Daemon to set scaling_max_freq according to CPU temperature

[logrusx]

OK, with k10temp it didn't work perhaps because I has switched amd-pstate in active mode after all. Zenpower is somehow broken. The user who brought it back to live last time has since vanished together with his repo and code. I think it should not be supported.

When I switched back amd-pstate to passive mode and loaded k10temp, it now works again.

Best Regards,
Georgi

[pa4wdh]

I'm sorry, i haven't seen your posts here. This topic popped up in a search i did and i more or less found them accidentally.

2025-03-24 21:16:20: Transition Latency: 0 (This seems low, consider setting it via -d)

This seems odd, try to set it with -d, start with a value of 50000.

2025-03-24 21:16:20: Error 22 occoured while writing to sysfs
2025-03-24 21:16:20: Failed to set governor, error 22 (Invalid argument)

My guess here is that the conservative governor is not available in your kernel. You can either add the conservative governor to your kernel or use -g to select an other one.

[logrusx]

I'm sorry, i haven't seen your posts here. This topic popped up in a search i did and i more or less found them accidentally.

2025-03-24 21:16:20: Transition Latency: 0 (This seems low, consider setting it via -d)

This seems odd, try to set it with -d, start with a value of 50000.

2025-03-24 21:16:20: Error 22 occoured while writing to sysfs
2025-03-24 21:16:20: Failed to set governor, error 22 (Invalid argument)

My guess here is that the conservative governor is not available in your kernel. You can either add the conservative governor to your kernel or use -g to select an other one.

In active mode there are only powersave and performance governors. I guess it's active because it actively predicts CPU load, rather than relying on the OS for that. The OS only gives hints.

Maybe you read my previous posts and missed the last one. I'm not sure if anything can be done in active mode. But in writing from memory because I'm not at my desk right now.

If you're interested, you can read about the modes in Documentation/admin/pm/amd-pstae.rst. If I'm getting the path wrong, it's similar.

Best Regards,
Georgi

[pa4wdh]

I've read the amd-pstate documentation. As far as i can see using any kind of software control doesn't make sense in active mode. The document states:

Code: Select all

- In autonomous mode, platform ignores the desired performance level request
  and takes into account only the values set to the minimum, maximum and energy
  performance preference registers.

And:

Code: Select all

Active Mode
------------

``amd_pstate=active``

This is the low-level firmware control mode which is implemented by ``amd_pstate
_epp``
driver with ``amd_pstate=active`` passed to the kernel in the command line.
In this mode, ``amd_pstate_epp`` driver provides a hint to the hardware if software
wants to bias toward performance (0x0) or energy efficiency (0xff) to the CPPC firmware.
then CPPC power algorithm will calculate the runtime workload and adjust the realtime
cores frequency according to the power supply and thermal, core voltage and some other
hardware conditions.

Which is probably why only powersafe and performance make sense in this mode, I wouldn't expect my daemon to do anything useful while in active mode.

[logrusx]

I wouldn't expect my daemon to do anything useful while in active mode.

I would add active mode is probably not a good idea for people who need what your daemon does.

Best Regards,
Georgi