T. Ment wrote:


So I swapped the K7S5A for a motherboard which has the 12v connector.
Now my multimeter reads 5.16v on the USB header pins. That's more like
it. To make sure, I swapped again for a third motherboard (12v). Same
result, 5.16v.

My $20 power supply is OK with a 12v motherboard. I'll match the K7S5A
with a better power supply.

And that's the cross-loading effect.

Balancing the loads leads to closer-to-correct
regulation.

If one rail of a three-rail supply is heavily loaded,
the fact there is only one feedback loop means
two rails will be high and one rail will be low.

Only a few supplies used separate circuits for
the rails. The concept never caught on. I think
the regulation on those could be in the 2% or so
ballpark.

There have also been a few supplies with four separate
12V rails. CPU1 and CPU2, motherboard, PCIe, some sort
of split like that. You had to make sure on those, that
the rails that were independent, didn't get joined
by accident. (That's because they're only nominally
12V, and if they are slightly different potentials,
a large current can flow between them. And also,
the ATX design is not push-pull, it's only push, and if
a "rail goes high", nothing actively works to bring
that voltage down.)

On the motherboard, there is only one push-pull
regulator, and that's the Vtt regulator for terminating
DIMM buses. Plus or minus two amps flows from that
one. It either has to sink 2 amps or source 2 amps,
on a nanosecond level basis.

You've done a good job by selecting a 12V Athlon board.
When I told people I had a board that ran the CPU
off 5V, not a lot of people believed it.

There are some other goofy practices. Biostar, on a
number of motherboards, joined ATX12V 2x2 12V to
the motherboard yellow 12V wire. One is 12V1 and the other
is 12V2. Now, because a lot of ATX supplies power
12V1 and 12V2 come from the same transformer, there is
no potential difference. It's only when you use
a supply with entirely separate 12V1 and 12V2, that
this would be a bad idea. I never did see an explanation
from Biostar, as to why this was a good idea.

I think the evidence for this came, when a Biostar owner
of a P4 board, accidentally forgot to plug in ATX12V
2x2, and the CPU and motherboard booted and ran without
it. That's the evidence they're joined on the motherboard.
I can't think of any other brands that do this.

If you need help on connectors, this site has a page on
ATX supplies.

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

Paul