AMD pstate driver steps and discussion

My doubt was whether the 2022 information is still valid.
My installation is fresh and totally updated. I have not installed/modified anything.
I expect proxmox to integrate the best drivers by default.

If you tell me that it is not so and that on the 8000 series mobile cpu it is better to switch to amd-pstate from acpi-cpufreq then I will do it.
It is always nice to have better performance but with lower consumption.
 
Mario Rossi said:
My doubt was whether the 2022 information is still valid.
My installation is fresh and totally updated. I have not installed/modified anything.
I expect proxmox to integrate the best drivers by default.

If you tell me that it is not so and that on the 8000 series mobile cpu it is better to switch to amd-pstate from acpi-cpufreq then I will do it.
It is always nice to have better performance but with lower consumption.
Click to expand...
I can't really answer that for you, but it's been better for me across two different machines, at least for keeping performance and dropping idle wattage. I imagine Proxmox uses acpi-cpufreq because it's supported by the most CPUs. Also, it's done at the kernel level, so they may not even have control over it. I don't really know.

My suggestion would be to try it if you want/need something better. You can always go back if you don't like it, just read the docs so you know what's going on. I seem to recall needing a kernel cmd-line parameter to re-enable acpi-cpufreq. Just removing the blacklist wasn't enough to go back to it. The info was easy to find last time I needed it, but I can't remember where I found that tid-bit.

Most if not all of the info about amd-pstate drivers can be found here:
https://docs.kernel.org/admin-guide/pm/amd-pstate.html
 
Among the various searches I found this link:

https://www.phoronix.com/review/amd-pstate-epp-ryzen-mobile

I added the string to my grub:
Code: 
GRUB_CMDLINE_LINUX_DEFAULT="quiet initcall_blacklist=acpi_cpufreq_init amd_pstate=active"

Then I had to modify the profile verification and set commands since I no longer use the governor but the drivers:
Code: 
cat /sys/devices/system/cpu/cpu0/cpufreq/energy_performance_available_preferences

my system has the following options available
Code: 
default performance balance_performance balance_power power

I chose balance_power because it almost reaches the ACPI power state but loses much less performance
Code: 
echo "balance_power" | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference


Code: 
root@pve2:~# grep "cpu MHz" /proc/cpuinfo
CPU MHz: 400,000
CPU MHz: 1771.374
CPU MHz: 400,000
CPU MHz: 4674.754
CPU MHz: 2336.478
CPU MHz: 2344.900
CPU MHz: 3276.278
CPU MHz: 1330.510
CPU MHz: 400,000
CPU MHz: 400,000
CPU MHz: 1597.927
CPU MHz: 400,000
CPU MHz: 400,000
CPU MHz: 400,000
CPU MHz: 400,000
CPU MHz: 400,000

root@pve2:~# cpupower frequency-info
analyzing CPU 0:
 driver: amd-pstate-epp
 CPUs which run at the same hardware frequency: 0
 CPUs which need to have their frequency coordinated by software: 0
 maximum transition latency: Cannot determine or is not supported.
 hardware limits: 400 MHz - 5.10 GHz
 available cpufreq governors: performance powersave
 current policy: frequency should be within 400 MHz and 5.10 GHz.
 The governor "powersave" may decide which speed to use
 within this range.
 current CPU frequency: Unable to call hardware
current CPU frequency: 400 MHz (asserted by call to kernel)
boost state support:
Supported: yes
Active: yes
Boost States: 0
Total States: 3
Pstate-P0: 3800MHz
Pstate-P1: 2200MHz
Pstate-P2: 1600MHz

now I have to find how to make this state fixed at boot
with the ACPI governor I had inserted in grub "cpufreq.default_governor=powersave", I have to find how to do it with the amd pstate
 
Mario Rossi said:
Among the various searches I found this link:

https://www.phoronix.com/review/amd-pstate-epp-ryzen-mobile

I added the string to my grub:
Code: 
GRUB_CMDLINE_LINUX_DEFAULT="quiet initcall_blacklist=acpi_cpufreq_init amd_pstate=active"

Then I had to modify the profile verification and set commands since I no longer use the governor but the drivers:
Code: 
cat /sys/devices/system/cpu/cpu0/cpufreq/energy_performance_available_preferences

my system has the following options available
Code: 
default performance balance_performance balance_power power

I chose balance_power because it almost reaches the ACPI power state but loses much less performance
Code: 
echo "balance_power" | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference


