[snippage]
Place the ground of your multimeter on an I/O screw on the back panel.
I like to clip the black probe to a screw, so I only need one hand
to make measurements with the red probe.

The ATX main power connector, you can probe in the back of the
connector, and just barely touch the metal on each wire crimp in
there. This allows every voltage on the connector to be verified.
As long as an older power supply is used, it has nice colored wire
and this makes it easier to determine which rail each wire is
connected to.

If you scroll down to 50% of this web page, there's a table of wire
colors.

http://www.playtool.com/pages/psucon...onnectors.html

PWR_OK gray Pin 8 || || Pin 20 white -5
volts (optional) === "blank pin"

You can see in that example, every wire has a color like gray or
white or red or orange. And for the ATX supplies that adhere
to the standard, this makes debugging a lot easier. I have a newer
PSU with all-black wires and that's nothing but a pain in the ass.

This kind of extension cable can be added to the power path, if
you need some "wire colors" for inspiration. I might use something
like this with my all-black-wire PSU, during a debug session.

https://www.newegg.com/p/N82E16812198008

Paul

Thanks for the valuable hardware info. I recall checking 5V when the
Q9650 was plugged in. Even then 5V was unnatural. It would peak at 5V and
then descend. Only boot 3 recycles, not infinite.

When you're watching the voltages of an ATX supply,
keep in mind it has only one regulation loop, and a
rail can be influenced by:

1) direct loading
2) cross loading

Let's make up an example. If you're monitoring 5V and decide to
draw 20 amps, maybe you see 4.9V from the PSU. That would be
an example of direct loading.

Now, instead, load the 12V rail with a 20A load, while monitoring
the 5V. The 12V drops to 11.9V say. But the 5V rises to 5.1V.
This is cross loading.

The supply measures the mean of the 3.3V/5V/12V voltages and uses
that value in the control loop. If something is getting loaded,
the "whole PSU works harder". The loaded rail (12V in the cross load
example), still demonstrates it "feels the load". The output has dropped
to 11.9V. But because the "entire PSU is turned up", the 3.3V and 5V
rails end up higher than they should be.

The PSU design aims for no more than 5% cross load. Which is
practically the whole range for regulation.

The output transformer establishes outputs via turns ratio of
the single output transformer. The high frequency transformer has
winds for 3.3V, 5V, 12V. Due to the turns ratio, there is a very
strong correlation between 3.3V, 5V, and 12V as a group. This is
why a single control loop is sufficient to control it. Rectifiers
are placed on each AC output of the three winds, giving DC voltages.
Since the switching frequency is so high, relatively small electrolytic
capacitors can give a reasonably smooth clean DC output.

That is the description of a "traditional" 0.7PF 70% efficient power
supply. I have several ATX PSUs in older computers that follow this
description of behavior. For more info, see this hand-drawn schematic
of a commercial PSU from slightly before 1999 perhaps. Don't quote
everything this does verbatim, because some of what is in here, isn't
exactly the way most are done. But it does illustrate the architecture
a bit at least. And how cross-loading could be present, as a behavior.

http://www.pavouk.org/hw/en_atxps.html

More modern supplies have changed that a bit, and it affects
the +5V behavior. Now, the architecture is called "double forward
conversion",
the conversion goes "120VAC to 12VDC", "12VDC to 3.3V/5V". There could be
a cross loading effect between 3.3V and 5V, which sit on a separate SMPS
regulator
board. There should be less of an effect, between 12V loading and 5V
output
voltage. I might only have one supply here of that architecture,
and I've never needed to measure any voltages on it, so can't
provide any first hand experiences in terms of "wandering voltage
values" with it.

This is just a quick capsule summary of what you might see or
expect from the primary DC voltages in the machine. Yes, the
voltages do tell you things about what the machine is doing.

In cases where the PSU is defective, sometimes the problem
is obvious. As an example, my very first PSU from the 1999 year PC,
it failed with "weakness". The supply still works today. However,
the 12V rail will only power a 0.1 ampere fan. If you connect
two fans to it, the voltage drops to around 6V or so. The supply
cannot run a processor. You can't say the thing has "totally failed",
but the supply is completely useless in any computer. It has
a "strength problem". Visual examination of the inside of the
supply, shows no leaking caps, and it's before the capacitor
plague era, so that's not an expected failure for that one.
I have other supplies that just conked out because their
safety circuits detected trouble (internal overload).

Supplies come in all shapes and sizes, some with rather bizarre
designs inside and strange behaviors. There is the 200W Bestec that
when it blows up, the 5V output rises to 9V, ruining hard drive,
optical drive, and other stuff related to 5V rail. There aren't
many other supplies that behave that badly, but it's not for
a lack of trying.

Paul

Thanks Paul. I found a pin on the 24-pin ATX power cable that gave steady
5V, each cycle. The MOBO would try to POST, give steady 5V, then shut down.
Ad nauseam. I don't think the extension cable would give a fruitful line of
inquiry.

I looked into whether my eBay purchase was covered by money back guarantee,
but the seller has gone out of business.

I found two GA-EP45-DS3L boards for sale on eBay. The best looking offer is
from the US, "factory refurbished" and fitted with an Intel E8400 CPU. I'll
wait until the seller responds to my email and then possibly buy it?