`Power Saver` profiles should set energy performance preference to `balance_power` or `power` when using the `amd-pstate-epp` CPU scaling driver.

changed the description

It already does that if you're using the intel-pstate driver, check the output of powerprofilesctl to see which driver you're using.

I have an AMD system, and more recently AMD provided support for their own amd-pstate driver that iirc does the same thing.

Here is the CPU scaling driver I'm using for reference.

A quick google search got me this read on the change. It appears to have been changed in kernel 6.5 https://www.phoronix.com/news/Linux-6.5-Power-Management

Here is the output of powerprofilesctl. I misunderstood your comment and didn't realize you meant just using the command powerprofilesctl:

  performance:
    Driver:     platform_profile
    Degraded:   no

  balanced:
    Driver:     platform_profile

* power-saver:
    Driver:     platform_profile

Seems like this is a similar or duplicate issue of https://gitlab.freedesktop.org/hadess/power-profiles-daemon/-/issues/124

And there is a draft PR that attempts to fix it: https://gitlab.freedesktop.org/hadess/power-profiles-daemon/-/merge_requests/123

closed

reopened

Reopening as it seems like this is more of a bug rather than a lack of a feature.

In this file, which appears to be related to the amd-pstate driver: https://gitlab.freedesktop.org/hadess/power-profiles-daemon/-/blob/be37aea01b10fb27be3a032f81a5388755589583/src/ppd-driver-amd-pstate.c

we have this function which appears to aim to set the epp preference for amd-pstate. However, this does not trigger for my system.

static const char *
profile_to_epp_pref (PpdProfile profile)
{
  /* Note that we don't check "energy_performance_available_preferences"
   * as all the values are always available */
  switch (profile) {
  case PPD_PROFILE_POWER_SAVER:
    return "power";
  case PPD_PROFILE_BALANCED:
    return "balance_performance";
  case PPD_PROFILE_PERFORMANCE:
    return "performance";
  }

  g_assert_not_reached ();
}

Also found another resource related to amd-pstate and amd-pstate-epp, the latter of these being amd-pstate when it is set to active instead of passive or guided. The active mode is new as of kernel 6.3, and set to default in 6.5. https://www.reddit.com/r/linux/comments/15p4bfs/amd_pstate_and_amd_pstate_epp_scaling_driver/

CC @hadess

changed title from Power Saver profiles should set energy performance preference to balance_power or power to Power Saver profiles should set energy performance preference to balance_power or power{+ when using the amd-pstate-epp CPU scaling driver.+}

The issue is that if platform_profile is what PPD uses then the EPP is never set. And if the pstate driver is used, then platform_profile is never used, meaning we have limited control and no fan-profile setting (or other related things happening in acpi). This applies to AMD and Intel.

These two things need to be used together if platform_profile is available. I'm looking at how to do this in kernel but it might be out of my ability.

When you say this do you mean you're looking at the linux kernel and plan to make a patch for it?

Yes, but I do not see any sort of easy way to adjust that one thing from the asus_wmi driver.

It seems like this issue of multiple drivers applies to some lenovo devices as well, since the guy who was working on !123 (closed) was developing on a lenovo device. We'd probably want a solution to work for them as well