Code: 
root@pve2:~# grep "cpu MHz" /proc/cpuinfo
CPU MHz: 400,000
CPU MHz: 1771.374
CPU MHz: 400,000
CPU MHz: 4674.754
CPU MHz: 2336.478
CPU MHz: 2344.900
CPU MHz: 3276.278
CPU MHz: 1330.510
CPU MHz: 400,000
CPU MHz: 400,000
CPU MHz: 1597.927
CPU MHz: 400,000
CPU MHz: 400,000
CPU MHz: 400,000
CPU MHz: 400,000
CPU MHz: 400,000

root@pve2:~# cpupower frequency-info
analyzing CPU 0:
 driver: amd-pstate-epp
 CPUs which run at the same hardware frequency: 0
 CPUs which need to have their frequency coordinated by software: 0
 maximum transition latency: Cannot determine or is not supported.
 hardware limits: 400 MHz - 5.10 GHz
 available cpufreq governors: performance powersave
 current policy: frequency should be within 400 MHz and 5.10 GHz.
 The governor "powersave" may decide which speed to use
 within this range.
 current CPU frequency: Unable to call hardware
current CPU frequency: 400 MHz (asserted by call to kernel)
boost state support:
Supported: yes
Active: yes
Boost States: 0
Total States: 3
Pstate-P0: 3800MHz
Pstate-P1: 2200MHz
Pstate-P2: 1600MHz

now I have to find how to make this state fixed at boot
with the ACPI governor I had inserted in grub "cpufreq.default_governor=powersave", I have to find how to do it with the amd pstate
Click to expand...
I just have crontabs set up to do it at boot:

Code: 
@reboot sleep 5 && echo "powersave" | tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor
@reboot sleep 10 && echo "balance_power" | tee /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference

EDIT: Another good thing is if you have emails set up from Proxmox, you'll get an email about the crontab completing. I use it as a sort of pseudo alert that I had a power or system failure.
 
Last edited:
Mario Rossi said:
Among the various searches I found this link:

https://www.phoronix.com/review/amd-pstate-epp-ryzen-mobile

I added the string to my grub:
Code: 
GRUB_CMDLINE_LINUX_DEFAULT="quiet initcall_blacklist=acpi_cpufreq_init amd_pstate=active"

Then I had to modify the profile verification and set commands since I no longer use the governor but the drivers:
Code: 
cat /sys/devices/system/cpu/cpu0/cpufreq/energy_performance_available_preferences

my system has the following options available
Code: 
default performance balance_performance balance_power power

I chose balance_power because it almost reaches the ACPI power state but loses much less performance
Code: 
echo "balance_power" | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference


Code: 
root@pve2:~# grep "cpu MHz" /proc/cpuinfo
CPU MHz: 400,000
CPU MHz: 1771.374
CPU MHz: 400,000
CPU MHz: 4674.754
CPU MHz: 2336.478
CPU MHz: 2344.900
CPU MHz: 3276.278
CPU MHz: 1330.510
CPU MHz: 400,000
CPU MHz: 400,000
CPU MHz: 1597.927
CPU MHz: 400,000
CPU MHz: 400,000
CPU MHz: 400,000
CPU MHz: 400,000
CPU MHz: 400,000

root@pve2:~# cpupower frequency-info
analyzing CPU 0:
 driver: amd-pstate-epp
 CPUs which run at the same hardware frequency: 0
 CPUs which need to have their frequency coordinated by software: 0
 maximum transition latency: Cannot determine or is not supported.
 hardware limits: 400 MHz - 5.10 GHz
 available cpufreq governors: performance powersave
 current policy: frequency should be within 400 MHz and 5.10 GHz.
 The governor "powersave" may decide which speed to use
 within this range.
 current CPU frequency: Unable to call hardware
current CPU frequency: 400 MHz (asserted by call to kernel)
boost state support:
Supported: yes
Active: yes
Boost States: 0
Total States: 3
Pstate-P0: 3800MHz
Pstate-P1: 2200MHz
Pstate-P2: 1600MHz

now I have to find how to make this state fixed at boot
with the ACPI governor I had inserted in grub "cpufreq.default_governor=powersave", I have to find how to do it with the amd pstate
Click to expand...
Thank you so much for sharing this!!

I run Proxmox 8.2-1 on an AMD Ryzen 5950X and had with chatGPT made a script to initialize the server. This was my GRUB part
# Modify /etc/default/grub
echo "Modifying /etc/default/grub..."
sed -i 's/^GRUB_CMDLINE_LINUX_DEFAULT=.*/GRUB_CMDLINE_LINUX_DEFAULT="quiet splash amd_pstate=active amd_iommu=on iommu=pt pcie_acs_override=downstream,multifunction video=HDMI-A-5:1920x1080@60e"/' /etc/default/grub
echo "Successfully modified /etc/default/grub."

I didnt blocklist the driver still it says
root@x:~# cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
amd-pstate-epp

root@x:~# cat /sys/devices/system/cpu/cpu0/cpufreq/energy_performance_available_preferences
performance

So it seems it only does performance?

After reading the link you shared I would love to use amd_pstate_epp powersave power

ChatGPT sais to do GRUB_CMDLINE_LINUX_DEFAULT="quiet splash amd_pstate=active amd_pstate_epp=powersave power" But I think it might halucinate.

Yet another instance ChatGPT advices me to make changes to the driver and make them persistent after reboot by making a service.

Would anybody be so kind to help me with the right steps to implement "amd_pstate_epp powersave power" on this AMD Ryzen 5950X running Proxmox 8.2-1 kernel 6.8.4-2-pve (yes because of modified nVidiaGRID drivers I can't update the kernel)?

Thank you!
 
