On Saturday 29 November 2003 05:46 pm, Marshal Newrock wrote:
> On Sat, 29 Nov 2003, Ernie Schroder wrote:
> > Memtest86 passes without error or crash, for that matter. But,
> > good point. I tried that yesterday. Hmmm. it's been up 4 1/2 hrs
> > now. That's the best so far today.
>
> I actually had bad RAM that passed a memtest86. Of course, I only
> had it on normal tests and not the full tests. Possibly memtest86
> also misses certain types of bad RAM, as it doesn't provide a
> "real-world" model of testing (I believe it's a linear test from
> beginning to end of memory). All I know is that with one stick of
> RAM in my machine, it kept randomly locking up and certain apps
> would crash, and with a different stick, nothing has crashed in
> months.
The machine has been solid for months. In fact it had been up 89 days
before this started. It has never crashed while stressed. The last
time it happened, top showed about 165 megs of memory free, 30 megs
cached and no swap being used CPU usage showed 98% idle, and CPU temp
was 40.5 degrees C.
The logs show nothing of use other than the daemon.log lets me nail
down the time of the crash to a 10 minute window due to the fact that
noip looks at my IP address periodicly. No changes to make.conf have
been made, and nothing of significance is different.
It seems to me that bad ram might show up under stress but not under
virtually no load. I've been thinking about adding another 512Megs so
maybe 2 new sticks would rule memory failure out. If it doesn't cure
the problem, what could be wrong with a gig and a half of RAM?
--
Regards, Ernie
100% Microsoft and Intel free
--
[EMAIL PROTECTED] mailing list
