[SteamOS 3.8 Beta] AMD EPP resets to "performance" on CPU1-7 after waking up from sleep

[khmm12]

Your system information

• Steam client version: 1776387948
• SteamOS version: 3.8.2 Beta
• Kernel version: 6.16.12-valve14-1-neptune-616-g4f70fa45dc69
• Opted into Steam client beta?: Yes
• Opted into SteamOS beta?: Yes
• Have you checked for updates in Settings > System?: Yes

Please describe your issue in as much detail as possible:

Expected Behavior:

After waking up from sleep (suspend), all CPU cores should preserve their energy_performance_preference (EPP) setting. For the Steam Deck using the amd-pstate-epp driver in "active" mode, the default should be balance_performance across all cores.

Actual Behavior:

Upon resume, only CPU0 retains the balance_performance setting. All other cores (CPU1-CPU7) are automatically reset to performance. This leads to higher power consumption and heat than expected after a sleep cycle.

I have verified this behavior using a custom debug script in /usr/lib/systemd/system-sleep/. The logs clearly show the state transition:

• Pre-suspend: All 8 cores are at balance_performance.
• Post-resume: CPU0 is at balance_performance, but CPU1 through CPU7 are at performance.

This seems to be a regression or a known issue in recent Linux kernels (6.16+).

Steps for reproducing this issue:

1. Boot Steam Deck (SteamOS 3.8 Beta).
2. Check EPP status: cat /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference (All should be balance_performance).
3. Put the device to sleep (Suspend).
4. Wake the device.
5. Check EPP status again. You will observe that only cpu0 is correct, while cpu1-7 have switched to performance.

debug script

sudo tee /usr/lib/systemd/system-sleep/00-log-epp >/dev/null <<'EOF'
#!/bin/sh
LOG=/tmp/epp-resume.log

dump_state() {
  echo "=== $1 @ $(date '+%F %T.%N %Z') ===" >> "$LOG"
  echo "--- scaling_driver ---" >> "$LOG"
  cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver 2>/dev/null >> "$LOG"
  echo "--- amd_pstate status ---" >> "$LOG"
  cat /sys/devices/system/cpu/amd_pstate/status 2>/dev/null >> "$LOG"
  echo "--- governors ---" >> "$LOG"
  for g in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor; do
    printf "%s = %s\n" "$g" "$(cat "$g" 2>/dev/null)" >> "$LOG"
  done
  echo "--- epp ---" >> "$LOG"
  for f in /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference; do
    printf "%s = %s\n" "$f" "$(cat "$f" 2>/dev/null)" >> "$LOG"
  done
  echo >> "$LOG"
}

case "$1" in
  pre)  dump_state "PRE-SUSPEND" ;;
  post) dump_state "POST-RESUME" ;;
esac
EOF

sudo chmod +x /usr/lib/systemd/system-sleep/00-log-epp

debug logs

cat /tmp/epp-resume.log

=== PRE-SUSPEND @ 2026-04-19 16:10:49.650911703 ===
--- scaling_driver ---
amd-pstate-epp
--- amd_pstate status ---
active
--- governors ---
/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor = powersave
/sys/devices/system/cpu/cpu1/cpufreq/scaling_governor = powersave
/sys/devices/system/cpu/cpu2/cpufreq/scaling_governor = powersave
/sys/devices/system/cpu/cpu3/cpufreq/scaling_governor = powersave
/sys/devices/system/cpu/cpu4/cpufreq/scaling_governor = powersave
/sys/devices/system/cpu/cpu5/cpufreq/scaling_governor = powersave
/sys/devices/system/cpu/cpu6/cpufreq/scaling_governor = powersave
/sys/devices/system/cpu/cpu7/cpufreq/scaling_governor = powersave
--- epp ---
/sys/devices/system/cpu/cpu0/cpufreq/energy_performance_preference = balance_performance
/sys/devices/system/cpu/cpu1/cpufreq/energy_performance_preference = balance_performance
/sys/devices/system/cpu/cpu2/cpufreq/energy_performance_preference = balance_performance
/sys/devices/system/cpu/cpu3/cpufreq/energy_performance_preference = balance_performance
/sys/devices/system/cpu/cpu4/cpufreq/energy_performance_preference = balance_performance
/sys/devices/system/cpu/cpu5/cpufreq/energy_performance_preference = balance_performance
/sys/devices/system/cpu/cpu6/cpufreq/energy_performance_preference = balance_performance
/sys/devices/system/cpu/cpu7/cpufreq/energy_performance_preference = balance_performance

