[vox-tech] When RAM goes bad...

[Rick Moen]

Quoting Peter Jay Salzman (p at dirac.org):

> In response to Rick, this machine is a gaming machine, and it does get
> rather hot.  I ran memtest86 three times, and all three times memtest86
> bombed with a "general protection fault".  It looks like the first DIMM may
> have gone bad.
> 
> I'll swap it out and see if memtest can run to completion...

Yes.  Sometimes, the only way you can narrow down the cause is to run
multiple 12-hour runs of memtest86, each with a different stick of RAM 
_extracted_ from the machine.

I even once had a horror story where every one of those runs ran to
completion with no errors, but then memtest86 showed RAM errors whenever
I put in _all_ the sticks, regardless of which stick sat in which
socket.  Eventually, I figured out that the _memory socket_ (i.e.,
socket on the motherboard) for one of the high-numbered memory banks had
gone bad.  (Technically, it was probably some of the motherboard's
support circuitry for that socket.)  _That_ was likely to have been 
heat damage.

In any event, your main weapon is logic.  (I hate that.  ;->  )