's requires more memory to run.  Could this be the main source of
> NT's lackluster performance?  There is a bug fix for "Out of memory"
> errors reported by Mathematica 3.0 and 3.0.1 under NT (see
> http://www.wolfram.com/support/Systems/Windows/OutOfMemory.html page).
> Were Stefan's results affected by this in any way?
> ...

After upgrading the CPU from P133 to P200MMX recently I re-ran Karl's
and Stefan's tests. 

Here the ratios:
CPU:	200 MHz / 133 MHz = 1.5
Karl:	1.15 / 0.706 = 1.6
Stefan:	902 s / 599 s = 1.5

I assume the better ratio with Karl's test to be due to the twice as big
L1-caches of MMX CPUs.

The memory bug fix shows no significant differences in the performance
of the tests - except that Stefan's test now can be run in one.

-- 
Harald Giese
Email: giese at dkrz.de
Phone: +49 (0)40 4123 5796; Fax: +49 (0)40 5605724 Institut fuer
Meereskunde der Universitaet Hamburg (Institute of Oceanography of the
University of Hamburg) Troplowitzstrasse 7, D-22529 Hamburg