On Mon, 8/25/2025 3:32 PM, David Samuel Barr wrote:
Hello again, all.˙ Some might vaguely remember
that back in October 2022 I started a thread
about doing a new build to replace (or as it
turned out, have to supplement) my 2007 build
that was similarly guided here on which I'm
writing this.˙ With your advice that morphed
from v3.0 to v4.0 a month later but it wasn't
until a year after that (Nov 23) that I
actually had time to buy the parts (by which
time I had also updated and revised some of
the components; see below) and then not until
this month (Aug 25) to actually piece them
together over the last two weeks.
CPU: Intel Core i5 13600K
CPU Cooler: Noctua NH-U12A
Motherboard: Asus Prime Z790-A WiFi
RAM: Crucial 2x16Gb DDR5-5600 UDIMM
Storage: WD SN850X nvME M.2 PCIe4.0 1TB SSD
Optical Drive: LG WH16NS40 CD/DVD/Blu-Ray
Power Supply: PC Power & Cooling FireStorm 750
Case: Cooler Master N400 with 3rd fan added
Monitor: Dell S2725QS
Keyboard & Mouse: Redragon K509 & M608 (just
˙because they were a freebie from MicroCenter)
There were some challenges in getting
everything in its proper place but eventually
I did so, plugged it in, powered it up and....
nothing.˙ Well, not nothing, the motherboard
power LED lit up and the PSU and fans all
started running but there was no boot beep
and the display said there was no HDMI signal
coming from the computer.˙ I tried it with
and without plugging in the USB drive which
has the updated BIOS file to be flashed but
it didn't work either way.
This being an ASUS board its POST diagnostic
Q-LEDs are CPU (Red), DRAM (Yellow), VGA
(White) and Boot (Yellow Green).˙ Upon
starting up the computer the red CPU light
turns on then immediately off but then the
yellow DRAM light goes on and stays on (the
next two lights don't light at all).˙ I've
removed and reseated the DRAM sticks, swapped
them, tried using just one.˙ Same result every
time.˙ (I am following ASUS's instructions to
use channels A2 and B2 for two sticks.)
Anyone got any ideas where to go from here?
(No, I don't have any additional parts here
to swap out with these nor any meters or other
test equipment.)
Given the constraints in your description, about the
only other step would be to remove, examine bottom
of CPU, and reseat CPU, as the memory path to CPU
involves contacts on the bottom of the CPU. Long ago,
there were "burn patches" on the bottom of the CPU,
where the contact compression force was not correct
and the pins didn't bite the LGA pads properly.
Your cooler is 1220 grams, so is below the limit for
mass. There are some larger coolers for which I would
use support columns under the fin banks so there is
less torque applied to the socket area.
It could be something as simple, as a failure on
the DRAM core regulator. Of the two motherboards I've
had fail, one of the boards the issue was traceable
to some small regulator on the motherboard. We always
have to look at the powering on the board, for clues.
The silk screens on normal boards, never mark scope
points on important voltages, for us to check.
One company, Diamond Flower (DFI), put some "test points"
on their motherboards, places you could hook a meter and
verify voltages. But that made too much sense, and as
far as I know, DFI is no longer in business.
It is probably not your CMOS battery, although the number
of years that have passed, the timing of the failure, it is
tempting to speculate on CMOS battery. But because the CPU
LED went out and the DRAM LED triggers, I think that is intended
to hint that some BIOS code ran and it's not related to the CMOS
battery.
In the old days, you would use a PCI POST card to analyze what
is going on. This is an example of what is offered today, you can
see this card uses an Actel FPGA for listening on the bus and
it has interfaces for two bus types (flip card over for usage on
modern motherboard). What you would do with it, is run the motherboard
with this in slot #1 (by CPU), then see if the digits stay at 0x00 or
0xFF value, or if any value other than that shows up. Any sort of
activity, changing values (0x23, 0x79, ...) indicates the CPU is alive.
It would be almost impossible to get a full table of values, to
actually decode the "progress indication" values seen, so this
would be a mere test for activity and not a quantitative analysis.
Without reviews of these, we do not know whether the BIOS has the
I/O instructions to drive the display digits or not (the BIOS plays
a part in making this card work!).
https://www.amazon.ca/Motherboard-Analyzer-Diagnostic-Laptop-Desktopmode/dp/B0874SXXTK
We have to assume the QLEDs are an honest effort, and "being stuck
in DRAM state" is not the result of the CPU not executing a single
instruction. If they cut corners on the design of QLED, that
would degrade our ability to debug the problem.
I don't own one of those POST cards, so I would head to my computer shop and "pay for a diagnose". At one time, that was about $25, and I can
save you the trip as they'd just say "it's your motherboard"
and charge you the $25 :-)
But pull the CPU and examine the bottom and look for burn marks.
There are two socket companies, Lotes and Foxconn, and Foxconn
only had one bad socket, and it was well before your board was made.
For anything else, the easiest assumption given symptoms don't change
as the DRAM is moved around, the DRAM regulators may have suffered a failure.
The board has Power_Good logic, but there is nothing saying that a
regulation stage cannot lie about its function. Power_Good is for
coordinating the startup of the board, and when all the Power_Good
signals are asserted, the board comes out of RESET and starts the
POST sequence. At least in terms of the Power_Good, it's telling
you that the status indications were all good. But for the
onboard regulators, there would not be much present there to verify
correct operation. It's not like the regulators have window comparators checking the voltage.
If you had yet another computer, you would take your RAM sticks
over and check the RAM sticks there. If neither RAM stick would
even allow the other computer to start, that might be an indication
the RAM was blown by the wrong voltage being applied.
Modern Asus boards are much more sophisticated on their regulators, than
in the "op amp" days, where the regulators were "barely functional". One
day, while checking voltages with a meter, there was a chip that
was supposed to output 2V, and when I read it with a meter, the
meter read "2.000" and that put a smirk on my face. As in the past,
the value could have been off by 10% and we would not have been
surprised. When the regulators work now, they work "very well"
(they use band gap references, and quantum mechanics properties). On
occasion, some of these run at 100C and the chip doing the work
really deserved a tiny heatsink. Someone checked their motherboard
with an infrared camera, and located the "hot spot" in that case.
The chip in such a case, may be able to run at 100C, but it may get
tired of doing that after a while.
Summary: If the QLED is working properly, it could be a DRAM power issue.
If the QLED takes shortcuts and it always "blows a DRAM code",
then there could be an issue with the ATX power supply. But given
the limitations on materials, disassembly and reseating are the
thing to do before discarding the materials. You can even take
the motherboard out and run it on top of a piece of cardboard
while testing.
You can ask a shop, how much time they will spend on a diagnose.
If they charge an hour labor, and claim to do one hour of
diagnose, those guys are pretty efficient at fault finding.
For example, at my work, I talked to a guy on the line who
used to diagnose my electronics design, and he could find
faults in it twice as fast as I could. I was suitably impressed.
He even had the same name as me :-) They can swap your DRAM into
one of their boards, swap your CPU into one of their boards, and
verify the components that way. And that's how the answer will
come back "it's your motherboard", as they've tested the CPU and
DRAM and those worked OK. They can check the power supply, with a
24 pin tester.
Paul
--- MBSE BBS v1.1.2 (Linux-x86_64)
* Origin: A noiseless patient Spider (3:633/280.2@fidonet)