Skip to content

FW16 (AMD 7040 / Ryzen 7840HS) + RX 7700S: dGPU power appears capped to ~35–40W under sustained load on Linux (firmware 0.0.4.2/0.0.4.3b) #151

@ChristophSchmidpeter

Description

@ChristophSchmidpeter

Device Information

System Model or SKU

Please select one of the following

  • Framework Laptop 12 (13th Gen Intel® Core™)
  • Framework Laptop 13 (11th Gen Intel® Core™)
  • Framework Laptop 13 (12th Gen Intel® Core™)
  • Framework Laptop 13 (13th Gen Intel® Core™)
  • Framework Laptop 13 (AMD Ryzen™ 7040 Series)
  • Framework Laptop 13 (AMD Ryzen™ AI 300 Series)
  • Framework Laptop 13 (Intel® Core™ Ultra Series 1)
  • Framework Laptop 16 (AMD Ryzen™ 7040 Series)
  • Framework Laptop 16 (AMD Ryzen™ AI 300 Series)
  • Framework Desktop (AMD Ryzen™ AI 300 PRO Series)

BIOS VERSION

Firmware: 0.0.3.2, 0.0.4.2, 0.0.4.3b (still reproducible).
04.03

DIY Edition information

Memory: 64GiB
Storage: 8TiB + 2TiB

Port/Peripheral information

Charger (peripheral): third-party USB-C 140W adapter.
Port used: [PLEASE FILL IN: which expansion bay USB-C / which numbered port per KB]
Adjacent port device(s): [PLEASE FILL IN if relevant]

  1. [ Fill In device / expansion card ]
  2. [ Fill In device / expansion card ]
  3. [ Fill In device / expansion card ]
  4. [ Fill In device / expansion card ]
  5. [ Fill In device / expansion card ]
  6. [ Fill In device / expansion card ]

Standalone Operation

Are you running your mainboard as a standalone device. Is standalone mode enabled in the BIOS?

  • Yes
  • No

Describe the bug

On Framework Laptop 16 (Ryzen 7840HS, RX 7700S module), the dGPU appears to be hard-limited to ~35–40W under sustained load on Linux. Even when the dGPU is clearly active (e.g. Steam/Proton game launched with DRI_PRIME=1; radeontop shows dGPU near 100% utilisation and iGPU relatively low), real-world performance is often nearly identical to running on the 780M iGPU. Power reported via sysfs (power1_average) does not exceed ~39W in my testing.

This severely impacts my primary use case for the RX 7700S module: local compute workloads on Linux (LLM inference / LoRA training, etc.). The cap can be "silent" and waste significant time because workloads just run much longer without an obvious error.

This is not obviously Linux-only: there are reports of a similar ~35W cap on Windows, and a separate issue about not reaching full dGPU power even with a 240W PD 3.1 charger.

Steps To Reproduce

  1. Boot Linux (Arch; KDE Plasma Wayland).
  2. Ensure on AC power (third-party USB-C 140W charger).
  3. Start a sustained dGPU workload:
    • Example: Steam/Proton game with launch options: DRI_PRIME=1 %command%
    • (Any sustained GPU stress test works as well.)
  4. Observe GPU utilisation (e.g. radeontop / amdgpu_top):
    • dGPU high utilisation, iGPU lower utilisation.
  5. Observe power and clocks:
    • cat /sys/class/drm/card*/device/hwmon/hwmon*/power1_average
    • cat /sys/class/drm/card*/device/pp_dpm_sclk
  6. Result: dGPU power remains around ~35–40W and performance is close to iGPU.

Evidence from my system (under load):

  • Power (uW):
    $ cat /sys/class/drm/card*/device/hwmon/hwmon*/power1_average
    29000000
    39255000
    (= ~29.0W and ~39.3W)

  • Clock tables:
    $ cat /sys/class/drm/card*/device/pp_dpm_sclk
    0: 255Mhz
    1: 794Mhz *
    2: 2208Mhz
    0: 800Mhz *
    1: 1100Mhz
    2: 2700Mhz

Expected behavior

On AC power, the RX 7700S should be able to exceed ~35–40W under sustained load (or, if there are intended caps based on charger/battery/power policy, they should be clearly documented and ideally configurable). The dGPU module should deliver a meaningful performance improvement over the 780M for GPU-bound workloads.

Screenshots

N/A (can provide MangoHud/amdgpu_top screenshots if requested).

Operating System (please complete the following information):

  • OS/Distribution: Arch Linux
  • Version: rolling (fully up-to-date at time of reporting)
  • Linux Kernel Version: [PLEASE FILL IN: output of uname -a]

Additional context

Questions:

  1. Is this considered the same underlying issue as DGPU isnt using full power limit on 240W charger while in performance mode #27 (EC/firmware/PD negotiation / power-policy behavior), or a different issue?
  2. Is a ~35–40W cap expected in any power/charger state? If yes, where is this documented and how can users control/raise caps on AC?
  3. What exact diagnostics do you want (dmesg, amdgpu logs, ectool chargestate output, specific charger tests, supported distro test)?

I can provide full logs and test with additional chargers if you tell me which configurations you consider valid.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions