Clear Linux on BD790i

I acquired a Minis Forum BD790i ITX motherboard and I have installed Clear Linux B42100 via the live desktop ISO and selected the installation.

The MinisForum BD790i has the following specification:
AMD Ryzen 9 7945HX (16c/32T)
(2) PCIe Gen 5 x4 SSD slots
(1) PCIe Gen 5 x16 slots
(2) 48Gb Mushkin DDR5-5600
(2) Samsung NVMe Gen 5 x2 SSD
(1) Intel Dual Port 10Gbe Network Adapter in the PCIe slot

UEFI:
Mode = Performance
Secure Boot = Off
Long Boot = Yes
All other options were default.

After installing the Clear build and booting, I was doing a control function check (devices, sound, networks, etc) to make sure there were no devices disabled from lack of support. (None that I can find)

I opened Firefox to check on a feature and actually had 3 tabs open when suddenly the screen went black. A few seconds later the BIOS Logo appeared and the Clear login screen reappeared. We are talking less than 6 to 7 seconds. Logged back in and my terminal cache was gone. Firefox reported an aborted exit.

I checked the boot logs and while there were several messages like thermald warning me that the boot CPUID is unrecognized, it seems when the system restarted there was a core dump in progress.

The log pull showed this:

Sep 08 18:34:40 clr-3e41015f01014fefb614cc118803ca84 systemd-coredump[1321]: [🡕] Process 959 (ibus-daemon) of user 311 dumped core.

Sep 08 18:34:40 clr-3e41015f01014fefb614cc118803ca84 systemd[1]: systemd-coredump@0-1313-0.service: Deactivated successfully.

6 stack traces of threads were identified in red. I would include them here if I knew a way to post without the formatting going awry. Otherwise I will post if someone wants to see them.

I will continue with the testing and see if it arises once again.

What’s your current CL version? You can check using swupd info

As noted in the first line…

Build 42100

42100, I don’t exactly remember, but one of the stable versions from that time was 42150. 42160/42170 had issues. Any reason for using 42100 in particular?

When I test Clear Linux, I always pull the live ISO recommended on the download page. As of the moment it was pointed to 42100.

I have a working theory for the core dump.

The Ryzen 9 7945HX is a mobile CPU but installed on a ITX board, sometimes called MoD (mobile on desktop). The UEFI is set to “performance” which is supposed to allow the CPU to reach its peak TDP (around 54W).

However, I think it took the dump when Clear started its power save steps. I changed the power mode on Clear to “always on” and the core dumps stopped. I also checked the default EPP from the install and it showed powersave (performance).

I am not saying that resolves it, more testing required, its just that I think that out of the box the UEFI thinks its a laptop and so it uses an aggressive powersave scheme.

I had a related issue with testing the 4900HX, another mobile CPU being used in a mini PC. The UEFI was taking the power on Core 0 too close to its fabrication tolerances and hanging the unit. Raising the power floor just a fraction of a watt stopped the hangs.

Testing with UEFI power modes, the EPP settings along with the OS settings, to see if there is a combination that is consistent.

I am going to wait until the 6.11 kernel is in a release for my next round of testing. Being a mobile CPU using the performance boost setting may be impacting the system. This patch (below) appears to be targeted at some of the areas I am looking at.

“AMD Core Performance Boost handling is added to the AMD P-State driver along with Fast CPPC support for select Zen 4 and newer laptops”