pm_policy(9)

💅 pm_policy(9) 💅

okay bestie, this one's from the Kernel API Manual, and it's literally all about letting your SOC pick a vibe for power management. let's get into it 🫶

---

✨ NAME ✨

pm_policy — the sysfs situation where you tell your GPU how to balance clock/throttling energy ⚡

---

🌸 DESCRIPTION 🌸

so, here's the tea: certain SOCs can support different power policies to optimize application performance. but — and this is the important part — this policy is provided only at SOC level, not at a per-process level. which honestly slaps when the entire SOC is utilized for a dedicated workload. main character SOC energy 👑

The amdgpu driver provides a sysfs API for selecting the policy. Presently, only two types of policies are supported through this interface. two. that's it. that's the roster 💅

• Pstate Policy Selection — this is to select different Pstate profiles which decides clock/throttling preferences. your whole speed-vs-chill situation ✨
• XGMI PLPD Policy Selection — when multiple devices are connected over XGMI, this helps to select policy to be applied for per link power down. link-by-link power down, iykyk 🔗

heads up babe: the list of available policies and policy levels vary between SOCs. they can be viewed under the pm_policy node directory. and if the SOC doesn't support any policy, this node won't be available at all — like, not even invited. the different policies supported will be available as separate nodes under pm_policy. 🎀

---

🦋 Reading the policy file

cat /sys/bus/pci/devices/.../pm_policy/<policy_type>

Reading the policy file shows the different levels supported. the level which is applied presently is denoted by * (asterisk). so look for the little star, that's your current fave ⭐ E.g.:

cat /sys/bus/pci/devices/.../pm_policy/soc_pstate
0 : soc_pstate_default
1 : soc_pstate_0
2 : soc_pstate_1*
3 : soc_pstate_2

cat /sys/bus/pci/devices/.../pm_policy/xgmi_plpd
0 : plpd_disallow
1 : plpd_default
2 : plpd_optimized*

see that 2 : soc_pstate_1*? that's the one currently serving looks 💅

---

🎀 To apply a specific policy

"echo  <level> > /sys/bus/pci/devices/.../pm_policy/<policy_type>"

for the levels listed in the example above, to select "plpd_optimized" for XGMI and "soc_pstate_2" for soc pstate policy —

echo "2" > /sys/bus/pci/devices/.../pm_policy/xgmi_plpd
echo "3" > /sys/bus/pci/devices/.../pm_policy/soc_pstate

and boom, policy applied, slay 🔥

---

🌈 SEE ALSO 🌈

okay this list is long but we don't skip, we don't summarize, we list every single one because that's what icons do 💜

• Kernel file ./drivers/gpu/drm/amd/pm/amdgpu_pm.c
• amdgpu_pm_dev_state_check(9), amdgpu_pm_get_access(9), amdgpu_pm_get_access_if_active(9), amdgpu_pm_put_access(9), power_dpm_state(9), power_dpm_force_performance_level(9), pp_table(9), pp_od_clk_voltage(9), pp_features(9), pp_dpm_sclk pp_dpm_mclk pp_dpm_socclk pp_dpm_fclk pp_dpm_dcefclk pp_dpm_pcie(9), pp_power_profile_mode(9), gpu_busy_percent(9), mem_busy_percent(9), vcn_busy_percent(9), pcie_bw(9), unique_id(9), thermal_throttling_logging(9), apu_thermal_cap(9), gpu_metrics(9), smartshift_apu_power(9), smartshift_dgpu_power(9), smartshift_bias(9), board(9), baseboard_temp(9), gpuboard_temp(9), cur_node_power_limit(9), node_power(9), npm_status(9), global_ppt_resid(9), max_node_power_limit(9), baseboard_power(9), baseboard_power_limit(9), hwmon(9), fan_curve(9), acoustic_limit_rpm_threshold(9), acoustic_target_rpm_threshold(9), fan_target_temperature(9), fan_minimum_pwm(9), fan_zero_rpm_enable(9), fan_zero_rpm_stop_temperature(9)

---

and that's the whole page, bestie 🧚 this one came straight out of ./drivers/gpu/drm/amd/pm, dated October 2026, and it's giving pm_policy(9) energy from start to finish. mwah 💖