Stopped displaying the watt consumption sensor for the GPU and CPU.

[arspacer]

Describe the bug Stopped displaying GPU watt consumption after updating to the current version. Also does not show CPU watt consumption; it has never shown it for some reason, regardless of the MangoHud version.

List relevant hardware/software information

• Void Linux x86_64 Kernel: Linux 6.12.49_1
• MangoHud version v0.8.1
• GPU: AMD Radeon RX 580 Series 8GB

To Reproduce Steps to reproduce the behavior:

1. Install sudo xbps-install -S MangoHud
2. Register in the Steam launch option: mangohud %command%
3. Used different presets (via Shift_R+F10) and tried my own configurations to display the wattage consumption for the CPU and GPU.
4. As a result, the consumption meters for the CPU and GPU are not displayed.

Expected behavior I installed it, launched it, everything should work, but I couldn't get it to run correctly so that everything would work :(

Screenshots

[flightlessmango]

cpu power requires intervention documented in readme. Do you have the ability to test latest git mangohud to see if gpu power has already been fixed?

[DanAlex121]

@flightlessmango Same issus here. Mangohud can't display GPU power but nvtop/amdgpu_top can display it normally.

List relevant hardware/software information

• Linux Distribution: Nobara Linux 43
• Kernel: 6.17.11-200.nobara.fc43.x86_64
• MangoHud version: 0.8.2-4.20251205git1115898.fc43
• GPU: AMD RX 590

I ran mangohud under debug mode and log shows: [MANGOHUD] [debug] [amdgpu.cpp:474] Failed to open gpu_metrics at '/sys/bus/pci/devices/0000:01:00.0/gpu_metrics' I think the problem is Old AMD cards don't have gpu_metrics info but use older hwmon info for status.

And for temperature Mangohud did fallback to use hwmon info. [debug] [cpu.cpp:628] hwmon: using input: /sys/class/hwmon/hwmon2/temp1_input Selected GPU 0: AMD Radeon RX 590 Series (RADV POLARIS10), type: DiscreteGpu

So I think why not fallback to /hwmon/hwmon1/power1_input and /hwmon/hwmon1/power1_average to display GPU power if gpu_metrics is unavailable? And for my case, I only have power1_input which path is /sys/class/drm/card1/device/hwmon/hwmon1/power1_input. There are also someone only have power1_average so maybe both fallback is needed for proper display.

[DanAlex121]

And for anyone need a instant fix, a temporary (and ugly) AI generated solution could be

# Disable the default broken power reading
# gpu_power

# Create a custom entry
exec=bash -c "read v < /sys/class/drm/card1/device/hwmon/hwmon1/power1_input; echo \$((v/1000000)).\$((v%1000000/100000)) W"

or with awk

exec=cat /sys/class/drm/card1/device/hwmon/hwmon1/power1_input | awk '{printf "%.1f W\n", $1/1000000}'

preview: