Sporadic freeze-ups and hair loss

[Grinder]

I've been having problems diagnosing a flakey Pentium II 266 MHz
(AL440LX mainboard) PC. It seems to determined to freeze-up after you
start to feel any confidence about having isolated the problem.

At any rate, these are the steps that I've followed, taken from the
mouth of a local wise man/guy:

| 1) Unplug all power supply leads except one to a hard
| drive, then plug the AC cord back in. Measure 5VSB,
| should be near enough to 5V. Measure PS_ON, should
| be 3-5V.

Check. The voltages look good.

| 2) Next short PS_ON to ground, at which point the PSU
| should turn on, fan spinning and hard drive spinning.
| You can then check 12V and 5V on another power plug.

Check. The PSU and hard drive behave as expected, and the voltages look
good.

| 3) If all is well so far, unplug from AC, plug back into
| motherboard and disconnect all non-essential parts from
| motherboard, leaving NOTHING in the board, no CPU or
| memory or video card. With nothing in the board the
| power-on switch connected, it should turn on the power
| supply.

Crud. Absolutely no response from pressing the power switch. I'll also
point out here that this diagnosis has been performed with the mainboard
liberated from the case -- it's sprawled on a wooden table, on one of
those anti-static mats.

| 4) If it does turn on in this barebones state, power off,
| remove AC cord, and add CPU, Memory, Video card.

Ploughing on ahead, I pop in Pentium II, a couple of sticks of RAM (that
I feel pretty good about in the short term at least,) and the Radeon
7000 into an AGPx2 slot. The system boots up normally.

The freeze-ups, though, remain.

I replace the video card with a Voodoo3 3500 card, that has always
served me well. (3dfx I mourn for you.) Much like the Radeon, the
system will work for fairly long stretchs (1-2 hours) then suddenly
freeze. And freeze. And freeze. Then, it comes to its senses and
behaves for awhile again.

I've tried clearing the CMOS in a few iterations, but to no apparent effect.

I have also inspected the mainboard for since of blown/vented
capacitors, toasted diodes, etc ... general damage, and can so nothing
wrong.

What's a boy to do?

If you've made it this far, thanks for taking the time. If you offer
additional speculation, that icing will receive its own share of gratitude.

 

[kony]

I've been having problems diagnosing a flakey Pentium II 266 MHz
(AL440LX mainboard) PC. It seems to determined to freeze-up after you
start to feel any confidence about having isolated the problem.

Did it come out of an OEM box, and if so, what system model?

At any rate, these are the steps that I've followed, taken from the
mouth of a local wise man/guy:

| 1) Unplug all power supply leads except one to a hard
| drive, then plug the AC cord back in. Measure 5VSB,
| should be near enough to 5V. Measure PS_ON, should
| be 3-5V.

Check. The voltages look good.

You'll need to check the voltages again with all the gear
hooked up. Some OEMs (notibly Gateway) used marginal PSU in
that era... the PSU was a good design but insufficient
capacity... but even so, now it might be 8 years old at
which point the function may have degraded further.

The motherboard itself is another story, they were very good
quality boards and are most likely to gracefully age
relative to most other contemporary motherboards. It
wouldn't make them immune to
misahanding/static-discharge/lightning damage/etc, though.

| 2) Next short PS_ON to ground, at which point the PSU
| should turn on, fan spinning and hard drive spinning.
| You can then check 12V and 5V on another power plug.

Check. The PSU and hard drive behave as expected, and the voltages look
good.

| 3) If all is well so far, unplug from AC, plug back into
| motherboard and disconnect all non-essential parts from
| motherboard, leaving NOTHING in the board, no CPU or
| memory or video card. With nothing in the board the
| power-on switch connected, it should turn on the power
| supply.

Crud. Absolutely no response from pressing the power switch. I'll also
point out here that this diagnosis has been performed with the mainboard
liberated from the case -- it's sprawled on a wooden table, on one of
those anti-static mats.

It does seem the board (or power supply) are failing/failed,
unless there is any possibility you are simply using the
wrong two (not the PS-On) pins to try and turn it on.

HOWEVER, do not put it on an anti-static mat. Anti-static
mats do, by design, conduct electricity. Several times I
have observed those mats preventing a board from working...
though IIRC, it was only temporary, after removing the mat
the boards then worked (of course, unless there was some
other problem besides their being on a mat). Test it again
without the mat.

| 4) If it does turn on in this barebones state, power off,
| remove AC cord, and add CPU, Memory, Video card.

