Subject: ram defects
To: None <port-sparc@netbsd.org>
From: Paul (NCC/CS) <pts@bom.gov.au>
List: port-sparc
Date: 12/11/2000 12:38:20
Hi there  People,

I've been having some signal 11 bomb outs during heavy compiles
on my sparcstation 1+.  I've been shuffling RAM in and out of the
machine to try and find the faulty chips and I've ended up with a new
set of 8, 4 meg simms (to get 32 mb) and these seem to work quite
well. That is after a few hours of compiling there were no bomb outs
like before.

Now I've got about 12, 4mb simms left out, of which one or some
are faulty and the rest are probably ok. The trouble is determining
which simms are ok so I can salvage some and increase my RAM
total back from 32 to somewhere closer to 64 Meg.

See each time I boot and run some compile tests in different RAM
amounts, the layout of programs in RAM could change. And hence
it could be any of the particular simms used that is really being
tested.
I don't know which one passes or fails as the case may be. It's almost
impossible to shuffle the old simms (you have to have blocks of 4 in or
out at a time) and test each time in order to find the simm(s) that is
damaged.

So my question:

1. Is there a program out there that can heavily test (like in a
compile)
a particular RAM simm in isolation so as to properly test it,
and then report and point the finger at one simm and say ok
or bad?

2. Alternatively what strategies do you other guys have for finding
bad RAM chips among the good ones when this sort of thing happens
to you?

PS remember the power on self test of RAM is not worth a pinch
of bulldust. It's only when it gets hammered by say a compile that
you really know whether or not the RAM is faulty.

Paul.





--

____________________________________________

Paul   (pts@bom.gov.au)
National Climate Centre
Australian Bureau Of Meteorology
____________________________________________