If you've recently installed Proxmox just like me and need to adjust kernel parameters (e.g., for CPU power management), you might run into issues when using proxmox-boot-tool. Unlike GRUB, which is commonly used on Linux, Proxmox often defaults to systemd-boot when using UEFI. This guide explains how to configure Proxmox’s boot tool correctly and adjust CPU frequency settings.

---

Step 1: Check Your Bootloader

First, determine whether your system is using proxmox-boot-tool (which means systemd-boot instead of GRUB).

Run:
Bash: 
proxmox-boot-tool status

If you see something like:
Bash: 
System currently booted with uefi
<UUID> is configured with: uefi (versions: 6.8.12-4-pve)
your system is using systemd-boot, and you should not edit /etc/default/grub. Instead, modify /etc/kernel/cmdline.

---

Step 2: Identify the Correct root= Parameter

Proxmox requires a valid root= parameter in /etc/kernel/cmdline. If this is missing, running proxmox-boot-tool refresh will result in:

Bash: 
No root= parameter in /etc/kernel/cmdline found!


To find the correct root partition, use:
Bash: 
findmnt /

Example output:
Bash: 
TARGET    SOURCE               FSTYPE OPTIONS
/         /dev/mapper/pve-root ext4   rw,relatime,errors=remount-ro
In this case, the root partition is /dev/mapper/pve-root (common with LVM installations). If using ZFS, it might look like:
Bash: 
/      rpool/ROOT/pve-1    zfs    rw,relatime
where rpool/ROOT/pve-1 is the ZFS root.

---

Step 3: Update Kernel Parameters

Now, edit /etc/kernel/cmdline:
Bash: 
nano /etc/kernel/cmdline

Modify it to include the correct root partition:

- For LVM-based installs (EXT4/BTRFS):
Bash: 
root=/dev/mapper/pve-root quiet amd_pstate=passive

- For ZFS-based installs:
Bash: 
root=ZFS=rpool/ROOT/pve-1 quiet amd_pstate=passive

⚠ Ensure that everything is on a single line with no line breaks!

---

Step 4: Apply Changes

After updating /etc/kernel/cmdline, refresh the boot configuration:
Bash: 
proxmox-boot-tool refresh

If no errors appear, reboot your system:
Bash: 
reboot

---

Step 5: Verify amd_pstate is Active

After rebooting, check whether amd_pstateis correctly loaded:
Bash: 
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver

If the output is:
Bash: 
amd_pstate
then the new driver is active.

To check the minimum allowed CPU frequency:
Bash: 
cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_min_freq

If you want to lower it further, try:
Bash: 
echo 800000 | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_min_freq
(800000 represents 800MHz; adjust as needed.)

---

What If My CPU Doesn't Go Below 2.2GHz?

If your CPU still doesn’t go below a certain frequency (e.g., 2.2 GHz), it might be limited by:
1. Your processor's hardware P-state restrictions.
2. BIOS/UEFI settings—some CPUs allow setting a lower minimum frequency via the BIOS.

---

Final Notes

- Do NOT edit /etc/default/grub on systems using proxmox-boot-tool. Use /etc/kernel/cmdline instead.
- Always run proxmox-boot-tool refresh after modifying kernel parameters.
- Check /proc/cmdline after reboot to confirm that the parameters are applied:
Bash: 
cat /proc/cmdline
- If using amd_pstate, ensure your kernel supports it (Proxmox 6.8+ is recommended).

This guide should help users properly configure proxmox-boot-tool and adjust CPU power management settings without running into boot issues.
 
Last edited:
rvdmeer said:
If you've recently installed Proxmox just like me and need to adjust kernel parameters (e.g., for CPU power management), you might run into issues when using proxmox-boot-tool. Unlike GRUB, which is commonly used on Linux, Proxmox often defaults to systemd-boot when using UEFI. This guide explains how to configure Proxmox’s boot tool correctly and adjust CPU frequency settings.

---

Step 1: Check Your Bootloader

First, determine whether your system is using proxmox-boot-tool (which means systemd-boot instead of GRUB).

Run:
Bash: 
proxmox-boot-tool status

If you see something like:
Bash: 
System currently booted with uefi
<UUID> is configured with: uefi (versions: 6.8.12-4-pve)
your system is using systemd-boot, and you should not edit /etc/default/grub. Instead, modify /etc/kernel/cmdline.

---

Step 2: Identify the Correct root= Parameter

Proxmox requires a valid root= parameter in /etc/kernel/cmdline. If this is missing, running proxmox-boot-tool refresh will result in:

Bash: 
No root= parameter in /etc/kernel/cmdline found!


To find the correct root partition, use:
Bash: 
findmnt /

Example output:
Bash: 
TARGET    SOURCE               FSTYPE OPTIONS
/         /dev/mapper/pve-root ext4   rw,relatime,errors=remount-ro
In this case, the root partition is /dev/mapper/pve-root (common with LVM installations). If using ZFS, it might look like:
Bash: 
/      rpool/ROOT/pve-1    zfs    rw,relatime
where rpool/ROOT/pve-1 is the ZFS root.

