To start down a road of (somewhat) structured testing based on what I've heard
from all of you I ran a couple of tests.
Video seems to be the key.
First I took the path of let's break it. I decided to do some video intensive
and also processor intensive work. To that end:
4 lucifer sessions (the first 4 examples from the readme)
setiathome
`xmatrix -root` (big eye candy on the background of the display)
configured KDE to randomly change the backgrounds on 6 windows every 60
seconds
It took less than an hour to freeze the machine.
Then to test system load without X I:
same lucifer script
setiathome - two instances
a script that tars up the /usr tree to a file and then erases it and starts
over again
a script that does make clean; make dep ; make bzImage ; make modules
repeat
a crontab that mails the tail of the logfile, an uptime, and `top n 1` to
another machine every 5 minutes
and top on a virtual display to keep an eye on things.
So lots of disk and processor activity. Load average is hovering between 9 and
10. Test has run for over 10 hours now.
While it's a little early to say that I've got a stable system, the high load
going for over 10 hours indicates the direction I'm going in is correct.
Next steps:
let the high load continue to run for another 24 hours or until it freezes.
repeat the first test without the seti. Perhaps modify it for more
intensive X usage by running 5 terminals of hacks from jwz's xscreensaver
package (they're stand alone programs that will do that).
find some screen intensive SVGALIB programs and runs graphics w/out X.
Assuming that heavy X will cause a lockup, try graphics without X to draw a
distinction between the X system and the AGP video.
find a PCI based video card and rerun X test, Is the problem in the system
<=> AGP bus?
Any suggestions are welcome.
--
Michael Rasmussen - [EMAIL PROTECTED]
Be Appropriate && Follow Your Curiosity
email to pager: [EMAIL PROTECTED]
--
=- To unsubscribe, email [EMAIL PROTECTED] with the -=
=- body of "unsubscribe linux-abit". -=
