Adding the AMD PSTATE kernel module

General Discussion
1

I would like to test Clear Linux with the AMD PSTATE scaling governor.

(amd-pstate CPU Performance Scaling Driver — The Linux Kernel documentation)

(Implement AMD Pstate EPP Driver [LWN.net])

Current install:
Clear Linux Build 39050
kernel: 6.3.1-1311.native

Currently this build defaults to acpi-cpufreq performance (Boost:Enabled)

When I run “cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors”
The output is as follows:
‘ondemand’
‘performance’
‘schedutil’

My understanding was that the latest AMD PSTATE_EPP module was integrated at the kernel level since the 6.3 kernel release.

I am testing Clear Linux on a AMD Ryzen 9 6900HX. (Zen 3)

So based on this, shall I assume that the specific scaling driver will have to be added manually?

2

Actually we do enable the AMD P State drive in our kernel:
CONFIG_X86_AMD_PSTATE=y

but note that AMD PSTATE is not a governor (which you list above) but a backend driver.
The governor is what does the policy choices, the backend driver is what programs the hardware

3

Excellent feedback. Let me take a look at this and see if the changes to activate the backend (and the subsequent policy) work as expected. Thank you for the response.

4

Ok I have it working now.

Procedure for those interested is as follows:

You have to add a kernel level boot parameter in Clear.

  • Create a directory under /etc/kernel called cmdline.d
  • Create a *.conf file under that directory that will contain your kernel params
  • In my case I called it “amdpstate.conf”
    then created the conf file ‘nano /etc/kernel/cmdline.d/amdpstate.conf’
  • I placed 2 lines in the conf file
    “initcall_blacklist=acpi_cpufreq_init”
  • This tells the kernel not to load the default scaler
    “amd_pstate=passive”
  • Tells the kernel to load the pstate driver in passive mode
  • Save your output
  • Then run “clr-boot-manager update”
    This updates the boot manager to look for the extra kernel parameters
  • Reboot your OS
  • Login and run
    “cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver”
  • The response should be “amd-pstate”

I can start to test and check with the utilities available to monitor the changes (if any).

1 Like
5

Why “amd_pstate=passive” in my kernel param?

AMD Pstate Driver Operation Modes

amd_pstate CPPC has two operation modes: CPPC Autonomous(active) mode and CPPC non-autonomous(passive) mode. active mode and passive mode can be chosen by different kernel parameters. When in Autonomous mode, CPPC ignores requests done in the Desired Performance Target register and takes into account only the values set to the Minimum requested performance, Maximum requested performance, and Energy Performance Preference registers. When Autonomous is disabled, it only considers the Desired Performance Target.

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.

Passive Mode

amd_pstate=passive

It will be enabled if the amd_pstate=passive is passed to the kernel in the command line. In this mode, amd_pstate driver software specifies a desired QoS target in the CPPC performance scale as a relative number. This can be expressed as percentage of nominal performance (infrastructure max). Below the nominal sustained performance level, desired performance expresses the average performance level of the processor subject to the Performance Reduction Tolerance register. Above the nominal performance level, processor must provide at least nominal performance requested and go higher if current operating conditions allow.

6

I ran my usual 4 hour stress test after the pstate change and the results were identical. A couple of tests varied by 2-3%. I didnt expect any difference since the governor for both was set to performance.

I did put in a set of DDR5-5600 and re-ran the test and it fell down after about 2 hours. But this build of Clear was 100% stable at DDR5-4800.

I will change the governor and retry at DDR5-5600 and see what/if happens.

This Zen 3 does not expose all of the pstate’s available in CPPC, i just want to see if I can find why I get a green screen of death.

1 Like