You only need amd_pstate=guided in your kernel cmdline and #reinstall-kernels if you are using systemd-boot, or a #grub-mkconfig -o /boot/grub/grub.cfg if you are running grub.
That and amd_pstate.shared_mem=1 are no longer necessary.
Then, check if it is running by:
$ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
amd-pstate
$ cat /sys/devices/system/cpu/amd_pstate/status
guided
cpupower frequency-info can also give you more details.
Thanks! works perfectly!
Gave āguidedā mode a quick try. While it scales nicely according to workload, balancing power usage and performance, it still uses slightly more power in governors other than powersave (which is not usable at all, stuck at 400 MHz throughout) compared to Active EPP with powersave governor. At least for my use case. Reverted back to said EPP mode, together with the convenience of switching between power modes with a slider, which is absent on āguidedā.
Another observation: Idle power draw is identical with all governors in Guided or Active mode. (Idle being on desktop, no active network transfers, no apps opened)
on Desktop AMD Ryzen 9 5900X i canāt remove amd_pstate.shared_mem=1
where it seems to work setting: amd_pstate=guided
also am not sure if it is used as i still can set governors using cpupower-- may i need to disble cpupower daemon?
For my use case and setup running amd p-state with guided has so far provided the best results - meaning sufficiently low power consumption in idle and low load and still being responsive in high load scenarios.
My usual use case are sporadic high / extremely high loads when running big data or other scientific loads or using VMās for those pesky few proprietary softwares (Damn you Nvivo!). Rest is mostly idle with loads of web browsing and word processing, with few games in between.
Setup is AMD 5900x and x570 board.
Idle package watts (read with turbostat) are around 24āis W, loads go up to 115W which concur with TDP and boosting.
Running following kernel arguments
initcall_blacklist=acpi_cpufreq_init amd_pstate=guided amd_pstate.enable=1
Governor etc info are
driver: amd-pstate
maximum transition latency: 20.0 us
hardware limits: 550 MHz - 4.95 GHz
available cpufreq governors: conservative ondemand userspace powersave performance schedutil
current policy: frequency should be within 550 MHz and 4.95 GHz.
The governor āschedutilā may decide which speed to use within this range.
current CPU frequency: 4.30 GHz (asserted by call to kernel)
boost state support:
Supported: yes
Active: yes
AMD PSTATE Highest Performance: 166. Maximum Frequency: 4.95 GHz.
AMD PSTATE Nominal Performance: 124. Nominal Frequency: 3.70 GHz.
Not running cpupower daemon.
Passive mode for me is a no go, I can feel some latency, high CPU spikes.
I didnāt feel any difference between guided and active modes. (B550-F | 5800x3D).
Both feel pretty good.
All it is required for me is to set amd_pstate= with the mode I want in my kernel cmdline.
cpupower has been updated recently and now it is compatible with amd_pstate, although Iām using it just to check, service is not currently enabled.
analyzing CPU 2:
driver: amd-pstate
CPUs which run at the same hardware frequency: 2
CPUs which need to have their frequency coordinated by software: 2
maximum transition latency: 20.0 us
hardware limits: 550 MHz - 4.55 GHz
available cpufreq governors: conservative ondemand userspace powersave performance schedutil
current policy: frequency should be within 550 MHz and 4.55 GHz.
The governor "schedutil" may decide which speed to use
within this range.
current CPU frequency: Unable to call hardware
current CPU frequency: 3.60 GHz (asserted by call to kernel)
boost state support:
Supported: yes
Active: yes
AMD PSTATE Highest Performance: 166. Maximum Frequency: 4.55 GHz.
AMD PSTATE Nominal Performance: 124. Nominal Frequency: 3.40 GHz.
AMD PSTATE Lowest Non-linear Performance: 64. Lowest Non-linear Frequency: 1.75 GHz.
AMD PSTATE Lowest Performance: 21. Lowest Frequency: 550 MHz.
Iām using the active mode.
amd-pstate-epp
Since kernel 6.4, active mode is the default, you donāt need any kernel cmdline option for that, it will be enabled by default now.
Okay Iāll try it now.
Edit: Nope!
[ricklinux@eos-plasma ~]$ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
acpi-cpufreq
[ricklinux@eos-plasma ~]$
Edit: If i set guided it shows amd-pstate but it makes fans run faster even at idle. I like active better.
Hmm, weird.
If I remove the amd_pstate option from my kernel cmdline, I get amd-pstate-epp active mode by default.
According to this phonorix article, it should be enabled by default from 6.3 onward.
I canāt hear my fans, too much noise here because I live very near to a very busy road..
Perhaps this is happening to me and I didnāt notice it yet, Iāll try to confirm checking the RPMs in both modes to compare.
If i take the kernel parameter out it goes back to:
acpi-cpufreq
Based on what Iāve been reading about this, you should be able to remove that kernel entry and have amd-pstate-epp active mode by default.
Iāll remove it here to test, one sec .
Edit: Same thing happened to me, Iām kind of lost right now 
This behavior is new to Kernel 6.4 because as far as I remember, Kernel 6.3 was setting the default to amd-pstate-epp active mode.
$ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
acpi-cpufreq
Well Iām going to stick with this mode as it seems to be the best for me. I have my 100w Cpu set to ECO mode for 65w also. I donāt change anything else. I have never gotten into undervolting or any of that overclocking stuff in all my years of computers. I tend to stick with the manufacturers settings and change only minor things. Iām more concerned with firmware updates and making sure i have the latest UEFI Firmware (Bios). I feel that is the most important until at such time as the hardware is getting that old and updates are not supported anymore.
Iām ādownvoltingā my CPU, all cores to -30mV, I have more performance according to CB23, I can sustain the boost for longer, less CPU temperatures and less overall power usage, thus more efficient.
Iām now using amd-pstate-epp active mode to confirm if my fans RPM will change.
My governors changed from schedutil to powersave.
amd-pstate use the default governors.
amd-pstate-epp has it owns governors, such as powersave and performance.
So, if you use amd-pstate-epp, cpupower should be up to date in case you enable it because the governors are different, schedutil is not even an option anymore.
active is available since 6.3, it will become the default in 6.5
Thanks BS86, you clarified things nowā¦ 
What do you mean? CPU power should be up to date?
$ LANG=C sudo pacman -Ss cpupower
extra/cpupower 6.3-2 (linux-tools) [installed]
Linux kernel tool to examine and tune power saving related features of your processor
Previous versions of cpupower governors didnāt include amd-pstate-epp governors.
[ricklinux@eos-plasma ~]$ sudo pacman -Ss cpupower
[sudo] password for ricklinux:
extra/cpupower 6.3-2 (linux-tools) [installed]
Linux kernel tool to examine and tune power saving related features of your processor
[ricklinux@eos-plasma ~]$
Do i need to set anything?
Images got from: https://wiki.archlinux.org/title/CPU_frequency_scaling#top-page
Previous versions of cpupower didnāt include these new governors from amd-pstate-epp driver.
Since you already have it updated, you donāt need to do anything, unless you want to change the governor from powersave to performance.
