[mythtvnz] Pointers on how to track down machine lockup

[Solor Vox]

On Mon, October 11, 2010 16:38, Wade Maxfield wrote:
> There was a break in the recording schedule  at lunch, so I've just
> swapped that in and will see how that goes over the next week. It's about
> 4 years old and at idle the machine now registers as using 134W (up from
> 122W that it used with the new PSU).

Each PSU has different efficiency rating.  It's very possible newer one
had a higher rating and that's why you're getting a higher usage for the
same system.

> You got me there. I haven't found a way to isolate that, other than trying
> a different PSU but leaving the rest of the machine the same.

You can get nice PSU testers that check all the rails, voltage levels and
more,  but I wouldn't recommend buying one for a one-off.  Best thing is
just swap with another as you're doing.

>
> Which components are we trying to stress test here? I've taken the new PSU
> out so should I be running these tests on the new machine with old PSU, or
> the old machine with the new PSU?
>
> I've done a quick 5 minute cpuburn on the new machine/old PSU.  8 copies
> of burnMMX resulted in CPU core temps quickly jumping from 24-28° to
> 43-46°. After 5 minutes they were still at 43-46°. Power draw went from
> 134W to 196W.

The point is to put as much load as possible on the system and make any
weak part(s) fail.  Then you reduce/remove things until you locate the
weak part causing the failure.  In this case, I suspect the PSU.  Putting
the max load on the system means it pulls the most load from the PSU.  If
you are able to run burn-in software with a different PSU, then it would
be a good sign your other PSU is bad.

sV