---

Step 3: Update Kernel Parameters

Now, edit /etc/kernel/cmdline:
Bash: 
nano /etc/kernel/cmdline

Modify it to include the correct root partition:

- For LVM-based installs (EXT4/BTRFS):
Bash: 
root=/dev/mapper/pve-root quiet amd_pstate=passive

- For ZFS-based installs:
Bash: 
root=ZFS=rpool/ROOT/pve-1 quiet amd_pstate=passive

⚠ Ensure that everything is on a single line with no line breaks!

---

Step 4: Apply Changes

After updating /etc/kernel/cmdline, refresh the boot configuration:
Bash: 
proxmox-boot-tool refresh

If no errors appear, reboot your system:
Bash: 
reboot

---

Step 5: Verify amd_pstate is Active

After rebooting, check whether amd_pstateis correctly loaded:
Bash: 
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver

If the output is:
Bash: 
amd_pstate
then the new driver is active.

To check the minimum allowed CPU frequency:
Bash: 
cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_min_freq

If you want to lower it further, try:
Bash: 
echo 800000 | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_min_freq
(800000 represents 800MHz; adjust as needed.)

---

What If My CPU Doesn't Go Below 2.2GHz?

If your CPU still doesn’t go below a certain frequency (e.g., 2.2 GHz), it might be limited by:
1. Your processor's hardware P-state restrictions.
2. BIOS/UEFI settings—some CPUs allow setting a lower minimum frequency via the BIOS.

---

Final Notes

- Do NOT edit /etc/default/grub on systems using proxmox-boot-tool. Use /etc/kernel/cmdline instead.
- Always run proxmox-boot-tool refresh after modifying kernel parameters.
- Check /proc/cmdline after reboot to confirm that the parameters are applied:
Bash: 
cat /proc/cmdline
- If using amd_pstate, ensure your kernel supports it (Proxmox 6.8+ is recommended).

This guide should help users properly configure proxmox-boot-tool and adjust CPU power management settings without running into boot issues.
Click to expand...
Thank you for your tutorial. Yet my system seems to succesfully use GRUB

root@xxxx:~# proxmox-boot-tool status
Re-executing '/usr/sbin/proxmox-boot-tool' in new private mount namespace..
E: /etc/kernel/proxmox-boot-uuids does not exist.

Could you please help me how to implement "amd_pstate_epp powersave power" on this AMD Ryzen 5950X running Proxmox 8.2-1 kernel 6.8.4-2-pve when defaulted to GRUB?

Thank you!
 
I also dont get amd pstate working on my older Cluster.
I have a machine with an amd 5700G and one with a 5900X
the 5900x was setup years ago with kernel 5.10
the 5700G later, actually I dont know when but at least a year later

machine with 5700G has amd pstate active
5900X has not

setting up the suggested kernel settings in the grub konfiguration and blacklisting acpi results in no scaling driver loaded at all.

Any suggestions on how to get AMDs pstate driver loaded on the 5900X machine too?

I was using

initcall_blacklist=acpi_cpufreq_init
amd_pstate.shared_mem=1

as suggested, no luck.
 
Dutch87 said:
Thank you so much for sharing this!!

I run Proxmox 8.2-1 on an AMD Ryzen 5950X and had with chatGPT made a script to initialize the server. This was my GRUB part
# Modify /etc/default/grub
echo "Modifying /etc/default/grub..."
sed -i 's/^GRUB_CMDLINE_LINUX_DEFAULT=.*/GRUB_CMDLINE_LINUX_DEFAULT="quiet splash amd_pstate=active amd_iommu=on iommu=pt pcie_acs_override=downstream,multifunction video=HDMI-A-5:1920x1080@60e"/' /etc/default/grub
echo "Successfully modified /etc/default/grub."

I didnt blocklist the driver still it says
root@x:~# cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
amd-pstate-epp

root@x:~# cat /sys/devices/system/cpu/cpu0/cpufreq/energy_performance_available_preferences
performance

So it seems it only does performance?

After reading the link you shared I would love to use amd_pstate_epp powersave power

ChatGPT sais to do GRUB_CMDLINE_LINUX_DEFAULT="quiet splash amd_pstate=active amd_pstate_epp=powersave power" But I think it might halucinate.

Yet another instance ChatGPT advices me to make changes to the driver and make them persistent after reboot by making a service.

Would anybody be so kind to help me with the right steps to implement "amd_pstate_epp powersave power" on this AMD Ryzen 5950X running Proxmox 8.2-1 kernel 6.8.4-2-pve (yes because of modified nVidiaGRID drivers I can't update the kernel)?

Thank you!
Click to expand...
One year later... Anybody get the amd_pstate_epp powersave power to work on Proxmox? if so, how?
 
Dutch87 said:
One year later... Anybody get the amd_pstate_epp powersave power to work on Proxmox? if so, how?
Click to expand...

