tuned-amdgpu/playbook.yml

---
- hosts: localhost
  become: yes
  vars:
    # the multipliers against power capability to determine power limits for the non-OC (default)/OC (custom) profiles
    # 0.5 = 50%
    # 1.0 = 100% (of card power capability, not stock limits)
    # should not exceed 1.0, must be a float.  driver will do some rounding/stepping
    # default is ~87% (281/323) on my 6900XT.  check '/sys/class/drm/card*/device/hwmon/hwmon*/power1_cap_default'
    power_max_custom_multi: 0.928793 # calculated to ~300W. used to control the effective power limit in the non-default AMDGPU tuned profiles
    power_max_default_multi: 0.82    # (typically) used to limit GPU power to some lower percentage on default perf mode/profile
    #
    # minimum/maximum GPU clocks using 'powerplay'
    # these do not apply to the resulting 'amdgpu-default' tuned profile
    # only the others (eg: 'VR'/'custom') under amdgpu_profiles (below)
    # effective clocks are dynamically adjusted by the driver in this range - based on utilization
    # can find baseline values in the '/sys/class/drm/{{ card }}/device/pp_od_clk_voltage' file
    #   OD_RANGE indicates acceptable value ranges for SCLK (core) and MCLK (memory)
    #   Under 'OD_SCLK' and 'OD_MCLK' you will see 0/1.  These are the minimum/maximum values for the respective component.
    gpu_clock_min: "2000"
    gpu_clock_max: "2615"
    #
    # memory clocks are static, we only set a 'max' - high-refresh rate display quirk
    # this allows the memory clock to be adjusted
    gpumem_clock_max: "1075"
    #
    # optional, applies offset to GPU voltage, eg: '+100' = to boost GPU core voltage 100mV or 0.1V. for the 'custom' GPU profile.
    gpu_mv_offset: "-25"
    #
    # the card for 'tuned' to control
    # this is *usually* 'card0', but may differ.
    # check '/dev/dri/card*'
    card: 'card0'  # default to card0
    # list of source tuned profiles available on Fedora (TODO: should dynamically discover)
    # further modified with AMD GPU power/clock parameters, creating new profiles. eg: 'balanced-amdgpu-VR'
    base_profiles:
      - balanced
      - desktop
      - latency-performance
      - network-latency
      - network-throughput
      - powersave
      - virtual-host
    amdgpu_profiles: # statically defined mapping of the contents in /sys/class/drm/card*/device/pp_power_profile_mode
      default:       # more may be added, but do not remove default/custom. new profiles require a script template, see 'templates'
        pwrmode: 0
      VR:
        pwrmode: 4
      custom:
        pwrmode: 6
  handlers:
    - name: restart tuned
      ansible.builtin.service:
        name: tuned
        state: restarted
  tasks:
    - name: Gather package facts
      ansible.builtin.package_facts:
        manager: auto
    - name: replace 'power-profiles-daemon' with 'tuned' on Fedora 35+
      dnf: # use with_items since 'dnf' module in Ansible doesn't support 'swap'
        name: "{{ item.name }}"
        state: "{{ item.state }}"
      with_items:
        - {name: 'power-profiles-daemon', state: 'absent'}
        - {name: 'tuned', state: 'present'}
      when: ('power-profiles-daemon' in ansible_facts.packages) or (('tuned' not in ansible_facts.packages) and ((ansible_distribution == 'Fedora') and (ansible_distribution_major_version|int > 35)))
      register: fed_ppdtuned_swap
      # 'power-profiles-daemon' was added/conflicts with 'tuned' since F35
      # otherwise, ensure the 'tuned' package is installed
    - name: install tuned
      package:
        name: tuned
        state: present
      when: (fed_ppdtuned_swap is not defined) or ('tuned' not in ansible_facts.packages)
    - name: find hwmon/max power capability file for {{ card }}
      find:
        paths: /sys/class/drm/{{ card }}/device/hwmon
        file_type: file
        recurse: yes
        use_regex: yes
        patterns:
          - '^power1_cap_max$'
      register: hwmon
    - name: find hwmon/current power limit file for {{ card }}
      find:
        paths: /sys/class/drm/{{ card }}/device/hwmon
        file_type: file
        recurse: yes
        use_regex: yes
        patterns:
          - '^power1_cap$'
      register: powercap_set
    - name: get max power capability for {{ card }}
      slurp:
        src: "{{ hwmon.files.0.path }}"
      register: power_max_b64
    - name: create custom profile directories
      file:
        state: directory
        path: /etc/tuned/{{ item.1 }}-amdgpu-{{ item.0.key }}
        mode: "0755"
      with_nested:
        - "{{ lookup('dict', amdgpu_profiles) }}"
        - "{{ base_profiles }}"
    - name: template AMDGPU control/reset scripts
      template:
        src: templates/amdgpu-clock.sh.j2
        dest: /etc/tuned/{{ item.1 }}-amdgpu-{{ item.0.key }}/amdgpu-clock.sh
        owner: root
        group: root
        mode: "0755"
      with_nested:
        - "{{ lookup('dict', amdgpu_profiles) }}"
        - "{{ base_profiles }}"
      notify: restart tuned
    - name: template custom tuned profiles
      template:
        src: templates/tuned.conf.j2
        dest: /etc/tuned/{{ item.1 }}-amdgpu-{{ item.0.key }}/tuned.conf
        owner: root
        group: root
        mode: "0644"
      with_nested:
        - "{{ lookup('dict', amdgpu_profiles) }}"
        - "{{ base_profiles }}"
      notify: restart tuned
    - name: ensure tuned is enabled
      service:
        name: tuned
        enabled: yes