"Thermal Event" Enigmatic issue that I need help on

[Niclo Iste]

ok so I go to a clients and their pc reboots every time they go to log in. I remove the MS patch that nerfs Norton and things work fine, it even works fine in safe mode. However PC still reboots now but at random. I thought ok maybe it's the system overheating... but the heat sink is only barely warm. I thought ok maybe a bad heat sensor coding so I update the bios and still get the issue. Mind you this all started last week he's not sure the exact day. Here's the kicker. I get a reboot if I'm starting windows, running linux off a cd or if I do a scan disk. I can run a motherboard/cpu/memory diagnostics tool and a stress test program for hours on end and no errors or reboots. Any ideas?

[slgrieb]

I think memory diagnostics are the least reliable of all software tests, so my first thought would be to try different RAM, or remove some of the installed modules and see what happens. One good clue can be the temperature of
the modules. If you touch a module and it blisters your finger, that's too hot. Of course, you can substitute a thermometer (I like the infrared jobs).

You might consider other environmental factors. You say you've done "hours" of tests. Did you do them on site or at the office? I've seen situations where an overloaded circuit can cause random reboots, or cause machines to reboot every time the AC kicks on.

I'd also just verify the physical stuff like the quality of thermal compound on the processor. It just takes a minute to pull the heatsink and inspect the thermal interface material. Has it crept? Do you see any signs that appear to indicate the heatsink didn't seat correctly on the processor? Either of these problems could cause a hot CPU and a cool heatsink.

[NooNoo]

Is there surge protection? It might be dirty power or the psu as well.

[Platypus]

For a "Thermal Event", definitely check for bad capacitors.

[Ferrit]

I am with slgrieb on this one. Memory is tough to test even with good testing software.
I have seen modules pass memtest and then fail on goldmem.
www.goldmemory.cz

[Niclo Iste]

For a "Thermal Event", definitely check for bad capacitors.

After trying everyone elses suggestions I finally looked all over on the motherboard and you were right it was capacitors. 3 underneath the exhaust fan enclosure were leaking. Thanks for the input everyone I was really stumped on this.

[Platypus]