RX 6000 series: Out-of-sync MCLK hardware and control values can cause system to reboot
Hello there,
Brief summary of the problem:
Due to the driver's inconsistent hardware and control values (97MHz or 674MHz resp.) for MCLK
, the system immediately reboots when using the advanced power management mode in CoreCtrl.
Hardware description:
- CPU: Ryzen 5 3600
- GPU: Radeon RX 6600 XT
- System Memory: 32GB
- Display(s): Eizo FlexScan EV2750 (27", WQHD)
- Type of Diplay Connection: DP
Output of cat /sys/class/drm/card0/device/pp_od_clk_voltage
:
OD_SCLK:
0: 700Mhz
1: 2619Mhz
OD_MCLK:
0: 97Mhz
1: 1000MHz
OD_VDDGFX_OFFSET:
0mV
OD_RANGE:
SCLK: 500Mhz 3150Mhz
MCLK: 674Mhz 1200Mhz
System information:
- Distro name and Version: Manjaro (21.1.2)
- Kernel version: 5.13.13-1-MANJARO
- AMD package version: amdgpu 21.0.0-1
How to reproduce the issue:
- Install CoreCtrl following the steps outlined on the Installation page
- Run corectrl
- System immediately reboots
The state 0 of
OD_MCLK
is changed even when not touched in the profile because all values used in controls must be in the operation range (excluding post-init restoration commands). As the hardware value (97MHz, bogus, out of range) is out of sync with the control value (674MHz is the minimum frequency operation value) the program tries to synchronize it. When the value is changed the system hangs.
In this thread, several users have reported similar issues.