Brooks (brooksmoses) wrote,

  • Mood:

Some ponderings on computer speeds.

For various random reasons, I recently ended up running a small benchmark of sorts -- just timing one of my research codes that takes a minute or so to run -- on various computers that I use. The results, for what it's worth, are thus:
Office computer:      36 seconds
"New" home computer: 115 seconds
"Old" home computer: 696 seconds
A friend's computer:  31 seconds
Those break down roughly as one would expect by age, with only a little lumpiness. The "Old" home computer was built in 1997 (a 200MHz PPro); the new one was built a month ago, but from surplus parts of roughly 2001-vintage (1.1GHz P3). The office computer is a couple of years old (2.4GHz P4), and my friend's computer is essentially new (2.15GHz Athlon). This raises the interesting observation that the P4 is nearly twice as fast, per MHz, as the PPro; I'm guessing that most of that is a matter of cache size, since this code doesn't have any processor-specific optimizations that would take advantage of most of the other reasons I'd expect the P4 to be faster -- the code takes about 5MB when running, so it's doing cache swap, but probably lots more on the smaller caches.

The interesting thing, though, is what those benchmarks actually mean. On one hand, they mean that computers have increased in speed by more than an order of magnitude since I started working on my Ph.D., and most of that time I've been working on the same program. The possibilities for what I can do have changed dramatically since I started, if I'm talking about number crunching. This particular program reads in a data file with over 60,000 numbers in it, and spits out a data file with about 9000 numbers in it, and every number in the output is a sum over most of the numbers in the input, with lots of trigonometry thrown in for good measure. That's a frightening amount of math to be doing in 36 seconds.

But, really, the two "slow" computers on that list aren't computers I use for number crunching. They're computers I use for everyday stuff -- reading news, running LaTeX, editing code, basic image-editing sometimes, and that sort of thing. And the main reason I have for retiring the older computer in favor of the newer one, aside from occasionally doing number-crunching on them while I'm finishing up my dissertation, is simply that it's maxed out on memory and I can't add any more, and when I have more than 30 or so Opera tabs, two newsgroup windows, email, a text editor, and a few other random things open, it starts having a few minor problems. As for speed on loading programs and nearly everything else (aside from a few badly-written things) that's not intense number-crunching, it simply isn't noticably slower than my office machine.

And that, I think, is the most interesting result: in my actual everyday practice other than research-related number crunching, that factor-of-twenty difference in the benchmarks amounts to a nearly negligible difference in the usability of the computers in question.

  • Post a new comment


    default userpic

    Your reply will be screened

    Your IP address will be recorded 

    When you submit the form an invisible reCAPTCHA check will be performed.
    You must follow the Privacy Policy and Google Terms of use.