OT?: Major PC hardware failure

I built a pc around Pentium cpu, Intel D865GVHZ board, Antec case/PS back in '06.
Have been running it about 16 hours/day ever since.

Whilst running a backup Sunday, I heard a constant little tic-tic, tic-tic,
tic-tic sound, looked down to find Speedfan registered about 74 C. then watched
the W2k system crash.

I immediately suspected a cpu fan failure, but all fans were running when I checked.
I replaced the cpu fan anyway. This morn it wouldn't even POST, showed only an
"Intel board" screen. A bit later it wouldn't power at all (no fans. lights,
nothing).

The Power Supply is Antec Smart-Power 350w. It could have gone bad and overjuiced
the cpu causing the 74 C. temp? After which the Intel overheated-cpu
function started shutting everything down?

I've not had a system fail in this manner. Hope someone has some idea
what happened.

Thx,
Puddin'

"Law Without Equity Is No Law At All. It Is A Form Of Jungle Rule."

Puddin' Man wrote:
I built a pc around Pentium cpu, Intel D865GVHZ board, Antec case/PS back in '06.
Have been running it about 16 hours/day ever since.

Whilst running a backup Sunday, I heard a constant little tic-tic, tic-tic,
tic-tic sound, looked down to find Speedfan registered about 74 C. then watched
the W2k system crash.

I immediately suspected a cpu fan failure, but all fans were running when I checked.
I replaced the cpu fan anyway. This morn it wouldn't even POST, showed only an
"Intel board" screen. A bit later it wouldn't power at all (no fans. lights,
nothing).

The Power Supply is Antec Smart-Power 350w. It could have gone bad and overjuiced
the cpu causing the 74 C. temp? After which the Intel overheated-cpu
function started shutting everything down?

I've not had a system fail in this manner. Hope someone has some idea
what happened.

Thx,
Puddin'

"Law Without Equity Is No Law At All. It Is A Form Of Jungle Rule."


Have a look at your Southbridge. You just had a latchup failure
on your ICH5 :-)

Because the motherboard won't start, that means you had a major failure.
Your ICH5 or ICH5R should look like this. (Back in that era, the
Southbridge didn't have a heatsink on it. The Northbridge, usually does.
The Southbridge is the chip you can visually inspect, without taking
anything apart.)

http://onfinite.com/libraries/179057/2ea.jpg

It would be fun to ask the warranty people at Intel, how many of
those blew. Musta been a few. It's a semiconductor problem,
not publicly acknowledged by Intel. Gigabyte is the only company
to post a warning about the problem.

Replace the motherboard, and your other components should be
OK. I haven't heard of collateral damage from that. It's just
the Southbridge which is "toast".

The burn mark in the picture, is over top of the contacts underneath,
which power USB I/O. A possible explanation, is static electricity
enters via a USB port, and causes the USB I/O pad to go into latchup.
That causes a conducting path to form between VCC and GND. If the bond
wires burn out, it's called a "minor failure" and you lose all your USB
interfaces. There would be no burn mark in that case, because the
bond wires let go, before anything gets fried. Device Manager
still shows USB entries (because the logic blocks are intact), but
no plugged in devices are detected. Since the bond wire is burned out,
the D+ and D- signals can no longer receive info from USB peripherals.

If the bond wires remain intact during the event, the chip heats up until
a hole is burned in the lid. The silicon is ruined, and the board will no longer
start.

While Gigabyte claimed, in their warning, that ICH4 and ICH5 were affected,
the vast majority of reported failures in newsgroups are ICH5. I think
I've only read one report of an ICH4 blowing.

Paul

Hi Paul,

I saved your notes re ICH5 problems some time ago.

My Southbridge has markings:
I N T E L
FW82801EB
F5441A13
Intel@@'02
Korea

and it's square, unlike on your link below.

And there's NO evidence of burnout. The chip and it's
peripheral connections look OK.

At the time of failure, I got RUNDLL Message Popup. Since then I get:

"The CPU was previously shutdown due to a thermal event(overheating)
Service the unit right away to resolve this"

Any/all advice on getting the unit "serviced" much appreciated.

Thx,
P
On Wed, 06 Apr 2011 17:35:35 -0400, Paul wrote:

Have a look at your Southbridge. You just had a latchup failure
on your ICH5 :-)

Because the motherboard won't start, that means you had a major failure.
Your ICH5 or ICH5R should look like this. (Back in that era, the
Southbridge didn't have a heatsink on it. The Northbridge, usually does.
The Southbridge is the chip you can visually inspect, without taking
anything apart.)

http://onfinite.com/libraries/179057/2ea.jpg

It would be fun to ask the warranty people at Intel, how many of
those blew. Musta been a few. It's a semiconductor problem,
not publicly acknowledged by Intel. Gigabyte is the only company
to post a warning about the problem.

