27 October 2004, 01:52 by mark hoekstra

It’s a new webserver!(2)

Well… I’ve done quite some work on my little webserver today. The SGI-SCSI-card didn’t work out (gave me strange errors and thought I had -8876MB on the disk (yes, that’s a minus)). Anyway, I threw in a Symbios-card and that one works flawlessly, so hardware-wise it’s finished. After that, installing NetBSD was a breeze and I’m compiling packages since.

bluedish.thumb-3.jpg

some pics

So… now Apache and PHP is running… how fast is it?

Well, an earlier non-scientific calculation of me guessed the Duron 1300 would be about 5.5-7.7 times as fast as the R5000 which is in my current webserver… time to get the facts! Bring in Bogoiips :-)

bogoiips-test on the Challenge S

bogoiips-test on Blue Dish

bogoiips-test on my Casio-calculator (no joke, well it is, but it’s real :-))

So:
Casio calculator = 23.2
Challenge S R5000SC = +/- 130,563
Duron 1300 on NetBSD = +/- 1,282,544

For anyone with a webserver that can handle PHP, here’s the source of this benchmark(ughe ughe)

bogoiips is nothing more than counting from 1 to 10,000 as fast as possible… even humans could be benchmarked this way!

permalink - add to del.icio.us

  1. Lothar @ 28 October 2004, 22:53 :
    Weird, my Athlon 1400 tops at 708,928. Could this be due to different processor integer performances ?


  2. mark @ 28 October 2004, 22:57 :
    Really? that’s strange indeed… Cause an Athlon TBird 1400 is a better/faster CPU than the Duron 1300 on every aspect, more L2, higher FSB etc… On what OS do you run this?


  3. Lothar @ 28 October 2004, 23:14 :
    Linux-2.4.27, php 5.0.2, apache 2.0.51. I haven’t looked into any optimizations though. Just plain vanilla installs. Just tested php from the cli, slightly faster, 774,228.


  4. Lothar @ 28 October 2004, 23:51 :
    Never mind, just did a cat /proc/cpuinfo.
    cpu MHz : 1000.004
    I had a power outage last monday. Apparently the bios has changed back to default settings.


  5. mark @ 29 October 2004, 01:43 :
    allright, but still on the low side I guess… even if we would ‘correct’ for the difference in MHz and multiply your result by 1.3, you would have 1,006,496 … against the 1,282,000 the Duron does (NetBSD 1.6.2, no optimizations, but I did compile PHP&apache on the box instead of binaries). I do run PHP 4.3.9 …

    http://mark.is-a-geek.org:85/testje.php


  6. Lothar @ 14 November 2004, 20:52 :
    Indeed, maybe PHP5 is somehow slower. I did try increasing the loop to see if the speed is affected by the new class handling, constructors and destructors etc, but the results seems to scale. At 100,000 it’s around 78000. Anyway, it’s not like this is some heavy-loaded LAMP server. It only serves static content. Actually I’m quite happy with it now, running at 1000 MHZ. The random crashes are gone :-D


previous: running Mac OS X on a 68k :-)

next: and yet another system