[mythtv] Myth box stability problem resolved (hangs, lockups)

[Colin Panisset]

On Saturday 25 January 2003 03:48, Henk Poley spake thus:
> > Van: Colin Panisset <mythtv at clabber.com>
> >
> > It turned out to be an L2 cache problem. Turning off the L2 cache ECC
> > checking in the BIOS has fixed the problem, and the box appears solid
> > now.
>
> Huh, you better return your CPU and get a replacement. The normal
> behaviour to ECC errors is halting the machine directly (that's the hangs
> you report). So unless your motherboard -though I guess the CPU does ECC
> by itself- is broken this means that memory degredation (bit toggling) in
> your L2 cache occure. Which can cause all sorts of horrible things inside
> programs when not detected/corrected by ECC...
>
> I hope you understand this shouldn't happen and that your CPU is actually
> broken then...

Actually, it looks now like it's the motherboard. I changed the CPU (a P3 
1GHz) for a Celeron 800 I had in another machine; swapped the RAM over as 
well. The box now hangs partway through booting linux, and not always at 
the same place :(

If only I could get a replacement that'd actually fit inside the shuttle 
box, but I can't find anywhere that sells FV25's, and the box itself has an 
FV24 inside it now.

I have to admit that I'm now kind of leery of Via's 686b South Bridge 
chipset, since there are known data corruption problems associated with it. 
This machine in particular has had random mysterious lockup problems, even 
when just running memtest86, and they've been getting worse of late.

I think it's time to dump the proprietary form factor, and go for a case 
that won't look too out of place (or can hide in the closet) with a 
conventional motherboard.

Dammit.

  -- C.