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

Re: Mma speed trials (Win'95)

• To: mathgroup at smc.vnet.net
• Subject: [mg2866] Re: Mma speed trials (Win'95)
• From: Ed Oswalt <efo at RDCS.kodak.com>
• Date: Sat, 30 Dec 1995 01:02:16 -0500
• Organization: Eastman Kodak - any opinions expressed are my own...

Richard Morey Porter <rmporter at unixg.ubc.ca> wrote:
>Hello all
>
>Happy Yule!
>
>Had a few spare moments, so I thought I'd do the speed trials on my =
>machine.  It would appear to be underperforming (or at best performing =
>inconsistently). Is this my imagination?  Am I merely quibbling over =
>nanoseconds?  I thought 16MB RAM would be enough, but my P-100 seems to =
>be outpaced by lower speed 486s running on 16MB.  Is Win'95 vs. MMa =
>front-end really slowing things down this much, that I am effectivly =
>running a 486 between 33 and 75 MHz?
>
>Anyone else experiencing a similar wholesale drop in performance?
>
>Machine Specs:
>
>Pentium, 100 MHz, 16MB RAM, Win'95 (Monitor: 800x600, 256, 1MB; MMa =
>2.2.4 student version)
>
< timings - snipped to save space >

I did see a performance drop with Mathematica 2.2 Student when I went from 
Windows 3.1 to Win 95 on a P-90 with 8Mb of RAM of 5%. 

It could be worse....
I only bought MMa a few months ago but it appears my school sold me
a old version of the student edition with NO FLOATING POINT (and of
course MATHEMATICA did not tell me that or offer the new 2.2.4 version
when I registered the product). I mention that because if you think your
numbers are slow wait till you run the speed tests without floating point!

For example I get 440 seconds verses your 27 seconds for the Eingenvalues
 timing test (a heavy Floating point computation). 

On interger operations I perform about the same as you....

But back to 3.1 verses Win 95 , I was suprized to see a 5% or so performance
drop (I don't have the results handy at the moment) - I expected Win 95
to be faster based on what MicroSoft said - not sure why but it may be
my Win 95 install is less than optimial...

I still need to tune my system for Win 95 and beat on Mathematica folks for letting
me get stuck with a non floating point version when the floating point was
available...

Ed Oswalt

A student again at RIT....