Re: Odd PSU problem

philo wrote:
> On 9/28/20 2:08 PM, Paul wrote:
>> philo wrote:
>>> Had a power supply die, so I replaced it with a used one I had that I 
>>> had marked "good."
>>>
>>> It booted to the bios and I just set if for default but it never 
>>> loaded the OS. It just returned to the bios.
>>>
>>> HD was detected and set for first boot option.
>>>
>>> Though I was sure I had a legacy Win10 install, I did try UEFI but 
>>> still could not get past bios.
>>>
>>>
>>> Finally a cracked open a new in the box PSU and all worked fine. I 
>>> assume there was some low voltage.
>>>
>>>
>>> Both sup[plies went to a metal scavenger
>>
>> You could have tried a memtest floppy.
>>
>> One of the slightly older single threaded versions.
>>
>> Then get out your multimeter and checked supply voltages.
>>
>> I suppose Power_Good was signaling a problem.
>>
>> I'd probably have the hard drive disconnected while
>> working on it at this point. To reduce the chances of
>> damaging it. Disconnect power and data.
>>
>> Or, you could put a dummy load on the PSU to test it out.
>> I have a dummy load, but it is "lite" and hardly a test.
>> It's enough though, to tip over my "weak" PSU, which is
>> the original Heroichi (HEC) bought for my first PC. It
>> hasn't blown up, but that supply can only run one 12V
>> fan before that's too much for it. Not enough power
>> to run a CPU. That supply is from well before the badcap
>> era. Haven't thrown it into the chipper yet.
>>
>> I like about 100W for a dummy load. As higher loads cost
>> money :-) And my electronics store is gone, so easy access
>> to their power resistors is now gone too.
>>
>>     Paul
> 
> 
> 
> I tossed it so I would not fool around trying to find the problem.
> 
> What I really need to do is a get a 24 pin PSU tester
> 
> 
> The 20 pin tester is not a good way to test a 24 pin PSU

The extra four pins on the end, are "ampacity extensions".
No new rails. No curve balls.

The 20 pin tester should have "indicators" for all rails,
like window comparators to check the voltages.

The key wire on the 24 pin, is an extra yellow 12V wire, which
improves the room for current flow for +12V used by PCIe slots.
And this is only an issue if the computer is "filled with NVidia
6600 cards" - those draw around 4.2 amps from +12V at the slot.
And for a pair of those in SLI, that's why you want the 24 pin providing
system power. That gives a total room of 12A for current flow for
PCIe slots. Enough for 8.4 amps for a couple 6600 cards. Higher power
cards, generally only draw 2 or 3 amps from the slot, the balance
is heavily drawn from the end-card connectors.

There are no new functions on the 24 pin. All the logic-type
signals are the same. Just a few extra "parallel wires" to carry
current. PSU testers generally don't have current flow
functions - the cheap ones just check voltage level (high or low).
Checking any red wire, is enough to quantify the voltage on a
red wire.

Now, if you're looking for continuity testing ("are all the wires
in the loom connected"), that's a different animal. A rack mount
lab tester probably checks for that. A $20 home tester is unlikely
to care.

And when a home tester presents a "load" to the PSU, it's usually
just on one rail. The 3.3V rail is the one with remote sense (the
little brown wire) and is the one that collects feedback
right at the load. But I don't think that's where they put the
load resistor on the tester :-) It might be on +5V. I've never
owned a consumer-grade PSU tester to know - I use a multimeter
instead.

    Paul