On my Home Server:
Code: 
root@pve:~# cpupower frequency-info
analyzing CPU 26:
  driver: amd-pstate-epp
  CPUs which run at the same hardware frequency: 26
  CPUs which need to have their frequency coordinated by software: 26
  maximum transition latency:  Cannot determine or is not supported.
  hardware limits: 1.22 GHz - 5.06 GHz
  available cpufreq governors: performance powersave
  current policy: frequency should be within 1.49 GHz and 5.06 GHz.
                  The governor "performance" may decide which speed to use
                  within this range.
  current CPU frequency: Unable to call hardware
  current CPU frequency: 4.74 GHz (asserted by call to kernel)
  boost state support:
    Supported: yes
    Active: yes
    AMD PSTATE Highest Performance: 166. Maximum Frequency: 5.06 GHz.
    AMD PSTATE Nominal Performance: 82. Nominal Frequency: 2.50 GHz.
    AMD PSTATE Lowest Non-linear Performance: 49. Lowest Non-linear Frequency: 1.49 GHz.
    AMD PSTATE Lowest Performance: 40. Lowest Frequency: 600 MHz.
root@pve:~# cat /proc/cpuinfo
processor       : 0
vendor_id       : AuthenticAMD
cpu family      : 26
model           : 68
model name      : AMD Ryzen 9 9955HX 16-Core Processor
stepping        : 0
microcode       : 0xb404024
cpu MHz         : 1493.903
cache size      : 1024 KB
physical id     : 0
siblings        : 32
core id         : 0
cpu cores       : 16
apicid          : 0
initial apicid  : 0
fpu             : yes
fpu_exception   : yes
cpuid level     : 16
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good amd_lbr_v2 nopl xtopology nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpuid_fault cpb cat_l3 cdp_l3 hw_pstate ssbd mba perfmon_v2 ibrs ibpb stibp ibrs_enhanced vmmcall fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local user_shstk avx_vnni avx512_bf16 clzero irperf xsaveerptr rdpru wbnoinvd cppc arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold v_vmsave_vmload vgif x2avic v_spec_ctrl vnmi avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid bus_lock_detect movdiri movdir64b overflow_recov succor smca fsrm avx512_vp2intersect flush_l1d amd_lbr_pmc_freeze
bugs            : sysret_ss_attrs spectre_v1 spectre_v2 spec_store_bypass srso spectre_v2_user vmscape
bogomips        : 4990.91
TLB size        : 192 4K pages
clflush size    : 64
cache_alignment : 64
address sizes   : 48 bits physical, 48 bits virtual
power management: ts ttp tm hwpstate cpb eff_freq_ro [13] [14]

On my Server:
Code: 
root@pve-hdr:~# cpupower frequency-info
analyzing CPU 23:
  driver: amd-pstate-epp
  CPUs which run at the same hardware frequency: 23
  CPUs which need to have their frequency coordinated by software: 23
  maximum transition latency:  Cannot determine or is not supported.
  hardware limits: 405 MHz - 4.31 GHz
  available cpufreq governors: performance powersave
  current policy: frequency should be within 2.60 GHz and 4.31 GHz.
                  The governor "performance" may decide which speed to use
                  within this range.
  current CPU frequency: Unable to call hardware
  current CPU frequency: 4.34 GHz (asserted by call to kernel)
  boost state support:
    Supported: yes
    Active: yes
    AMD PSTATE Highest Performance: 255. Maximum Frequency: 4.31 GHz.
    AMD PSTATE Nominal Performance: 228. Nominal Frequency: 3.85 GHz.
    AMD PSTATE Lowest Non-linear Performance: 154. Lowest Non-linear Frequency: 2.60 GHz.
    AMD PSTATE Lowest Performance: 24. Lowest Frequency: 400 MHz.
root@pve-hdr:~# cat /proc/cpuinfo
processor       : 0
vendor_id       : AuthenticAMD
cpu family      : 25
model           : 17
model name      : AMD EPYC 9374F 32-Core Processor
stepping        : 1
microcode       : 0xa101154
cpu MHz         : 4342.726
cache size      : 1024 KB
physical id     : 0
siblings        : 64
core id         : 0
cpu cores       : 32
apicid          : 0
initial apicid  : 0
fpu             : yes
fpu_exception   : yes
cpuid level     : 16
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good amd_lbr_v2 nopl xtopology nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpuid_fault cpb cat_l3 cdp_l3 hw_pstate ssbd mba perfmon_v2 ibrs ibpb stibp ibrs_enhanced vmmcall fsgsbase bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local user_shstk avx512_bf16 clzero irperf xsaveerptr rdpru wbnoinvd amd_ppin cppc arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif x2avic v_spec_ctrl vnmi avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq la57 rdpid overflow_recov succor smca fsrm flush_l1d debug_swap
bugs            : sysret_ss_attrs spectre_v1 spectre_v2 spec_store_bypass srso spectre_v2_user tsa vmscape
bogomips        : 7776.94
TLB size        : 3584 4K pages
clflush size    : 64
cache_alignment : 64
address sizes   : 52 bits physical, 57 bits virtual
power management: ts ttp tm hwpstate cpb eff_freq_ro [13] [14]

powersave seems to be working without issues and no special cmdline is required:

Code: 
cat /proc/cmdline
BOOT_IMAGE=/vmlinuz-6.17.9-1-pve root=UUID=7826786a-27c0-43f6-a1b7-7c46a3210775 ro quiet

But there are usually much more modes available, im not sure but maybe those are only for laptops.
However the Server is set in BIOS everywhere to Performance with Overclocking and some other High-Performance profiles.
The Home Server has default Bios Settings, i believe everything balanced or with whatever default Bios Settings those Mini-PC's come.

