X170SMG Black Screen after Sleep

[Bidelloman]

Greetings,

 

The moment when last year I changed my stock x170smg's cpu and gpu from 10600k and rtx 2080s to 10900k and rtx 3080, sleep has been an issue regardless of OS ( tested on various linux distributions, windows 10 ltsc 2019, 2021, windows 11 ltsc) where I have black screen on wake. 

 

I also changed bios and EC, have one with subsystem ID support for the 3080, doesn't matter, sleep fails with black screen.

I changed gpu when the previous one stopped working, problem persists.

 

Reset CMOS and used stock optimised defaults BIOS settings, no matter, after 1 2 days max sleep issue returns ( occasionally it works, especially during shorter duration sleeps).

 

Full Specs:

X170SMG ( originally from PCSpecialist UK)

CPU: 10900k, delidded with copper IHS

Ram: BL16G32C16S4B.16FE Crucial Ballistix 2x16GB 3200MHZ (XMP), 2667MHZ stock frequency

GPU: 3080 mxm

Storage is comprised of 2 nvme ssds and 1 sata..doubt that this matters much to be honest.

Cooling mods: (though I do not think this matters much either) taobao fans, stock heatsink, PTM 7950 on both cpu and gpu, utp-8 putty on the rest, rockitcool copper IHS on cpu and LM between CPU and IHS.

 

There are only 2 combinations of ram slots that work ( with or without xmp) and those are RAM1+RAM3 (bottom) or RAM1+RAM4 (1 bottom + 1 under keyboard), both present the same issue. 

 

S3 SLEEP TESTS RUN THUS FAR:

 

Note: Bios settings, apart from those mentioned, stock, nothing else has been changed and have gone through many CMOS resets to ensure equal results, or as much as you can get with this fucking clevo platform.

 

SK Hynix set (2x8gb 2667mhz no XMP available) do work on any OS, any xmp or safe UV applied, no matter the power profile or anything.

 

Crucial Ballistix 2x16 3200 (XMP) CL 16:

-6 core 12 threads, -90mv stock UV, XMP, fails with uv

-8 core 16 threads, XMP on, fail on undervolt ( 90 or 50, values known to be stable)

-10 core 20 threads, XMP on, fail with or without undervolt, sometimes even without XMP, fail even when only 1 stick or even when trying other slots.

 

Current main ideas are:

 

weak CPU IMC/Voltage Margin - main suspect here currently

weak board / compatibility with higher core cpus with this fast a ram kit - less probable, in linux I've done various kernel stages tests to see the board behaviour directly, all of which passed in the most "extreme" conditions.

 

As far as I can tell, with various AI's worth of help and some technicians, we can clear everything else of blame, but I can't be sure as I am no expert. We do know how temperamental this board / pc can be with ram, especially with xmp involved.

 

If anyone has any suggestion...they'd be more than welcome, thank you

[harkaz]

Try installing the latest chipset drivers: https://www.intel.com/content/www/us/en/download/19347/chipset-inf-utility.html

Another suggestion would be integrating the latest microcode for the new 10900k CPU to your BIOS.

If all software/firmware options fail, consider a hardware issue (CPU or mobo).

[Bidelloman]

Chipset drivers like every other driver are new or updated, as I said, issue carries over various OSes.

Bios has been updated and is known to fully work with this machine and cpu gpu combo ( so does in mine to be honest if it weren't for this sleep issue ).

Motherboard also never had an issue with sleep with 2080s and 10600k and various ram models. 

 

[Bidelloman]

New Update:

 

I sorta simulated a 10600k and a 10700k with the number of active cores in the bios, and both have sleep working on all OSes without issues even with XMP on my crucial ram set and with both version of control centers ( clevo on windows, tuxedo on linux), however, adding what I know to be a stable during-active-operation undervolt (-50mv) breaks sleep instantly.

 

This is starting to look more and more like a weak cpu like @harkaz suggested, the other option is that the board is delicate and not robust enough, weird, considering other people have my identical config ( 3080, 10900k, same bios and ec) and do not have any of my issues.

[harkaz]

Cleaning the board using isopropyl alcohol may help. Also, use compressed air to remove any dust particles from the CPU socket.

[Bidelloman]

18 minutes ago, harkaz said:

Cleaning the board using isopropyl alcohol may help. Also, use compressed air to remove any dust particles from the CPU socket.

Both have been cleaned and checked many times. 

Did also a 1h sleep with all active cores (and xmp on) solely on battery and it woke up. But failed if i changed to AC power during sleep or wakeup.

Ultimate test would obviously get another 10900k, or kf or 10850k and repeat tests and see if they all work fine.

[harkaz]

Try with 'High Performance' power profile while on battery to see if system resumes ok. If the problem exists only on AC power, the PSU, the power socket, or even a cable could be implicated.

[Bidelloman]

Confirmed, it's these crucial sticks which do not play well with 10900k and the rest of the platform, least of all under xmp ( memtest errors etc). 

 

Tested a sk hynix set I had around, no problem even with hefty UV.

 

Tested crucial 2 sticks with a samsung 4gb stick i had lying around, no issues either.

 

I guess they are too much on their own, simple as that.

 

Increasing SA and IMC voltage offset breaks sleep completely regardless of it helping memtest / stability.

 

PC refuses to boot on many custom memory profiles, even if I try the exact same xmp profile at like 1.4v, it would not boot.