GPU fan's speed does not go down after setting performance level from "profile_peak" to "auto"
Brief summary of the problem:
After setting the performance level from "profile_peak" to "auto", GPU's fan speed does not go down.
Hardware description:
- GPU: Ellesmere [Radeon RX 470/480/570/570X/580/580X/590]
System information:
- Distro name and Version: Arch Linux
- Kernel version: 6.3.6-arch1-1 #1 (closed) SMP PREEMPT_DYNAMIC Mon, 05 Jun 2023 15:12:57 +0000 x86_64 GNU/Linux
- Driver version: amdvlk 2023.Q2.2-1, mesa/vulkan-radeon 23.1.1-1
How to reproduce the issue:
echo "profile_peak" | sudo tee /sys/class/drm/card0/device/power_dpm_force_performance_level
Followed by:
echo "auto" | sudo tee /sys/class/drm/card0/device/power_dpm_force_performance_level
The issue happens with both drivers. In same very rare occasion the fan will slow down, but after a couple of switches between the two levels, the fan will get stuck at maximum speed.
Here's the output of sensors
:
RADV
At system startup
vddgfx: 750.00 mV
fan1: 795 RPM (min = 0 RPM, max = 3200 RPM)
edge: +47.0°C (crit = +91.0°C, hyst = -273.1°C)
PPT: 31.12 W (cap = 185.00 W)
After setting profile_peak
vddgfx: 1.19 V
fan1: 3173 RPM (min = 0 RPM, max = 3200 RPM)
edge: +53.0°C (crit = +91.0°C, hyst = -273.1°C)
PPT: 106.09 W (cap = 185.00 W)
After setting auto
vddgfx: 818.00 mV
fan1: 3161 RPM (min = 0 RPM, max = 3200 RPM)
edge: +41.0°C (crit = +91.0°C, hyst = -273.1°C)
PPT: 32.18 W (cap = 185.00 W)
AMDVLK
At system startup
vddgfx: 725.00 mV
fan1: 1669 RPM (min = 0 RPM, max = 3200 RPM)
edge: +41.0°C (crit = +91.0°C, hyst = -273.1°C)
PPT: 34.09 W (cap = 185.00 W)
After setting profile_peak
vddgfx: 1.19 V
fan1: 3173 RPM (min = 0 RPM, max = 3200 RPM)
edge: +50.0°C (crit = +91.0°C, hyst = -273.1°C)
PPT: 106.18 W (cap = 185.00 W)
After setting auto
vddgfx: 818.00 mV
fan1: 3163 RPM (min = 0 RPM, max = 3200 RPM)
edge: +38.0°C (crit = +91.0°C, hyst = -273.1°C)
PPT: 39.15 W (cap = 185.00 W)
Here's the output of dmesg -T | grep "amdgpu"
in case it could useful:
[Sat Jun 10 10:09:17 2023] [drm] amdgpu kernel modesetting enabled.
[Sat Jun 10 10:09:17 2023] amdgpu: Ignoring ACPI CRAT on non-APU system
[Sat Jun 10 10:09:17 2023] amdgpu: Virtual CRAT table created for CPU
[Sat Jun 10 10:09:17 2023] amdgpu: Topology: Add CPU node
[Sat Jun 10 10:09:17 2023] amdgpu 0000:09:00.0: No more image in the PCI ROM
[Sat Jun 10 10:09:17 2023] amdgpu 0000:09:00.0: amdgpu: Fetched VBIOS from ROM BAR
[Sat Jun 10 10:09:17 2023] amdgpu: ATOM BIOS: 113-4E3661U-X6I
[Sat Jun 10 10:09:17 2023] amdgpu 0000:09:00.0: vgaarb: deactivate vga console
[Sat Jun 10 10:09:17 2023] amdgpu 0000:09:00.0: amdgpu: Trusted Memory Zone (TMZ) feature not supported
[Sat Jun 10 10:09:17 2023] amdgpu 0000:09:00.0: amdgpu: VRAM: 8192M 0x000000F400000000 - 0x000000F5FFFFFFFF (8192M used)
[Sat Jun 10 10:09:17 2023] amdgpu 0000:09:00.0: amdgpu: GART: 256M 0x000000FF00000000 - 0x000000FF0FFFFFFF
[Sat Jun 10 10:09:17 2023] [drm] amdgpu: 8192M of VRAM memory ready
[Sat Jun 10 10:09:17 2023] [drm] amdgpu: 7963M of GTT memory ready.
[Sat Jun 10 10:09:17 2023] amdgpu: hwmgr_sw_init smu backed is polaris10_smu
[Sat Jun 10 10:09:17 2023] kfd kfd: amdgpu: Allocated 3969056 bytes on gart
[Sat Jun 10 10:09:17 2023] amdgpu: sdma_bitmap: f
[Sat Jun 10 10:09:17 2023] amdgpu: SRAT table not found
[Sat Jun 10 10:09:17 2023] amdgpu: Virtual CRAT table created for GPU
[Sat Jun 10 10:09:17 2023] amdgpu: Topology: Add dGPU node [0x67df:0x1002]
[Sat Jun 10 10:09:17 2023] kfd kfd: amdgpu: added device 1002:67df
[Sat Jun 10 10:09:17 2023] amdgpu 0000:09:00.0: amdgpu: SE 4, SH per SE 1, CU per SH 9, active_cu_number 36
[Sat Jun 10 10:09:17 2023] amdgpu 0000:09:00.0: amdgpu: Using BACO for runtime pm
[Sat Jun 10 10:09:17 2023] [drm] Initialized amdgpu 3.52.0 20150101 for 0000:09:00.0 on minor 0
[Sat Jun 10 10:09:17 2023] fbcon: amdgpudrmfb (fb0) is primary device
[Sat Jun 10 10:09:17 2023] amdgpu 0000:09:00.0: [drm] fb0: amdgpudrmfb frame buffer device
[Sat Jun 10 10:09:19 2023] snd_hda_intel 0000:09:00.1: bound 0000:09:00.0 (ops amdgpu_dm_audio_component_bind_ops [amdgpu])
[Sat Jun 10 10:15:10 2023] amdgpu 0000:09:00.0: amdgpu: Disabling VM faults because of PRT request!