So Bios doesnt seem to matter, probably the Kernel version does.
Cheers
 
Dutch87 said:
One year later... Anybody get the amd_pstate_epp powersave power to work on Proxmox? if so, how?
Click to expand...
Three years ago (post #8) and until some more recent time, I would see speeds go down to 550Mhz (Lowest) using guided mode with schedutil, but currently (kernel 6.17) it only goes down to 1.7Ghz (Lowest Non-linear). It appears to run fine in active mode with powersave (instead of the default performance) but only down to 1.7Ghz and not 550Mhz anymore.
 
Ramalama said:
On my Home Server:
Code: 
root@pve:~# cpupower frequency-info
analyzing CPU 26:
  driver: amd-pstate-epp
  CPUs which run at the same hardware frequency: 26
  CPUs which need to have their frequency coordinated by software: 26
  maximum transition latency:  Cannot determine or is not supported.
  hardware limits: 1.22 GHz - 5.06 GHz
  available cpufreq governors: performance powersave
  current policy: frequency should be within 1.49 GHz and 5.06 GHz.
                  The governor "performance" may decide which speed to use
                  within this range.
  current CPU frequency: Unable to call hardware
  current CPU frequency: 4.74 GHz (asserted by call to kernel)
  boost state support:
    Supported: yes
    Active: yes
    AMD PSTATE Highest Performance: 166. Maximum Frequency: 5.06 GHz.
    AMD PSTATE Nominal Performance: 82. Nominal Frequency: 2.50 GHz.
    AMD PSTATE Lowest Non-linear Performance: 49. Lowest Non-linear Frequency: 1.49 GHz.
    AMD PSTATE Lowest Performance: 40. Lowest Frequency: 600 MHz.
root@pve:~# cat /proc/cpuinfo
processor       : 0
vendor_id       : AuthenticAMD
cpu family      : 26
model           : 68
model name      : AMD Ryzen 9 9955HX 16-Core Processor
stepping        : 0
microcode       : 0xb404024
cpu MHz         : 1493.903
cache size      : 1024 KB
physical id     : 0
siblings        : 32
core id         : 0
cpu cores       : 16
apicid          : 0
initial apicid  : 0
fpu             : yes
fpu_exception   : yes
cpuid level     : 16
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good amd_lbr_v2 nopl xtopology nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpuid_fault cpb cat_l3 cdp_l3 hw_pstate ssbd mba perfmon_v2 ibrs ibpb stibp ibrs_enhanced vmmcall fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local user_shstk avx_vnni avx512_bf16 clzero irperf xsaveerptr rdpru wbnoinvd cppc arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold v_vmsave_vmload vgif x2avic v_spec_ctrl vnmi avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid bus_lock_detect movdiri movdir64b overflow_recov succor smca fsrm avx512_vp2intersect flush_l1d amd_lbr_pmc_freeze
bugs            : sysret_ss_attrs spectre_v1 spectre_v2 spec_store_bypass srso spectre_v2_user vmscape
bogomips        : 4990.91
TLB size        : 192 4K pages
clflush size    : 64
cache_alignment : 64
address sizes   : 48 bits physical, 48 bits virtual
power management: ts ttp tm hwpstate cpb eff_freq_ro [13] [14]

On my Server:
Code: 
root@pve-hdr:~# cpupower frequency-info
analyzing CPU 23:
  driver: amd-pstate-epp
  CPUs which run at the same hardware frequency: 23
  CPUs which need to have their frequency coordinated by software: 23
  maximum transition latency:  Cannot determine or is not supported.
  hardware limits: 405 MHz - 4.31 GHz
  available cpufreq governors: performance powersave
  current policy: frequency should be within 2.60 GHz and 4.31 GHz.
                  The governor "performance" may decide which speed to use
                  within this range.
  current CPU frequency: Unable to call hardware
  current CPU frequency: 4.34 GHz (asserted by call to kernel)
  boost state support:
    Supported: yes
    Active: yes
    AMD PSTATE Highest Performance: 255. Maximum Frequency: 4.31 GHz.
    AMD PSTATE Nominal Performance: 228. Nominal Frequency: 3.85 GHz.
    AMD PSTATE Lowest Non-linear Performance: 154. Lowest Non-linear Frequency: 2.60 GHz.
    AMD PSTATE Lowest Performance: 24. Lowest Frequency: 400 MHz.
root@pve-hdr:~# cat /proc/cpuinfo
processor       : 0
vendor_id       : AuthenticAMD
cpu family      : 25
model           : 17
model name      : AMD EPYC 9374F 32-Core Processor
stepping        : 1
microcode       : 0xa101154
cpu MHz         : 4342.726
cache size      : 1024 KB
physical id     : 0
siblings        : 64
core id         : 0
cpu cores       : 32
apicid          : 0
initial apicid  : 0
fpu             : yes
fpu_exception   : yes
cpuid level     : 16
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good amd_lbr_v2 nopl xtopology nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpuid_fault cpb cat_l3 cdp_l3 hw_pstate ssbd mba perfmon_v2 ibrs ibpb stibp ibrs_enhanced vmmcall fsgsbase bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local user_shstk avx512_bf16 clzero irperf xsaveerptr rdpru wbnoinvd amd_ppin cppc arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif x2avic v_spec_ctrl vnmi avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq la57 rdpid overflow_recov succor smca fsrm flush_l1d debug_swap
bugs            : sysret_ss_attrs spectre_v1 spectre_v2 spec_store_bypass srso spectre_v2_user tsa vmscape
bogomips        : 7776.94
TLB size        : 3584 4K pages
clflush size    : 64
cache_alignment : 64
address sizes   : 52 bits physical, 57 bits virtual
power management: ts ttp tm hwpstate cpb eff_freq_ro [13] [14]

