philo wrote:
On 02/21/2018 11:21 AM, Paul wrote:
philo wrote:

Thanks Paul.

It seems to be something to do with the mobo's behavior with a PSU
other than the original.

The mobo has a 20 pin power connection and the original supply had a
24 pin connection (with the 4 pin connector left dangling of course.)

I replaced it with a simple 20 pin PSU


I know that there the 20 pins of a 24 pin PSU are somehow a bit
different than that of a straight 20 pin PSU because when I test 24
pin supplies with my 20 pin tester, one of the red warning lights
goes on even though the supply is good.

The four leftover pins on the 24 pin are for "ampacity". You might not
find that in your dictionary, and some staff at work introduced me
to the (apparently made up) word.

There is no functional difference as such between 20 and 24 pins.
The 24 pins, the actual useful benefit, is the additional yellow
+12V wire, which was badly needed. That raises the current available
for PCIe slots from 6A to 12A. Before the 24 pin came along, some
companies were putting a 4 pin Molex on the motherboard, to carry
a bit more current for +12V.

There are two tables here, you can use to compare a 20 pin to
a 24 pin. The difference, the extra four pins, look like this.

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

Description Color Pin Pin Color Description

+12 volts yellow 11 23 red +5 volts
+3.3 volts orange 12 24 black ground

There are already a fair number of 3.3V, 5V, and Ground wires
so adding one more of each isn't as big an improvement, as that
Yellow wire for +12V is.

The worst card for slot power, was a 6600 video card, with 4.2A
of 12v slot power. Two of those would exceed the conservative 6A
limit for the single Yellow wire on the 20 pin. You could safely
run two 6600 cards on a 24 pin connector, as the 12A available,
exceeded the 8.4A load.

When you buy a $1000 video card, those only draw around 2A from
the slot. It's the mid-range cards, the one without a PCIe power
connector on the end, that push towards the slot limit of 12V @ 5A max.

It's really hard to guess where the 3.3V and 5V current flow
levels are today. But you'll notice that the "gigantic" modern
80+ supplies only offer 20A on 3.3V and 5V. And it is hard to
get a supply with more (there are traditional design PSUs that
go up to 40A on the low rails).

Some video cards have used as much as 3A on the 3.3V, so two video
cards would be 6A out of 20A. And the chipset is pretty thirsty
and could well be running from the wrong rail as well. There will
be some things that can run directly from 3.3V. And a lot of lower
voltages (1.2, 1.5, 1.0V) that you could run off 12V if you had the
ampacity to do it. Switching regulators (like the switching supply for
the DIMMs) get less efficient at lower voltages (but that's never
stopped anyone in the past).

With the "sleeved" cables of modern supplies, I can't get my
ammeter jaws around stuff, to characterize my newer machines.
And the new machine, the heatsink is so big, I can't even get
my hands inside to seat half of the DIMMs, and have to snap
them into place with a sort of dowel to press down on them.
Not the nicest place to be doing power testing. No room
for anything. And spinning fan blades to nick me if I get
too adventurous (VCore has a fan on it now).

Paul





I am familiar with the portmanteau"ampacity" but shy away from using the
term.

It too much reminds me of laymen who use the term "amperage" rather
than "current."



At any rate, I decided to have a look at all the spare power supplies I
have in my workshop and the older 20 pin supplies have a white (-5v)
wire at pin 18.

My tester which is evidently ancient gives all green lights on those
supplies.

I have some newer 20 pin supplies which do not have that white wire
which corresponds to the "missing" wire at pin 20 on the 24 pin supplies.


My tester gives an over-all fault (red) for those supplies but all other
green lights are lit up but one.


Now that I have the machine all put back together and working I don't
think I'll fool with it anymore, but one of these days when I get a
chance I'll see what can cause a fault on that old Patriot UPS I have.



Though I always use 24 pin supplies with 24 pin motherboards, in this
case, since the mobo was just 20 pin, I figured it would be a good way
to use up one of my old, 20 pin supplies.

The need for -5V was removed from the standard years ago. The
dates can be a bit deceiving, as the industry could see the writing
on the wall, before it became official. So even though spec 1.3
says "year 2003" on it, mobo designers stopped using -5V before that.

Spec 1.1 has -5V on it. 20 pin connector.

http://web.archive.org/web/200304240...12V_PS_1_1.pdf

Spec 1.3 has -5V removed. 20 pin connector.

http://www.formfactors.org/developer...X12V_1_3dg.pdf

Spec 2.2 introduces... 24 pin connector.

http://www.formfactors.org/developer...public_br2.pdf

Formfactors.org belongs to Intel.

*******

My P2B-S (year 2000???) has a voltage monitor input for -5V, but
no circuitry on the motherboard uses -5V. The two RS232 chips use
-12V for the bipolar RS232 output swing. Even my P4C800-E Deluxe
(a P4 mobo) didn't use a Maxim "pumped" RS232 and the TI 75232
chip on that one continued to use -12V. That's about the only
reason to still have -12V. You can use -12V for op amp circuits,
or maybe for a little 2W power amp for passive speakers, but other
than that, there aren't a lot of excuses for -12V. And the -5V
back in the day, was used for digital chips. And all those chips
stopped using -5V. For maybe a year or two, you could build a
design purely out of +5V circuitry, but then the industry moved
on and new voltage requirements were introduced. Today, a Southbridge
could have as many as five rails on it (with many of the voltages
made by circuits on the motherboard). Even a NIC is dual rail, with
maybe 3.3V for digital I/O signals, but 1.8V or lower to power
the core circuits inside the chip.

Core logic today can run off 1.0V. They even make op amps that
run off voltages like that. I don't expect they'll go too much
lower. The 200W video card you own, the core of that runs at
around 1.0V. And the voltage regulator has to provide 200A
of current, and the copper planes in the video card have to
be able to withstand a 200A flow.

*******

You might notice that one pin on the ATX PSU, has two wires.
That's for remote sense of 3.3V output. That remote sense leads
to tighter regulation on the 3.3V output. To make the ATX PSU
supply 1.0V to the motherboard say, you'd need to run the
remote sense right to the load, to do a good job. This is one
reason that power supplies will not be making voltages lower
than the 3.3V they make today. Tolerances. And the cabling
that results from doing it that way, would be quite inconvenient.

Paul