DIMM fault comes and goes

[skotl]

Hi all,

I swapped the DIMM in my main web server (win2003) for another 256MB
module and started to experience hangs / bluescreens in Windows.
I booted up a copy of memtest (http://www.memtest86.com) which I have
always found very reliable, and after a couple of hours it reported
faults around the 220MB location.
Cool, I thought - the DIMM is stuffed, so that's what the problem is.

For completeness, I moved the DIMM to slot 2 and installed the original
DIMM and re-ran memtest. Results were exactly as expected; faults
around 470MB (256MB OK for the first one, and then failing at 220MB on
the second one).

So, I pulled them out of the machine but got them jumbled up. No probs,
I thought, I'll put them back in individually and the one that fails on
MemTest goes in the bin.

Here's the odd bit, though. Both of them now appear to be absolutely
fine! I've ran MemTest solid for days on each of the DIMMs and don't
get a single error!
At first, I thought maybe there had been a bad connection, but that
would have been far more catastrophic than just having Windows hang
every 24 hours or so.

Is it possible for a DIMM to appear to be faulty, and then to regain
it's health?

Confused,
Skot.

 

[Bob Day]

Is it possible for a DIMM to appear to be faulty, and then to regain
it's health?

Not likely, if everything else stayed the same.
But memory *slots* can be bad.

-- Bob Day
http://bobday.vze.com

 

[skotl]

Cheers Bob.

I take your point about slots being bad, but the same question applies.
Can a slot be bad for a while and then appear healthy. Also, all of the
tests should alleged errors at 220MB where you would have though a duff
slot would have had errors up and down the range.

 

[David Maynard]

Cheers Bob.

I take your point about slots being bad, but the same question applies.
Can a slot be bad for a while and then appear healthy.

Dirty connection can do it.

Also, all of the
tests should alleged errors at 220MB where you would have though a duff
slot would have had errors up and down the range.

Why? If it were a bad data line then yes. A bad address line, not
necessarily, especially if an upper address or bank select line.

 

[David Maynard]

Hi all,

I swapped the DIMM in my main web server (win2003) for another 256MB
module and started to experience hangs / bluescreens in Windows.
I booted up a copy of memtest (http://www.memtest86.com) which I have
always found very reliable, and after a couple of hours it reported
faults around the 220MB location.
Cool, I thought - the DIMM is stuffed, so that's what the problem is.

For completeness, I moved the DIMM to slot 2 and installed the original
DIMM and re-ran memtest. Results were exactly as expected; faults
around 470MB (256MB OK for the first one, and then failing at 220MB on
the second one).

So, I pulled them out of the machine but got them jumbled up. No probs,
I thought, I'll put them back in individually and the one that fails on
MemTest goes in the bin.

Here's the odd bit, though. Both of them now appear to be absolutely
fine! I've ran MemTest solid for days on each of the DIMMs and don't
get a single error!
At first, I thought maybe there had been a bad connection, but that
would have been far more catastrophic than just having Windows hang
every 24 hours or so.

Not necessarily because it depends on what windows is using and, in
addition, memtest stresses the memory timing and 'normal' operation may be
slacker on it.

Is it possible for a DIMM to appear to be faulty, and then to regain
it's health?

Dirty contacts.

 

[skotl]

Gets more curious. I switched everything off and let it cool down for a
while, then re-inserted the DIMM and it now shows errors.

I think it's down to a heat thing - the colder the DIMM, the more
chance it will error, the warmer it is then it seems fine.

Still a little odd, as normally components fail when they get hotter.

Still, it was more an academic query as it's on $30 and it's going in
the bin!

Cheers,
Scott

 

[Edward A. Weissbard]

Scott,

That is interesting..............first I've heard of that issue, but that's
always possible.