tuned-amdgpu/README.md

tuned-amdgpu

Hacky solution to integrate AMDGPU power control and overclocking in tuned with Ansible

Takes a list of existing tuned profiles and creates new ones based on them. These new profiles include AMDGPU power/clock parameters

An attempt is made to discover the active GPU using the 'connected' state in the DRM subsystem, example:

~ $ grep -ls ^connected /sys/class/drm/*/status | grep -o card[0-9] | sort | uniq | sort -h | tail -1
card1

Warning: This is only tested with RX6000 series GPUs, it is probable that other generations will not work properly. Use at your own risk!

Profiles

Two 'profiles' are in each name:

• before amdgpu is the source profile provided with tuned
• after amdgpu tells the GPU clock profile offered, outlined below

Output profile	Description
balanced-amdgpu-default	Includes the (assumed) existing balanced tuned profile. Only adjusts the GPU power limit (typically lower). Clocks/voltage curve remain the default.
desktop-amdgpu-overclock	Includes the (assumed) existing desktop tuned profile. Adjusts the GPU power limit, clocks, and the voltage curve.
desktop-amdgpu-peak	Includes the (assumed) existing desktop tuned profile. Same as the overclock profile, but locks clocks to their highest configured values

Notable variables

These are the variables you're likely to want to change. They are defined in playbook.yml

Variable	Description	In-playbook
gpu_clock_min	Sets the minimum (dynamic) GPU clock (in Mhz) for the non-default amdgpu profiles	700
gpu_clock_max	Sets the maximum (dynamic) GPU clock (in MHz) for the non-default amdgpu profiles	2600, results in 2.6GHz (rounded); mild overclock
gpumem_clock_static	Sets the static memory clock for the GPU (in MHz). This is not the effective data rate. That is a multiple of this depending on the type of VRAM. To avoid flickering this does not change dynamically with load.	1050, results in just over 1GHz; mild overclock Actual effective clock depends on this being multiplied against the data/pump rate of the GDDR? GPU memory
gpu_mv_offset	GPU core voltage offset. Takes +/- some integer in millivolts. Can be used to both over and under volt.	-50 (undervolt 50mV or 0.05V)
base_profiles	List of base tuned profiles to clone in the new AMDGPU profiles. Defaults based on Fedora	balanced desktop latency-performance network-latency network-throughput powersave virtual-host
gpu_power_multi	Dictionary with two keys, default and overclock. Expects two floats to set a power limit relative to the board capability. Example: 1.0 is full board capability, 0.5 is 50%.