powersave seems to be working without issues and no special cmdline is required:

Code: 
cat /proc/cmdline
BOOT_IMAGE=/vmlinuz-6.17.9-1-pve root=UUID=7826786a-27c0-43f6-a1b7-7c46a3210775 ro quiet

But there are usually much more modes available, im not sure but maybe those are only for laptops.
However the Server is set in BIOS everywhere to Performance with Overclocking and some other High-Performance profiles.
The Home Server has default Bios Settings, i believe everything balanced or with whatever default Bios Settings those Mini-PC's come.

So Bios doesnt seem to matter, probably the Kernel version does.
Cheers
Click to expand...
Yes, So you got the same problem, you only have:
available cpufreq governors: performance powersave

But how to get it to do amd_pstate_epp powersave power?

My system is:
# uname -r
6.8.4-2-pve
# cpupower frequency-info
analyzing CPU 0:
driver: amd-pstate-epp
CPUs which run at the same hardware frequency: 0
CPUs which need to have their frequency coordinated by software: 0
maximum transition latency: Cannot determine or is not supported.
hardware limits: 550 MHz - 5.08 GHz
available cpufreq governors: performance powersave
current policy: frequency should be within 550 MHz and 5.08 GHz.
The governor "performance" may decide which speed to use
within this range.
current CPU frequency: Unable to call hardware
current CPU frequency: 550 MHz (asserted by call to kernel)
boost state support:
Supported: yes
Active: yes
Boost States: 0
Total States: 3
Pstate-P0: 3400MHz
Pstate-P1: 2800MHz
Pstate-P2: 2200MHz

And it does go down to 550mhz and It is actually using P-states..
I use this line in GRUB:
GRUB_CMDLINE_LINUX_DEFAULT=.*/GRUB_CMDLINE_LINUX_DEFAULT="quiet splash amd_pstate=active amd_iommu=on iommu=pt pcie_acs_override=downstream,multifunction video=HDMI-A-5:1920x1080@60e"

I still like to know how he enabled amd_pstate_epp powersave power:
https://www.phoronix.com/review/amd-pstate-epp-ryzen-mobile

This was in 2023 and with a 6.3 Kernel.... Somebody must know something?
 
Last edited:
Dutch87 said:
Yes, So you got the same problem, you only have:
available cpufreq governors: performance powersave
Click to expand...
This is normal. If it is set to active (instead of passive or guided) then only performance and powersave are supported: https://docs.kernel.org/admin-guide/pm/amd-pstate.html
EDIT: Also, https://wiki.archlinux.org/title/CPU_frequency_scaling as all this is not Proxmox specific.
Dutch87 said:
But how to get it to do amd_pstate_epp powersave power?
Click to expand...
You get active (epp) by default and then set the governor to powersave: echo powersave | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor
 
Last edited:
  • Like
Reactions: Dutch87
Dutch87 said:
Yes, So you got the same problem, you only have:
available cpufreq governors: performance powersave

But how to get it to do amd_pstate_epp powersave power?

My system is:
# uname -r
6.8.4-2-pve
# cpupower frequency-info
analyzing CPU 0:
driver: amd-pstate-epp
CPUs which run at the same hardware frequency: 0
CPUs which need to have their frequency coordinated by software: 0
maximum transition latency: Cannot determine or is not supported.
hardware limits: 550 MHz - 5.08 GHz
available cpufreq governors: performance powersave
current policy: frequency should be within 550 MHz and 5.08 GHz.
The governor "performance" may decide which speed to use
within this range.
current CPU frequency: Unable to call hardware
current CPU frequency: 550 MHz (asserted by call to kernel)
boost state support:
Supported: yes
Active: yes
Boost States: 0
Total States: 3
Pstate-P0: 3400MHz
Pstate-P1: 2800MHz
Pstate-P2: 2200MHz

And it does go down to 550mhz and It is actually using P-states..
I use this line in GRUB:
GRUB_CMDLINE_LINUX_DEFAULT=.*/GRUB_CMDLINE_LINUX_DEFAULT="quiet splash amd_pstate=active amd_iommu=on iommu=pt pcie_acs_override=downstream,multifunction video=HDMI-A-5:1920x1080@60e"

