Re: PowerPCs

*To*: mathgroup at yoda.physics.unc.edu*Subject*: Re: PowerPCs*From*: Richard Mercer <richard at rmercer.wright.edu>*Date*: Thu, 07 Jul 1994 11:34:35 -0700

Truth Is Stranger Than Fiction, Part III >From a previous message of mine: > An example of what I am trying to avoid is on p.5 of the 5.30.94 MacWeek, > where a "Timing" result of ListPlot is given as 4.48 seconds on a PowerMac, > which does not include the considerable time for rendering the Postscript > graphic. This is claimed to take 27.3 seconds on a Quadra 700, a result which > almost certainly is total elapsed time. The actual command in question, printed in MacWeek, is Timing[ListPlot[Table[Prime[i],{i,10000}]], PlotJoined->True]] I now have actual facts, and they are weird. The Timing command actually does give results between 4.5 and 5.0 seconds on a PowerMac 7100AV, so 4.48 is legit. The actual elapsed time is between 11 and 12 seconds, due to the time required for Postscript rendering of 10,000 points and 9,999 line segments. (Why anyone felt it necessary to join these points is beyond me.) On an actual Quadra 700, the Timing results were clustered around 24 seconds, however the actual elapsed time was around 29-30 seconds. The weird thing is that the Postscript rendering of the graph began consistently started after 13-14 seconds. Thus it would seem the kernel calculations were completed in at least 10 seconds less than the Timing command would indicate. Conclusions: (1) The published timings are legitimate and represented no attempt to deceive. (2) The ratio of the Timing results (roughly 6) is a wildly inaccurate representation of the actual relative speed of the two machines. In total elapsed time the PowerMac was between 2.5 and 3 times faster than the Quadra 700. It may have been 3-4 times faster in performing the kernel calculations. (The PowerMac seemed to begin Postscript rendering after 3-4 seconds, slightly less than the Timing results.) (3) The claim in that same MacWeek article that "Mathematica performs numerical operations up to 10 times faster on a PowerMac than on any 680x0-based machine and renders graphics up to 5 times faster" is at best highly misleading, and probably cannot be substantiated by accurate timings. I stand by my earlier claim that Mathematica on the PowerMac is 2-4 times faster than on Quadras (well, maybe 5 times faster than a Quadra 605!). Richard Mercer