Replace the motherboard, and your other components should be
OK. I haven't heard of collateral damage from that. It's just
the Southbridge which is "toast".

The burn mark in the picture, is over top of the contacts underneath,
which power USB I/O. A possible explanation, is static electricity
enters via a USB port, and causes the USB I/O pad to go into latchup.
That causes a conducting path to form between VCC and GND. If the bond
wires burn out, it's called a "minor failure" and you lose all your USB
interfaces. There would be no burn mark in that case, because the
bond wires let go, before anything gets fried. Device Manager
still shows USB entries (because the logic blocks are intact), but
no plugged in devices are detected. Since the bond wire is burned out,
the D+ and D- signals can no longer receive info from USB peripherals.

If the bond wires remain intact during the event, the chip heats up until
a hole is burned in the lid. The silicon is ruined, and the board will no longer
start.

While Gigabyte claimed, in their warning, that ICH4 and ICH5 were affected,
the vast majority of reported failures in newsgroups are ICH5. I think
I've only read one report of an ICH4 blowing.

"Law Without Equity Is No Law At All. It Is A Form Of Jungle Rule."

Puddin' Man wrote:
Hi Paul,

I saved your notes re ICH5 problems some time ago.

My Southbridge has markings:
I N T E L
FW82801EB
F5441A13
Intel@@'02
Korea

and it's square, unlike on your link below.

And there's NO evidence of burnout. The chip and it's
peripheral connections look OK.

At the time of failure, I got RUNDLL Message Popup. Since then I get:

"The CPU was previously shutdown due to a thermal event(overheating)
Service the unit right away to resolve this"

Any/all advice on getting the unit "serviced" much appreciated.

Thx,
P

OK, have you pulled the CPU heatsink/fan and checked for damage ?
Do you still have thermal paste between the CPU and heatsink ?

Maybe the Vcore circuit overvolted and ran the CPU at higher than
normal voltage ? The last datasheet I looked at, Vcore is supposed
to have a check for overvoltage - the regulator will shut down,
if the output is off by a certain percentage or more. So normally
you're protected from that.

The Vcore circuit doesn't have "total control", and I suppose it
would be possible for a fault to occur, that the regulator chip
cannot stop. Only a certain class of faults can be detected
by the regulator chip, and stopped relatively quickly.

*******

The processor has two levels of thermal protection. It uses throttling
first, to try to stop the overheat. If it gets too hot, it uses
THERMTRIP, which should cause the power supply to shut off.

THERMTRIP or a Vcore failure, might be latching faults. That means,
to clear then, all power must be removed by the computer. The easiest
way to do that, is to flip the rear switch to OFF for 60 seconds, then
flip it back on again.

*******

Relative to other components in the computer, processors are
extremely reliable. It is possible for a silicon die to crack,
if it's under enough mechanical stress. And perhaps a resulting
short circuit on the die, can result in overheating.

The only reason I'm not blaming your power supply at all, is
your observation of the ongoing overtemperature indication,
which would be coming from the processor. Vcore can handle excursions
on +12V, without transferring that through to the processor. So if
the 12V shot up to 16V for example, it might fry all the hard
drives, but it wouldn't bother Vcore at all. Onboard regulators
help to "insulate" motherboard components, from power supply
problems. Components directly connected to the supply (PCI cards,
hard drives) are more at risk.

So right now, all I can suggest, is check to see whether the
heatsink is firmly attached to the motherboard. Sometimes, a plastic
latch snaps or the like. Also, visually inspect the capacitors
around the motherboard socket, for bulging or leakage.

http://upload.wikimedia.org/wikipedi...pacitor_01.jpg

*******

On some motherboards, the Northbridge heatsink falls off, when the
solder joint fails on the steel wire holding the heatsink to the
motherboard. Your symptoms don't seem to be consistent with that,
but you can check the Northbridge heatsink for being securely in
place, while you're doing your other inspections. At least one
Dell computer, electrically checks that the heatsink is secure, and
the computer won't boot unless the mechanical wire is soldered
back into place.

Paul

"Puddin' Man" wrote in message
...

