A while ago, I posted a graphics-dependent benchmark that did a lot of floating-point operations. I now have a small benchmark that, while running for considerably longer for me at least, should depend only on CPU and memory system performance. <b>It's <A HREF="http://us.f2.yahoofs.com/bc/775690c7/bc/fotos/bench11.z..." target="_new">HERE</A></b> (unfortunately, the file params.txt has to go with the .exe), and those of you who have any interest in comparing our systems can download it and execute (no, it will not cause hard drive disintegration or anything... regretably).
It does some integer math, but it's mostly floating point. And it is a ridiculous amount of FP operations.
What it's actually doing is expanding a two-dimensional wave in eigenfunctions of some differential equation and then drawing a two-dimensional projection which requires reevaluating the expanded wave function a lot of times... It's not an expertly-written program for doing that which it
does, which is why it's slow. (which is what makes it a reasonable benchmark!)... It is currently using only 25 harmonics of a square infinite potential well; the most consuming task here is the projection, which is not shown on screen and is fixed at 800x600 resolution and is done fully in software.
My time in it is 85-90 seconds, with a P3 933Mhz with 256MB of PC-133 SDRAM... I also tried a P4 1.7Ghz (willamette...) with 256MB of PC-800 RDRAM, for a total time of about 100 seconds while still running some background tasks.
<i><font color=red>You never change the existing reality by fighting it. Instead, create a new model that makes the old one obsolete</font color=red> - Buckminster Fuller </i>