X870E Nova D4 on restart/cold boot

Not a bad way to go really, so long as you are still within your warranty just
use it. If something dies you can get it replaced (for certain) under warranty.
If the symptoms worsen just RMA the CPU and/or motherboard. Odds are pretty high
that a BIOS update might resolve the issue sometime soon.

Thanks again for all your detailed info.

Hello, i have some actual news.

So yesterday had a very dedicated debug session (7hr+, well into early morning hours...)

Memory speed, voltages, etc perhaps play some role, but it ain't down to that.

Sometime in last 3rd of that long session, i have OC'ed the memory to 6000 MT/s @1.32V - same timings as XMP-1 profile from it. (kit is rated for 5600 MT/s, 1.25V).

Also, i have noticed sometime tRDWR ends up different (1 clock cycle) between A2 and B2 (not always).

Memory is still @6000, currently little under one hour in y-cruncher Zen5 stress test.

Most of those 7+ hours yesterday where spent doing the following:
1) restart from windows
2) restart to UEFI (from windows) using ASRock utility
3) Save and exit from UEFI (restart from UEFI after a change)
4) Shutdown from Windows (S5 state, plugged in, 5VSB) -> Power on (boot)
5) Shutdown from Windows (S5 state) -> Flip switch on PSU (G3 state) -> Toggle switch on PSU -> Power on (boot) state
6) Press RESET key whenever 4d error occurred
7) Flip switch on PSU (G3 state) whenever 4d error occurred


That is what i have been testing and noting down if there is a pattern when 4d occurs. Right towards the end i realized something which i actually managed to prove, but more on that later.

So:

1) restart from windows:
4d occurred 0 (ZERO) times.

2) restart to UEFI (from windows) using ASRock utility:
4d occured 0 (ZERO) times.

3) Save and exit from UEFI (restart from UEFI after a change:
4d occured 0 (ZERO) times.

4) Shutdown from Windows (S5 state, plugged in, 5VSB) -> Power on (boot)- MOST of the time unsuccessful, 4d occurred in over 80% of cycles carried.

5) Shutdown from Windows (S5 state) -> Flip switch on PSU (G3 state) -> Toggle switch on PSU -> Power on (boot) state:
MOST of the time successful, 4d occurred in about 30% of these type of cycles

6) Press RESET key whenever 4d error occurred - works sometimes.

7) Flip switch on PSU (G3 state) whenever 4d error occurred:
Almost total success, 4d re-appeared less then 5% of these cycles carried (these i was forced into).

For all of the tests, all timings were set manually - but taking "healthy" XMP-1 or values known good training had produced (readout by ZenTimings tool).

2 3rds of the time memory was in XMP-1 mode, rest OCe'd (still running same OC - made no observable difference to 4d randomness).



Final notes.
While testing, i got to learn when training will fail / 4d appear - VRM and CPU fans raise their RPM (not to max, but enough to be noticable).

Most important and somewhat confusing note. I re-used one of 120mm fans i had lying around. Non-PWM fan, spins very fast. I didn't measure it's current draw, but it was plugged into AIO_Pump
header, set to DC mode. Once i unplugged it, 4d occurence became way less.

I have measured 12V, and it's bang on startup, 12.05 rock stable (probed somewhere).

Enabling Deep sleep mode also dropped occurence of 4d error a bit more so - 4d occured in about 40% of cycles carried with this ACPI setting.

If i have external 3.5'' USB HDD plugged in, i see code 99 persisting for a bit longer, and 4d again trips more often.

I have plugged Corsair PSU USB cable (terminated in the form of USB 2.0 header) - into USB 2.0 header of xHCI Prom #1.
This seems to also decrease occurence, otherwise this Prom21 chip is in power down mode until something is plugged to it.

I can't draw any conclusion, but empirically as if some power-hiccup or current limit occurs after cold-start and training, and 4d gets tripped - this really looks like it has nothing with voltages set in BIOS,
speed options, timings...

A question. How much does mobo rely on grounding via standoffs? Case i have was painted inside and is no longer conductive -
only partially where standoffs screw in (due to threads chewing thru paint and clear coat).

3A fan header specification is per-header?


Following BIOS settings were used, but for all i know - it doesn't seem like there is defacto settings that would completely remove this oddity.

PS. Wheter "Fast boot=Disabled"; "DF CStates=Disabled"; "PCIE Devices Power On=Disabled" have any influence on 4d occurence - i doubt and can't tell - my test data ain't granular as much.

BIOS Settings list part 1 - on the left are settings used, on the right default BIOS values.


BIOS Settings list part 2 - on the left are settings used, on the right default BIOS values.


BIOS Settings list part 3 - on the left are settings used, on the right default BIOS values. (memory OC mode, Active memroy timing settings, ignore since timings used were ones from OC Tweaker)


BIOS Home page..Note 6000 MT/s for this kit.. Very relaxed timings - basically whatever was in XMP-1.


DF Cstates


OC Tweaker


Y-Cruncher, Dimm A2 and B2 temps:
[img]https://i.postimg.cc/XqSZ7SfM/09-memory-temps-ycrunch-test.png[/img

Y-cruncher, CPU temps


Spec from HWinfo

This is brilliant info, thanks dom0xDee4. Hopefully ASRock will see this and get
something useful from it.

To answer your questions:
Yes, 3A is per header as far as I know.
Grounding isn't critical but certainly doesn't hurt, particularly if you are experiencing
electrical interference for some reason. Adding a grounding wire on a spade connector
between the board and the case (being careful not to short anything around the
mounting hole/ground plane) might be worth giving a shot. I would also ground
the PSU casing to the PC case, that one is easy and safe to do.

I am not sure if it will help but given your suspicion of some kind of power issue
it's worth a shot. It may also be worth trying another PSU, just incase you have
some ripple happening that isn't easy to measure with a regular multimeter. You
typically need an oscilloscope to measure the ripple on power rails.

EDIT: That is the case - Fast Boot ignores "Setup timeout delay".

Question 2. Over the interwebs, Code 15 = Memory training. Is there more to it? Since enabling the Setup timeout option and disabling FastBoot - as you've seen in previous message, code 15 is shown for as long as the delay set in BIOS.