I built a pc around Pentium cpu, Intel D865GVHZ board, Antec case/PS back
in '06.
Have been running it about 16 hours/day ever since.

Whilst running a backup Sunday, I heard a constant little tic-tic,
tic-tic,
tic-tic sound, looked down to find Speedfan registered about 74 C. then
watched
the W2k system crash.

I immediately suspected a cpu fan failure, but all fans were running when
I checked.
I replaced the cpu fan anyway. This morn it wouldn't even POST, showed
only an
"Intel board" screen. A bit later it wouldn't power at all (no fans.
lights,
nothing).

The Power Supply is Antec Smart-Power 350w. It could have gone bad and
overjuiced
the cpu causing the 74 C. temp? After which the Intel overheated-cpu
function started shutting everything down?

I've not had a system fail in this manner. Hope someone has some idea
what happened.

Thx,
Puddin'


I'd test the PSU.
Disconnect all leads from the M/B, drives etc.
Short the green (PS_ON#) to black (Com) and check the voltages.
See http://www.smpspowersupply.com/connectors-pinouts.html

Around 06 I built 4 systems based on Antec cases with PSUs.
3 out the four PSUs malfunctioned and destroyed M/Bs and drives within about
a month of each other.
I changed the PSU in the fourth and the system is still running.

Eric

On Thu, 07 Apr 2011 02:30:09 -0400, Paul wrote:

OK, have you pulled the CPU heatsink/fan and checked for damage ?
Do you still have thermal paste between the CPU and heatsink ?

I installed new HS/fan with paste the night it crashed.

Maybe the Vcore circuit overvolted and ran the CPU at higher than
normal voltage ?

Sounds likely.

The last datasheet I looked at, Vcore is supposed
to have a check for overvoltage - the regulator will shut down,
if the output is off by a certain percentage or more. So normally
you're protected from that.

The Vcore circuit doesn't have "total control", and I suppose it
would be possible for a fault to occur, that the regulator chip
cannot stop. Only a certain class of faults can be detected
by the regulator chip, and stopped relatively quickly.

*******

The processor has two levels of thermal protection. It uses throttling
first, to try to stop the overheat. If it gets too hot, it uses
THERMTRIP, which should cause the power supply to shut off.

THERMTRIP or a Vcore failure, might be latching faults. That means,
to clear then, all power must be removed by the computer. The easiest
way to do that, is to flip the rear switch to OFF for 60 seconds, then
flip it back on again.

I've done that numerous times. No help.

*******

Relative to other components in the computer, processors are
extremely reliable. It is possible for a silicon die to crack,
if it's under enough mechanical stress. And perhaps a resulting
short circuit on the die, can result in overheating.

The only reason I'm not blaming your power supply at all, is
your observation of the ongoing overtemperature indication,
which would be coming from the processor.

Check. I've installed a new PS and it's made no difference
in system behavior. Problem persists.

Vcore can handle excursions
on +12V, without transferring that through to the processor. So if
the 12V shot up to 16V for example, it might fry all the hard
drives, but it wouldn't bother Vcore at all. Onboard regulators
help to "insulate" motherboard components, from power supply
problems. Components directly connected to the supply (PCI cards,
hard drives) are more at risk.

The primary HD didn't fry: looked OK on the other pc. There
were no PCI cards.

So right now, all I can suggest, is check to see whether the
heatsink is firmly attached to the motherboard. Sometimes, a plastic
latch snaps or the like. Also, visually inspect the capacitors
around the motherboard socket, for bulging or leakage.

All that has been done. No evident problem.

http://upload.wikimedia.org/wikipedi...pacitor_01.jpg

*******

On some motherboards, the Northbridge heatsink falls off, when the
solder joint fails on the steel wire holding the heatsink to the
motherboard. Your symptoms don't seem to be consistent with that,
but you can check the Northbridge heatsink for being securely in
place, while you're doing your other inspections. At least one
Dell computer, electrically checks that the heatsink is secure, and
the computer won't boot unless the mechanical wire is soldered
back into place.

The Northbridge heatsink is fine. Everything on the board looks OK.

I power it up for about 20 secs, then it shuts all power down.
Should there be some sort of Vcore reset in bios? Any other way
to effect a reset?

Thanks,
P

"Law Without Equity Is No Law At All. It Is A Form Of Jungle Rule."

