FW16 7840HS CPU/GPU throttling aggressively with PROCHOT_CPU, PROCHOT_CPU

Heh, I just posted about this yesterday to a github issue.

I’ll paste the contents from that post to here, quite a long post

I overall think it’s a BIOS issue

==== COPY PASTED ===

Framework 16 AMD Ryzen 7 7840HS using Radeon 780M
RAM/Memory: 32GB ( 2x16GB ) Framework
NVME 2280: Western Digital SN850X 2TB - Firmware 620361WD
NVME 2230: Western Digital SN770M 2TB - Firmware 731120WD
BIOS: 3.05
Linux firmware: linux-firmware d864697f
Gentoo Linux 2.17 ( Linux 6.15-rc1 mainline realtime, compiled by clang 20.1.2 march and mtune set to znver4)
KDE Plasma 6.2.5 Wayland

I’ve also tried this on Linux 6.12.22, 6.13.10, 6.14.1 and now 6.15-rc1

jcdutton ( user on framework github issues )

I’m not sure where to report this but I thought of you because you mentioned something about the EC slowing down the CPU

I’ve recently received my PTM7958 from Framework and things are looking good on average.

Using amdgpu_top ( 0.10.4 ), watching the temperatures and power/TDP, one thing I noticed ( 2~ weeks before PTM7958 ) was that the throttle status flags showing

Throttle Status: [PROCHOT_CPU, PROCHOT_GPU]

and I thought this was due to the processor package overheating because of the liquid thermal issue, but after applying the PTM7958, I’m still seeing this and thought this is weird because the temperatures are well within the safe range.

I rebooted and flags are gone ( as some others mentioned ), did some tests again ( PTM7958 related ) and thought “hmm, wonder what frequencies are like on battery”, unplugged Framework power adapter and the PROCHOT_CPU and PROCHOT_GPU flags showed again, I’m like “huh?”

I plugged the Framework power adapter back in and then another flag showed for 1 second ( EDC_CPU ) while PROCHOT_CPU and PROCHOT_GPU flags is still showing.

I went away from my FW16 and it auto suspended in that time while amdgpu_top was still running, later resumed and flags are gone

so I tried this numerous times in power profiles ( power save, balanced and performance ) via KDE power devil

  • Unplug the Framework power adapter, the PROCHOT_CPU and PROCHOT_GPU flags show
  • Suspend and then resume, flags are gone
  • Plug in the Framework power adapter and EDC_CPU flags shows for 1 second ( it’s probably less but I have amdgpu_top refresh period to 1 second ) while PROCHOT_CPU and PROCHOT_GPU still showing
  • Suspend and resume, flags are gone

The EDC_CPU flag though only shows when I plug ( never at unplug ) the Framework power adapter, CPU is idle and power profile set to “power save”

again “power save” in KDE power devil ( communicates to power profile daemon ), in my case also turns off CPB ( Core Performance Boost ) and sets iGPU performance level to “low”, in “balanced” or “performance” CPB is on and iGPU performance level is higher

There was one instance ( after PTM7958 applied but has happened numerous times before PTM7958 and again, thought it was liquid thermal issue ) where the flags were showing, suspended the system overnight, resumed later next morning and regardless of the power profiles the processor frequency would not go above 3.8GHz~ and the temperatures were ranging 70-75 degree’s celsius, after reboot ( suspend/resume didn’t fix this ), frequencies were good again and I’ve not been able to trigger this again

Since applying the PTM7958, I’ve never seen the processor temperature go above 97.8 degree’s celsius ( average is 90.8-97.8, depending on ambient temperature ) degree’s celsius, before with liquid thermal, it was hitting 100, sometimes 101 many times

so it seems changing the power state from AC to DC or DC to AC enables these flags and after suspend/resume or reboot flags are gone, I’m guessing something is not being configured correctly in the firmware/BIOS/EC before or after AC/DC/suspend/resume/reboot?

and with those flags enabled it’s possible the firmware/BIOS/EC elsewhere is doing something which may eventually lead to the FTR issue ( #41 )?

The FTR issue for me is low frequency ( weeks )

edit: maybe a bug with amdgpu_top, that only just occurred to me but there no open issues and I’ve been using this software on multiple AMD GPU’s for some years and when looking at the frequencies, I use btop and htop too and see similar numbers ( btop just shows and average of all the cores )

edit 2: it’s just done it again. On battery ( down to 59% ), plugged in, flags show, suspended, resumes, flags still showed and processor frequency won’t go above 3.8GHz, suspended and resumed again, flags gone and frequency still won’t go above 3.8GHz, this is while I’m in performance mode and temps no higher than 81 degree’s celsius ( weather quite warm today ). I’m now wondering if this frequency issue is to do with battery discharging, I’ll find out later and if so, maybe another issue ( BIOS or OS? )

===