Re: Power Macs
- To: mathgroup at yoda.physics.unc.edu
- Subject: Re: Power Macs
- From: "Andrew B. Watson" <beau at vision.arc.nasa.gov>
- Date: Fri, 1 Jul 1994 14:23:36 -0700
I hope I am not contributing to the noise, but my causal tests show:
Sparc 10:
In[2]:= Timing[Do[Gamma[1.1],{100000}]]
Out[2]= {18.2333 Second, Null}
PowerMac 8100
In[2]:= Timing[Do[Gamma[1.1],{100000}]]
Out[2]= {14.0833 Second, Null}
Both were run on a raw kernel, without front end. Very similar results
were obtained with:
tmp=AbsoluteTime[];Do[Gamma[1.1],{100000}];AbsoluteTime[]-tmp
which suggests that Timing is relatively trustworthy under these conditions.
The subject of benchmarking is notoriously slippery, and I am afraid
that misinformation (perhaps mine included) may unwittingly be dispersed to the
net. The slipperiness is compounded by the multiplicity of timing
methods, and whether or not one is using a Notebook Front End (I have
found this can make a substantial difference under some conditions).
Perhaps someone at WRI could offer a benchmark suite, and timing
method(s), so that we could all be talking the same language.
Andrew B. Watson
MS 262-2 NASA Ames Research Center
Moffett Field CA 94035-1000
(415) 604-5419 -3323 fax
beau at vision.arc.nasa.gov