On Thu, 7 Apr 2011 11:42:45 +0100, "Eric Parker" wrote:

I'd test the PSU.
Disconnect all leads from the M/B, drives etc.
Short the green (PS_ON#) to black (Com) and check the voltages.
See http://www.smpspowersupply.com/connectors-pinouts.html

Around 06 I built 4 systems based on Antec cases with PSUs.
3 out the four PSUs malfunctioned and destroyed M/Bs and drives within about
a month of each other.
I changed the PSU in the fourth and the system is still running.

I'll assume the Antecd PS caused the problem 'till I have evidence
to the contrary.

I have new Extreme 460w PS installed: problem persists.

Never again Antec PS for me.

Thanks,
P

"Law Without Equity Is No Law At All. It Is A Form Of Jungle Rule."

Puddin' Man wrote:


The Northbridge heatsink is fine. Everything on the board looks OK.

I power it up for about 20 secs, then it shuts all power down.
Should there be some sort of Vcore reset in bios? Any other way
to effect a reset?

Thanks,
P

"Law Without Equity Is No Law At All. It Is A Form Of Jungle Rule."


Your D865GVHZ should have THERMTRIP, and shutting off after 20 seconds
could be due to that. The BIOS obviously has the skill set to shut
off the computer too, but then the question would be, why. Some
BIOS will check for a minimum fan rotation speed on the CPU fan,
and shut off the computer if they don't detect it. That would be for
the CPU fan, while other fans are allowed to die if they want. Usually,
on the CPU fan would be considered as a reason to shut off the
computer.

Vcore is programmable a couple ways.

On a non-enthusiast motherboard, they would use the facilities provided
by default. The processor has VID signals on the bottom of it. The bit
pattern specifies a voltage. Those can be connected directly to a
similar interface on the Vcore regulator chip. Modern processors even
adjust that voltage dynamically, as a function of computing load
(a feature of Intel SpeedStep, otherwise known as EIST).

On an enthusiast motherboard, the Vcore may have an additional input feature,
that allows adding an offset to the "regular" voltage value. That is used
when overclocking. On my last motherboard, I found a pin on the
Vcore regulator, by which I could offset the voltage by 0.1 volts, so I could
try overclocking (even though the motherboard didn't support it). When a
BIOS supports such a feature, there is an actual Vcore setting in the BIOS
screen (sometimes it'll be "+0.1" implying a bump of 0.1 volts, rather than
the BIOS screen giving an absolute voltage as such).

Your Intel board probably doesn't have that, and just uses the VID code as
it comes from the processor and goes to Vcore. Since you can't boot the
system right now, there is no way to get into Windows and use CPUZ, to check
what CPUZ sees as the current Vcore value. I also checked the user manual
for the motherboard, and I don't see a hardware monitor page in there. One
option would have been, to use the 20 seconds to visit the hardware monitor
page and check a Vcore reading there. (Some SuperI/O chips have eight channels
of ADC or analog to digital conversion, for making crude 8 bit voltage measurements.
They use that to measure the three main power supply rails, Vcore, and present
that in a BIOS hardware monitor page.)

Without that, you'd have to get out a multimeter, and figure out where to probe,
if you thought it was actually a Vcore problem.

Note that, a motherboard can "run" without a processor installed in it. You would
remove the memory and the processor, and operate with just the bare board.
Pressing the switch on the front of the computer, with the panel wiring attached,
should allow the power supply to be turned on and off. In this case, the purpose
of the test, would be to verify the control circuit is working. If the
thing still died after 20 seconds, then you'd know it likely wasn't
the processor. In such a scenario, motherboard running without processor
installed, the VID signals are "floating", they float to all 1's logic
value, and VCore gets set to zero volt output when that happens. That
means effectively, the VCore regulator is not loaded. So the motherboard
is pretty well neutered. You would expect the power to remain running,
because many of the "switch-off" mechanisms have been removed.

I don't particularly see a reason to try a CMOS reset or the like
in this case. You can try it, but at the very least, remove all
power from the system before doing it. Intel tends to do things
differently than other manufacturers, so I don't even know what
the gotchas are with their procedure. I don't think the BIOS can set
the voltage to anything dangerous - it doesn't give the impression
of an enthusiast board when I read the manual. I mean, not finding
a hardware monitor page, seemed pretty bad. You'd find that feature
available on a lot of other brands.

Paul

Problem persists. Clear CMOS doesn't help.

I strongly suspect that the "thermal event" kicked THERMTRIP in, and
that THERMTRIP isn't allowing the default reset that would enable
resumption of operation (deficiency in THERMTRIP design).

IIRC, the system has a Winbond chip, and there is a
temp, etc monitor program (that doesn't function properly).
Runs under Windoze, of course.

