What would cause a computer to simply stop?

  The major components are a Gigabyte mobo, Antec Earthwatts 380 watt
PSU, 2.33 gig Core2 Duo, XP Pro, 320 gig WD Caviar PATA C: drive, 320
gig WD Caviar SATA aux drive.

After replacing some bad ram a couple of days ago, the machine was
running like the proverbial Swiss watch until today when it suddenly
winked off. Just quit as if it were unplugged. This has never happened
before. I tried turning the PSU switch off and on, tried disconnecting
the PSU main board connect and reconnecting. All that would happen
would be that if I turned off the PSU switch for a minute, turned it
back on, hit the ON button, the case fans would nudge briefly and then
nothing.

Took the PSU out, took it to the local computer shop, they said it
tested good.

Put the PSU back in, hooked up enough components to make the machine
run, still nothing other than the aforementioned case fan nudge.

Took out the C: drive, stuck it in another machne, it fired right up,
no problems.

Put the C: drive back into original machine (that was exhibiting the
symptoms) and now it does fire up. Hmmm....

Only thing is, after a few minutes of browsing the machine winks out
again -but- this time it spontaneously reboots.

Anoter thing I noticed is that with the aux SATA drive disconnected
the C: drive wasn't showing up in bios, even though XP was coming up.
Admittedly, I didn't change the jumper on the C: drive after
disconnectng the SATA drive. When I reconnected the SATA drive, both
were recognized in bios.

After all this, any theories as to what might be going on?

Thanks for all input.

 Doc wrote:

OK. Let's review.

You've been inside the machine. You changed RAM.

Occasionally, a stick of RAM will have a bad capacitor or a sloppy solder
joint, and "stuff burns". Since your computer was running like a Swiss Watch
after the install, that probably isn't it. Generally, when that happens
(infrequently), it happens to brand new product, at first switch-on.

You may have bumped something while in the PC.

Verify the ATX 12V 2x2 power connector is firmly in place.

Check that the CPU heatsink/fan is making good contact with the CPU,
and that the fan is spinning. Set the fan to maximum speed (i.e. if
it is currently under Automatic Control, set it to manual and let the
fan have full 12V). Now, try the machine again, and see if extra
cooling effort helps.

A computer will wink off, if THERMTRIP is triggered on a CPU overheat.

But your computer rebooted, rather than just shutting down.

Rebooting suggests you next check for bad capacitors around the CPU socket.
Check the aluminum cylinders with plastic sleeves. The top of each, should
be flat. On electrolytics at least, there are pressure relief seams. If
they burst, electrolyte leaks out. The power circuit the cap is part of,
them becomes unstable and spontaneous reboots can happen.

A number of newer Gigabyte motherboards, use Polymer caps, rather than
electrolytic, and then the motherboard bad cap theory is out the window.

The same thing can happen inside the power supply (bad caps). They don't use
Polymer caps in the power supply, so it's always a candidate for analysis.

If the power supply is under warranty, opening the lid voids the warranty.
A sticker will usually cover one of the four screws on top. If it is outside
the warranty period, remove the four screws that hold on the top of it.
Have a look for bad caps. (You're looking for orange goo, as shown. This
power supply is bad, and looks very similar to my failed, older model Antec.)

http://upload.wikimedia.org/wikipedia/commons/8/84/PSU_Caps.jpg

If you entirely remove the power supply from the PC, then connect
the ends of a paper clip between PS_ON# and an adjacent COM pin,
you can turn on the supply. The shop says it tested good. We could
either take their word for it, or test it again. I use a load
box for that, which places a light load on the supply If it ran for
a couple hours, without complaint, with a 100W load on it, I'd probably
pass it.

If you don't have a load box, have no interest in building one, then
the next best thing is to purchase a spare power supply and install it.
I try to keep a 500W around for fun and games. If I have my suspicions,
I pop it in for a trial. Since power supplies are always failing,
you'll be using the spare, sooner or later...

Another thing - take a look and verify the power supply fan is still
turning, and that the power supply is being cooled.

    Paul

  On 3/22/2011 10:06 PM, Paul wrote:

  I use a load

Hey Paul

Care to tell how to build a load box?

 TVeblen wrote:

Blast from the past :-)

http://groups.google.com/group/alt.comp.periphs.mainboard.asus/msg/19647caf2c65504b?dmode=source

Now, when I built mine, I was able to get an ATX connector shell and crimp pins
at a local electronics store. Since then, another store offered an ATX
extension cable for sale at a reasonable price, and if you chop that up, it's
a better starting material, because you get a prewired connector (no crimping
needed). Just strip the wires and solder the resistors to it.

One of the enthusiast sites built a "real one", with banana plug holes so you
could
adjust the load on each rail. If you have, say four resistors in parallel, you
can
set them up such that external wires are used to connect them to the load. A
load box
you might have in a lab, would likely have a rotary switch to switch in various
loads,
bu it's too hard to find rotary switches. The banana plug route is more crude,
and still
costs a fortune to put together.

Mine is a fixed load, and not a stress test. If I had a power supply with leaky
caps,
perhaps I could connect it, and see if the thing "stays up" for any period of
time.

At the factory, some of the load testing is dynamic, such as load dump tests or
the like.
There are specifications on the ATX supply, for how long it should take to
respond to
a dynamic event, but the static load boxes I'm talking about, aren't designed
for that.

My load box is mainly to protect me from an "infant failure" on an ATX supply.
When
I get a new supply, I put it on the load box for two hours, and check the rails
with
a multimeter. That isn't a guarantee all is well, but intended to weed out a
supply damaged in transit.

    Paul

 On 3/23/2011 1:21 PM, Paul wrote:

http://groups.google.com/group/alt.comp.periphs.mainboard.asus/msg/19647caf2c65504b?dmode=source

Thanks Paul
A fun weekend project!