I still like to know how he enabled amd_pstate_epp powersave power:
https://www.phoronix.com/review/amd-pstate-epp-ryzen-mobile

This was in 2023 and with a 6.3 Kernel.... Somebody must know something?
Click to expand...
So ChatGPT got a bit smarter and I guess I got a step closer to solving it. The BIOS/firmware must expose it:

-----------------------------------------------------
ChatGPT 5.2 Thinking
Manual: enable amd-pstate-epp + governor=powersave + EPP=power

Step 1 — Add kernel parameters (bootloader)

If you use GRUB:
nano /etc/default/grub
Add to GRUB_CMDLINE_LINUX_DEFAULT=:
amd_pstate=active initcall_blacklist=acpi_cpufreq_init

Apply + reboot:
update-grub
reboot

If you use Proxmox boot-tool (systemd-boot on UEFI):
nano /etc/kernel/cmdline
Append:
amd_pstate=active initcall_blacklist=acpi_cpufreq_init

Apply + reboot:
proxmox-boot-tool refresh
reboot

Step 2 — Verify amd-pstate-epp is active
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
cat /sys/devices/system/cpu/amd_pstate/status

Expected:
amd-pstate-epp
active

Step 3 — Verify the BIOS exposes power as an EPP choice
cat /sys/devices/system/cpu/cpu0/cpufreq/energy_performance_available_preferences
You must see power (often also balance_power, etc.). If you only see performance, you cannot set power.

Step 4 — Set cpufreq governor to powersave (all CPUs)
echo powersave | tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

Step 5 — Set EPP to power (all CPUs)
echo power | tee /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference

Step 6 — Confirm the configuration
cpupower frequency-info | sed -n '1,25p'
echo -n "governor="; cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
echo -n "epp="; cat /sys/devices/system/cpu/cpu0/cpufreq/energy_performance_preference

Step 7 — Make it persistent (@reboot, forum-style)
crontab -e

Add:
@reboot sleep 5 && echo "powersave" | tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor
@reboot sleep 10 && echo "power" | tee /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference

-----------------------------------------

Ad I am a noob I do not take any responsability for this tutorial, but it makes sense to me...
Im going to dig in further to see how to let the ASUS BIOS expose it to the driver. I did enable CPPC/CPPC preferred cores.

To be updated...
 
leesteken said:
This is normal. If it is set to active (instead of passive or guided) then only performance and powersave are supported: https://docs.kernel.org/admin-guide/pm/amd-pstate.html
EDIT: Also, https://wiki.archlinux.org/title/CPU_frequency_scaling as all this is not Proxmox specific.

You get active (epp) by default and then set the governor to powersave: echo powersave | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor
Click to expand...
Thank you! Im reading the links and working on it. Lets see if I get it exposed from the bios and then to get it to work
 
Dutch87 said:
So ChatGPT got a bit smarter and I guess I got a step closer to solving it. The BIOS/firmware must expose it:

-----------------------------------------------------
ChatGPT 5.2 Thinking
Manual: enable amd-pstate-epp + governor=powersave + EPP=power

Step 1 — Add kernel parameters (bootloader)

If you use GRUB:
nano /etc/default/grub
Add to GRUB_CMDLINE_LINUX_DEFAULT=:
amd_pstate=active initcall_blacklist=acpi_cpufreq_init

Apply + reboot:
update-grub
reboot

If you use Proxmox boot-tool (systemd-boot on UEFI):
nano /etc/kernel/cmdline
Append:
amd_pstate=active initcall_blacklist=acpi_cpufreq_init

Apply + reboot:
proxmox-boot-tool refresh
reboot

Step 2 — Verify amd-pstate-epp is active
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
cat /sys/devices/system/cpu/amd_pstate/status

Expected:
amd-pstate-epp
active

Step 3 — Verify the BIOS exposes power as an EPP choice
cat /sys/devices/system/cpu/cpu0/cpufreq/energy_performance_available_preferences
You must see power (often also balance_power, etc.). If you only see performance, you cannot set power.

Step 4 — Set cpufreq governor to powersave (all CPUs)
echo powersave | tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

Step 5 — Set EPP to power (all CPUs)
echo power | tee /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference

Step 6 — Confirm the configuration
cpupower frequency-info | sed -n '1,25p'
echo -n "governor="; cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
echo -n "epp="; cat /sys/devices/system/cpu/cpu0/cpufreq/energy_performance_preference

Step 7 — Make it persistent (@reboot, forum-style)
crontab -e

Add:
@reboot sleep 5 && echo "powersave" | tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor
@reboot sleep 10 && echo "power" | tee /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference

-----------------------------------------

Ad I am a noob I do not take any responsability for this tutorial, but it makes sense to me...
Im going to dig in further to see how to let the ASUS BIOS expose it to the driver. I did enable CPPC/CPPC preferred cores.

To be updated...
Click to expand...
End conclusion, without the BIOS exposing the full CPPC its stuck at Performance since only that is exposed. The bios is set correctly but wont expose on the Asus ROG STRIX B550-F GAMING Motherboard with the AMD 5950x... When it comes to my system case closed
 
You must log in or register to reply here.
Top Bottom
Back