Tater wrote:
On Mon, 11 Aug 2014 12:00:21 -0700, Tater wrote:


Paul, thanks for your highly detailed reply. After you and Bob
suggested overheating as a good possibility, I did some investigation
and have interesting (and embarrassing) results to report. My case is
an Antec P280, which has dust filters on the front intake fans, and on
the power supply fan. The power supply arrangement in this case is on
the bottom of the case, with the intake facing down, and the filter is
between the power supply intake and the bottom of the case (which has
a perforated section to let air into the power supply).

I checked both filters and found them clean. I blew out all the fans
with compressed air and found almost no dust on them. The only thing
I couldn't easily examine was the power supply fan because of the
location of the power supply. So I turned the case upside down so I
could verify the power supply fan was running. I was astounded to
find the perforated case intake area for the power supply almost
completely clogged with dust!

What was happening was that the dust was accumulating on the
perforations and not even making it to the filter. Since I frequently
checked the filters and found very little dust, I assumed everything
was OK.

After cleaning out the dust and verifying that the power supply fan
was indeed running, I tried running FurMark again with the GTX 580 and
it ran fine! I thought I was home free until I tried running a
refresh on the Windows Experience Index and the system shut down
again. However, repeat refreshes on the Index and further running of
FurMark didn't result in any shutdowns.

Is it possible that running the computer for months (maybe over a
year?) with the power supply air intake clogged has degraded the power
supply and it is in need of replacement?

BTW, my BIOS is not set to autostart after a power failure.
Unfortunately I have not been able to make the system shutdown again
so I could check on your suggestion of observing the behavior of the
LED on the motherboard.

Jerry

Well I was thinking that cleaning out the air intake to my power
supply might have cured the problem, but I ran anotherWindows
Experience Index refresh and it shut down the computer again. This
time the side of the case was open and I was able to observe the
lights during this episode. My motherboard doesn't have a single
green LED as Paul suggested, rather it has an illuminated power
button, an illuminated reset button, and a two digit alpha-numeric
diagnostic display (all of these items are mounted directly on the
motherboard).

When the computer shut down, I quickly looked in the case and the two
digit LED display turned off, but the illumination for the power and
reset switches remained on.

I guess this continues to leave me confused as to where the problem
lies.

Jerry

In some of these situations, you can't tell if it is motherboard side
or power supply side that is doing it.

The processor has THERMTRIP, which can turn off the power supply. If
the processor overheats, it is protected, and THERMTRIP from the CPU
feeds into some motherboard logic, causing PS_ON# to go off.

To test that path, you use a "CPU burn" or "100% loading" CPU test
(no big video loading is required for this test). If you can run Prime95
for hours without the PSU going off, that tells you the CPU is not
asserting THERMTRIP for no reason. Running the CPU at 100% with Prime95,
doesn't draw as much power as running Furmark would on a high-end video
card.

If your CPU is stable in this sort of test, that helps take the motherboard
off the suspect list a bit. There is a need to "isolate to the nearest
subsystem", and it isn't always easy to do that. When testing, right
up to the end there can be some uncertainty as to what is broken.

*******

Your LED results suggest the Power and Reset illumination are running
off +5VSB. The fact they remain illuminated, says the |5VSB portion
of the PSU is still operating. Just the main section switched off.

The PSU can switch off for two reasons. It can switch off on overcurrent
or overheat (internally detected PSU issue). It can also switch off
if the motherboard turns off the signal. It's the symptom correlation
that matters here - if the PS_ON# was going off at random, we don't
know what is doing it. If the supply appears to be going off only
under significant load, and the test in the first part of my posting
is passing, then it suggests the problem is the PSU.

So do the 100% loading CPU-only test. If the system always stays up
with that, and yet it fails on Furmark or other video activity, it's
most likely to be the PSU that needs replacement.

Proper PSU testers are not all that common (expensive), and if you
take your computer to a shop, they would simply swap out the supply
and retest. If the swap out didn't fix it, they'd then be focused
on motherboard (or CPU THERMTRIP). Watching which cases are failing,
helps eliminate some things and make it less likely that swapping them
will fix the problem.

Checking for the LEDs going out, was to see whether the +5VSB portion
of the PSU was winking out. You can overload that section with enough
electrical load, such as charging two iPads, running a USB lamp,
or an excessive number of USB toys. Since they starting running
USB ports of +5VSB, we have to be a bit aware of the consequences
of using too much USB power. At one time, the USB ports could be
configured to run off +5V (20 amps or more available), and then
that was one less thing to worry about. The +5VSB section only
has two or three amps, and the motherboard can be using around
one amp to start with. Leaving one or two amps for toys, which is
not much.

But what triggers the failure, is intended to narrow down
where the switch-off is happening. The symptoms still point
to PSU, and running more tests is to narrow it down a bit
better (to justify spending money on another supply, to test
with if necessary). I've used up all my supplies here, so I
have to pull one from a working system, to test with.

Paul