=== POST-RESUME @ 2026-04-19 16:10:56.639612150 ===
--- scaling_driver ---
amd-pstate-epp
--- amd_pstate status ---
active
--- governors ---
/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor = powersave
/sys/devices/system/cpu/cpu1/cpufreq/scaling_governor = powersave
/sys/devices/system/cpu/cpu2/cpufreq/scaling_governor = powersave
/sys/devices/system/cpu/cpu3/cpufreq/scaling_governor = powersave
/sys/devices/system/cpu/cpu4/cpufreq/scaling_governor = powersave
/sys/devices/system/cpu/cpu5/cpufreq/scaling_governor = powersave
/sys/devices/system/cpu/cpu6/cpufreq/scaling_governor = powersave
/sys/devices/system/cpu/cpu7/cpufreq/scaling_governor = powersave
--- epp ---
/sys/devices/system/cpu/cpu0/cpufreq/energy_performance_preference = balance_performance
/sys/devices/system/cpu/cpu1/cpufreq/energy_performance_preference = performance
/sys/devices/system/cpu/cpu2/cpufreq/energy_performance_preference = performance
/sys/devices/system/cpu/cpu3/cpufreq/energy_performance_preference = performance
/sys/devices/system/cpu/cpu4/cpufreq/energy_performance_preference = performance
/sys/devices/system/cpu/cpu5/cpufreq/energy_performance_preference = performance
/sys/devices/system/cpu/cpu6/cpufreq/energy_performance_preference = performance
/sys/devices/system/cpu/cpu7/cpufreq/energy_performance_preference = performance

Additional context

This issue seems to be a known regression in the Linux kernel affecting AMD EPP persistence. Similar behavior has been reported by other users here: Reddit.

It appears that this issue is addressed in the upstream kernel by the following commit: cpufreq/amd-pstate: Fix EPP being reset on resume.

[matte-schwartz]

I can reproduce the issue on SteamOS 3.8.2 with my Steam Deck. However, the commit you're referencing already seems to be in the kernel on this version of SteamOS:

~/src/kernel-src/linux tags/6.16.12-valve15
❯ git log --oneline | grep "cpufreq/amd-pstate: Fix a regression leading to EPP 0 after resume"
9c48608e3243 cpufreq/amd-pstate: Fix a regression leading to EPP 0 after resume

the issue does not persist on our development kernel which is based on Linux 6.18, so I'll see if I can figure out which commit(s) we're still missing.

[matte-schwartz]

looks like we need this very similarly-named commit: torvalds/linux@85d7dda

[matte-schwartz]

the fix for this is in the next kernel release for the beta/preview channel

[khmm12]

@matte-schwartz

I confirm that the issue has been fixed.

Thank you 🙇

logs

(1)(deck@sdeck-maxim ~)$ cat /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference
balance_performance
balance_performance
balance_performance
balance_performance
balance_performance
balance_performance
balance_performance
balance_performance
(deck@sdeck-maxim ~)$

Broadcast message from deck@sdeck-maxim (Fri 2026-05-01 13:31:57):

The system will suspend now!

(deck@sdeck-maxim ~)$ cat /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference
balance_performance
balance_performance
balance_performance
balance_performance
balance_performance
balance_performance
balance_performance
balance_performance