Thanks for the cpu-less test info: I didn't know about that.

I have an old Celeron S-478 cpu that I can sub. I guess I'll
wait until tomorrow, do the cpu-less test, and if it keeps
"running", install the Celeron, which is not THERMTRIP-ed.

Many thanks, more info tomorrow.

Prost,
P

On Thu, 07 Apr 2011 18:59:28 -0400, Paul wrote:

Your D865GVHZ should have THERMTRIP, and shutting off after 20 seconds
could be due to that. The BIOS obviously has the skill set to shut
off the computer too, but then the question would be, why. Some
BIOS will check for a minimum fan rotation speed on the CPU fan,
and shut off the computer if they don't detect it. That would be for
the CPU fan, while other fans are allowed to die if they want. Usually,
on the CPU fan would be considered as a reason to shut off the
computer.

Vcore is programmable a couple ways.

On a non-enthusiast motherboard, they would use the facilities provided
by default. The processor has VID signals on the bottom of it. The bit
pattern specifies a voltage. Those can be connected directly to a
similar interface on the Vcore regulator chip. Modern processors even
adjust that voltage dynamically, as a function of computing load
(a feature of Intel SpeedStep, otherwise known as EIST).

On an enthusiast motherboard, the Vcore may have an additional input feature,
that allows adding an offset to the "regular" voltage value. That is used
when overclocking. On my last motherboard, I found a pin on the
Vcore regulator, by which I could offset the voltage by 0.1 volts, so I could
try overclocking (even though the motherboard didn't support it). When a
BIOS supports such a feature, there is an actual Vcore setting in the BIOS
screen (sometimes it'll be "+0.1" implying a bump of 0.1 volts, rather than
the BIOS screen giving an absolute voltage as such).

Your Intel board probably doesn't have that, and just uses the VID code as
it comes from the processor and goes to Vcore. Since you can't boot the
system right now, there is no way to get into Windows and use CPUZ, to check
what CPUZ sees as the current Vcore value. I also checked the user manual
for the motherboard, and I don't see a hardware monitor page in there. One
option would have been, to use the 20 seconds to visit the hardware monitor
page and check a Vcore reading there. (Some SuperI/O chips have eight channels
of ADC or analog to digital conversion, for making crude 8 bit voltage measurements.
They use that to measure the three main power supply rails, Vcore, and present
that in a BIOS hardware monitor page.)

Without that, you'd have to get out a multimeter, and figure out where to probe,
if you thought it was actually a Vcore problem.

Note that, a motherboard can "run" without a processor installed in it. You would
remove the memory and the processor, and operate with just the bare board.
Pressing the switch on the front of the computer, with the panel wiring attached,
should allow the power supply to be turned on and off. In this case, the purpose
of the test, would be to verify the control circuit is working. If the
thing still died after 20 seconds, then you'd know it likely wasn't
the processor. In such a scenario, motherboard running without processor
installed, the VID signals are "floating", they float to all 1's logic
value, and VCore gets set to zero volt output when that happens. That
means effectively, the VCore regulator is not loaded. So the motherboard
is pretty well neutered. You would expect the power to remain running,
because many of the "switch-off" mechanisms have been removed.

I don't particularly see a reason to try a CMOS reset or the like
in this case. You can try it, but at the very least, remove all
power from the system before doing it. Intel tends to do things
differently than other manufacturers, so I don't even know what
the gotchas are with their procedure. I don't think the BIOS can set
the voltage to anything dangerous - it doesn't give the impression
of an enthusiast board when I read the manual. I mean, not finding
a hardware monitor page, seemed pretty bad. You'd find that feature
available on a lot of other brands.

Paul

"Law Without Equity Is No Law At All. It Is A Form Of Jungle Rule."

Nothin' works!

I meant to wait 'till tomorrow, but these thangs bother and bother me,
and I went ahead and fiddled the damn thang.

I cleared CMOS again, then pulled cpu and mem. Powered up and ...
nuthin'!. No fans, no anything.

Installed old Celeron (which ran in this system years ago), HS/fan
(w paste), mem. Powered up and ... it spun the fans for a second or
2, then shut down everything. This is what it has been doing lately:
a sec or 2 then shut down.

Beats me. I was hoping at least the board would spin the fans
w/o cpu, mem.

Gawd, whotta headache!

Any ideas? Mobo is now highly suspect??

Thx,
P

"Law Without Equity Is No Law At All. It Is A Form Of Jungle Rule."