testing hardware - use what software ?

[Alan Cox]

On Fri, 23 May 2008 23:35:31 +1000
David Timms <dtimms at iinet.net.au> wrote:

> I have a PC that is failing intermittently. What are the best choice for 
> attempting to pinpoint a hardware fault - in terms of either fedora 
> packaged or other open sourcesoftware ?
> 
> Perhaps there is something that can both perform one of and 'to death' 
> tests of various PC components, and components in combination, and also 
> provide peak workload for all subsystems at once, in an attempt to 
> generate a slow to occur failure.
> 
> I know about memtest, and read-only disk tests from fsck. Any others ?

You can certainly stress bits of the system - eg boot into text mode and
run dbench tests to trash the disk subsystem.

I actually usually start from the hardware end with a misbehaving PC

- Reseat all the cards
- Remove and reinsert all the RAM (taking care to earth yourself first)
- Blow dust out of the fans
- Optionally unplug/replug all the cables although that is rarely a
problem

Run it with the lid off for a bit and see if anything seems oddly hot

memtest86 and smart reporting from the drive are good useful starters as
a positive bad result saves a lot of guessing. Its often hard to pin down
a hardware fault however that is more complex or subtle (or to tell it
from software)

Alan