Ploughing on ahead, I pop in Pentium II, a couple of sticks of RAM (that
I feel pretty good about in the short term at least,) and the Radeon
7000 into an AGPx2 slot. The system boots up normally.

The freeze-ups, though, remain.

It's still on the mat?
Had you taken voltages at the motherboard connector with a
multimeter at this point?

You might also check on whether the cards (particularly
video) are sitting very straight in the AGP slot. A cockeyed
video card can freeze a system and if it had been cockeyed
bad enough it can bend the slot contacts a little. A strong
light, magnifying glass, and very thin needle might be used
to straighten up any contacts that look bend when the card
is removed, carefully.

I replace the video card with a Voodoo3 3500 card, that has always
served me well. (3dfx I mourn for you.) Much like the Radeon, the
system will work for fairly long stretchs (1-2 hours) then suddenly
freeze. And freeze. And freeze. Then, it comes to its senses and
behaves for awhile again.

I've tried clearing the CMOS in a few iterations, but to no apparent effect.

I have also inspected the mainboard for since of blown/vented
capacitors, toasted diodes, etc ... general damage, and can so nothing
wrong.

What's a boy to do?

If you've made it this far, thanks for taking the time. If you offer
additional speculation, that icing will receive its own share of gratitude.

Sometimes an old power supply may measure with acceptible
voltages but still be failing. If you had another spare
(any decent, not ancient, 250W should suffice for the
described parts).

 

[Grinder]

Did it come out of an OEM box, and if so, what system model?

It's from a Gateway G6-266m.

You'll need to check the voltages again with all the gear
hooked up. Some OEMs (notibly Gateway) used marginal PSU in
that era... the PSU was a good design but insufficient
capacity... but even so, now it might be 8 years old at
which point the function may have degraded further.

The motherboard itself is another story, they were very good
quality boards and are most likely to gracefully age
relative to most other contemporary motherboards. It
wouldn't make them immune to
misahanding/static-discharge/lightning damage/etc, though.

I've since tried the setup with a known good 300W replacment -- with the
same results.

It does seem the board (or power supply) are failing/failed,
unless there is any possibility you are simply using the
wrong two (not the PS-On) pins to try and turn it on.

HOWEVER, do not put it on an anti-static mat. Anti-static
mats do, by design, conduct electricity. Several times I
have observed those mats preventing a board from working...
though IIRC, it was only temporary, after removing the mat
the boards then worked (of course, unless there was some
other problem besides their being on a mat). Test it again
without the mat.
Check.



It's still on the mat?

No, I've since retested without the mat.

Had you taken voltages at the motherboard connector with a
multimeter at this point?

You might also check on whether the cards (particularly
video) are sitting very straight in the AGP slot. A cockeyed
video card can freeze a system and if it had been cockeyed
bad enough it can bend the slot contacts a little. A strong
light, magnifying glass, and very thin needle might be used
to straighten up any contacts that look bend when the card
is removed, carefully.

The cards and slots look good -- perfectly perpendicular and nice and tight.

Sometimes an old power supply may measure with acceptible
voltages but still be failing. If you had another spare
(any decent, not ancient, 250W should suffice for the
described parts).

Oddly enough, the PC has been a real run of properly behaving. I have a
busy program running on it now -- 4 hours and running with no freeze-up.

 

[Grinder]

Sounds like a heat problem to me. Is the processor a slot 1, or socket
7? Many of these old gateways had little provisions for cooling major
components. Feel the memory, CPU, hard drive, north / south bridge,
and video card for abnormal temperatures. Please report. :? :?

Slot 1.

Heat sensitivity seems like a good suspicion as well, but there is
apparently no cooling that qualifies as adequate. I can still get
freeze-ups with the system out of its case, with box fans blowing upon it.

 

[kony]

Slot 1.

Heat sensitivity seems like a good suspicion as well, but there is
apparently no cooling that qualifies as adequate. I can still get
freeze-ups with the system out of its case, with box fans blowing upon it.

During that era Gateway commonly used a PSU fan shroud to
cool the CPU, and CPU only had tall tinned passive 'sink.
Are your case fans situated close enough to force enough air
through the 'sink? Does the 'sink feel hot? P2/266 was
relatively hot running for that era, but not too hard to
cool put in context of semi-modern systems.

