Re: Power Mac Performance
- To: mathgroup at christensen.cybernetics.net
- Subject: [mg256] Re: [mg247] Power Mac Performance
- From: Richard Mercer <richard at rmercer.wright.edu>
- Date: Thu, 01 Dec 1994 10:06:30 -0500
> MathGroupers,
>
> I had the opportunity to compare the performance of the
> native version of MMa (2.2.2) running on a Power Mac
> 7100/66 with that of a Quadra 700 running version 2.2.1.
> The notebook used represents a fairly good mix of symbolic
> manipulation, number crunching, and rendering of postcript
> plots (including 3D plots). The Power Mac has 24 MB of
> memory and virtual memory was turned off. The Quadra
> 700 has 20 MB of memory and virtual memory was also turned
> off. The results were as follows:
>
> Machine Elapsed Time --------
> ------------ 7100/66 230 seconds
>
> Quadra 700 1815 seconds
>
> The Power Mac was nearly 8 times faster than the Quadra.
>
> I also ran a simpler test with the following results:
>
> Quadra 700
>
> Timing[Inverse[Table[Random[],{100},{100}]];] {7.95
> Second, Null} Timing[N[Pi,500];] {0.55 Second, Null}
> Timing[10000!][[1]] 57.6833 Second
>
> Power Mac 7100/66
>
> Timing[Inverse[Table[Random[],{100},{100}]];] {0.966667
> Second, Null} Timing[N[Pi,500];] {0.05 Second, Null}
> Timing[10000!][[1]] 9.16667 Second
>
> I would be interested in seeing the results of this simple
> test from other platforms such as the Sparc workstations
> and the Pentium-based PC.
>
> Ed Boss NOAA/Pacific Marine Environmental Laboratory
>
There was a discussion on this issue several months ago in this group.
At that time I strongly cautioned against relying on the results of the
Timing command, which has some quirks, and insisted on using actual elapsed
time (as measured e.g. by the stopwatch feature on many inexpensive digital
watches).
One quirk that I observed was that the Timing command on a Quadra 700 often
included most (though not all) of the time used to render a graphic, which
often takes longer than the actual kernel calculations, especially in 3D plots.
However on the PowerMac the Timing command did not appear to include any
rendering time. This can lead to exaggerated results in favor of the PowerMac.
I specifically tested the Quadra 700 (because it was available) and found the
PowerMac 7100 to be 2.5 to 4 times as fast depending on the task. While you may
get higher figures for some specific tasks, I suggest you retime your test with
a stopwatch.
(I have a PowerMac 7100 at home and love it.)
Richard Mercer