Please fix the ACPI/SBIOS tables so that the amd_pstate governor can be used

[Nyan~]

As is, the amd_pstate governor can not be used on the Steam Deck despite cpuid advertising support for cppc due to the lack of a _CPC object. Furthermore, enabling the CPC option on BIOS 116 with the unlock tool results in an unbootable deck without the CMOS reset key combo, so evidently it's just entirely absent or broken.

Please fix this so that I can save some additional power. Being stuck on one of three performance states that can only scale slightly up or down using boost frequencies is not ideal.

[deaddoof]

i knew that it was broken but i didn't know why.
the missing pstate surprised takashi iwai

[Rodomar705]

Yep, it's broken and the driver do not load since the ACPI tables are missing. Please fix this glaring issue, so we can save additional power by being able to switch to more than the couple of frequency states that the standard ACPI allows.

[deaddoof]

Op. Do you know whether these issues affect software sleep?

echo freeze | tee -a /sys/power/state

[Nyan~]

Publicado originalmente por deaddoof:

Op. Do you know whether these issues affect software sleep?

echo freeze | tee -a /sys/power/state

I don't see why you would do that when suspend-to-RAM/deep sleep works, has the expected power consumption, wakes up extremely quickly and is the default sleep method. You should be changing mem_sleep to deep and writing mem.

I would imagine the power consumption of suspend-to-idle is probably slightly less than the device idling, which would be lower if the firmware actually supported the amd_pstate governor. It'd probably still have higher power consumption than suspend-to-RAM, though, so I don't see the point in testing this.

[deaddoof]

Publicado originalmente por Nyan~:

Publicado originalmente por deaddoof:

Op. Do you know whether these issues affect software sleep?

echo freeze | tee -a /sys/power/state

I don't see why you would do that when suspend-to-RAM/deep sleep works, has the expected power consumption, wakes up extremely quickly and is the default sleep method. You should be changing mem_sleep to deep and writing mem.

I would imagine the power consumption of suspend-to-idle is probably slightly less than the device idling, which would be lower if the firmware actually supported the amd_pstate governor. It'd probably still have higher power consumption than suspend-to-RAM, though, so I don't see the point in testing this.

I find it kinda awkward software sleep is broken in upstream kernel but hardware sleep somehow works. The deck have many power issues. Hard to reproduce.

Steam OS software sleep is hit and miss.

[Nyan~]

Suspend-to-idle isn't even proper sleep. It's turning off everything possible and idling.

Nothing is broken with regards to sleep, nor does sleep have anything to do with what I brought up in the OP.