However, some P2 Katmai (maybe all?) had cache chips that
didn't contact the heat spreader behind the heatsink itself.
I vaguely recall some article about moddin' a P2 to cool the
cache chips better by lapping a three pennies (one per cache
chip and one for the controller chip) to the right thickness
then epoxying them onto the chips. It seems like a lot of
work to do for a 266MHz CPU, today, though.

You migth try isolating things further. Does the board have
the last/newest BIOS? You might reflash newest bios then
clear CMOS w/AC disconnected. Is it using ACPI power
management in windows? You might try removing one of the
memory modules, then alternating, removing the other, and
testing with memtest86. Test with Prime95's Torture Test
too.

Is it the same windows installation? If so, is it possible
that prior instability has corrupted OS/other files and now
even after regaining hardware stability, the corrupt files
are a problem?

 

[Grinder]

During that era Gateway commonly used a PSU fan shroud to
cool the CPU, and CPU only had tall tinned passive 'sink.
Are your case fans situated close enough to force enough air
through the 'sink? Does the 'sink feel hot? P2/266 was
relatively hot running for that era, but not too hard to
cool put in context of semi-modern systems.

Yes, I believe so. After shutting the system down, I would sometimes
touch the heatsink -- it was warm (bathwater warm) but not hot to the touch.

However, some P2 Katmai (maybe all?) had cache chips that
didn't contact the heat spreader behind the heatsink itself.
I vaguely recall some article about moddin' a P2 to cool the
cache chips better by lapping a three pennies (one per cache
chip and one for the controller chip) to the right thickness
then epoxying them onto the chips. It seems like a lot of
work to do for a 266MHz CPU, today, though.

You migth try isolating things further. Does the board have
the last/newest BIOS? You might reflash newest bios then
clear CMOS w/AC disconnected.

There is a version of the BIOS that's about six months newer that what I
have.

Is it using ACPI power
management in windows?

Yes, power management seems to be working under Windows 98se.

You might try removing one of the
memory modules, then alternating, removing the other, and
testing with memtest86.

I have tested the large stick (128 mb) by itself, with Memtest86. Also,
I've run the system quite a bit with just that module, with freeze-ups,
of course. I'll try the other (32 Mb) one.

Test with Prime95's Torture Test too.

Hadn't heard of that one -- I'll try it.

Is it the same windows installation? If so, is it possible
that prior instability has corrupted OS/other files and now
even after regaining hardware stability, the corrupt files
are a problem?

I have a clean Windows 98se install in the system now.

- - - - - - -

I've had the system up and running, with Windows 98se loaded, for about
30 hours, with no freeze up. It would not be very satisfying for the
problem to have gone away by itself, but I'm frustrated enough to start
to hope for it.

 

[kony]

Yes, I believe so. After shutting the system down, I would sometimes
touch the heatsink -- it was warm (bathwater warm) but not hot to the touch.

You can touch it while running- it won't bite you. ;-)
If it didn't feel overly hot then it's plenty cool enough,
assuming the CPU is still in original factory state, IE-
someone didn't get ahold of it before you did and start
taking it apart then not putting it back together right and
reapplying compound, since those too have a one-use-only
thermal material that has to be cleaned off if the 'sink
(heat spreader in this case) is removed.

There is a version of the BIOS that's about six months newer that what I
have.

I wouldn't expect this problem then, but if you can find
notes on what changes it makes, that might be useful
information. "Sometimes" the notes aren't prominently
displayed but are in the bios file download (compressed
file).

I've had the system up and running, with Windows 98se loaded, for about
30 hours, with no freeze up. It would not be very satisfying for the
problem to have gone away by itself, but I'm frustrated enough to start
to hope for it.

With a board that old (and not knowing the history of the
system) it could be something as simple as
dirt/residue/contaminants on some of the slot contacts and
removing and reinserting some parts a few times has cleaned
them off a bit.

 

[Grinder]

With a board that old (and not knowing the history of the
system) it could be something as simple as
dirt/residue/contaminants on some of the slot contacts and
removing and reinserting some parts a few times has cleaned
them off a bit.

It has been running for a considerable amount of time now without
freeze-up. Oddly, though, it still froze up some after having been
disassembled, cleaned, and reassembled numerous times. There's a French
term to describe my ambivalence, but I don't want to look like Doris Day.

This PC's future, repaired or otherwise, is likely to be at the back of
a closet, so I have spent my concern for its stability. I do appreciate
your advisements, however, and will cache them for the next
gremlin-infested machine that crosses my path.

Thanks again to everyone that offered suggestions.