Hi Proxmox Community,
I have been primarily using HPE servers (G7,8,9) for years and have always had the Power Regulator set to OS Mode without any issues.
However, I recently switched to the DL325 Gen10 Plus and encountered some problems.
On the Gen10 host, I have configured the Workload Profile as "Virtualization - Power Efficient," which utilizes the OS mode power regulator accordingly. You can find an overview of the Workload Profiles here: Workload Profiles Overview
I have manually activated Turbo Boost.
The only difference between "Virtualization - Power Efficient" and "Virtualization - Max Performance" is the static max performance power regulator (which I want to avoid) and C-State 6 (I have not observed a core being turned off).
Here you can read what the exact workload profiles change in the bios settings: Workload Profiles BIOS Settings
You can check the available power settings here: Power Settings Overview
A pdf about HPE Power Management with RHEL and SUSE for Gen10 (intel) can be found here: Server Power Management PDF
Before putting the system into "production", I conducted some benchmarks and monitored the CPU frequency.
As you can see, the CPU is fully utilized, and the highest frequency of my AMD 7F52 CPU is achieved in the "Virtualization - Power Efficient" Workload Profile.
However, now the system is in "production", I've noticed that only about half of the CPU frequency is being used most of the time. The CPU usage averages around ~24%, unlike during the benchmark.
This leads to significant performance fluctuations in the VMs.
I can't quite explain this. I don't think it's a problem related to PVE 8 (my first system with PVE 8, the others were on version 7 or older).
Could the C-State 6 setting in the BIOS be responsible for this? I thought C-State 6 meant a core could be turned off, not that the cores would run at half frequency. Am I mistaken?
Now it gets really strange. When I check via cpupower, I see the following output:
It's using a significantly lower frequency than what is shown here.
What is going on? Do I need to install additional packages for acpi-cpufreq usage? I thought they were already installed automatically, and I don't recall manually installing packages for that.
Another peculiar aspect is that the boost state is not being displayed. It should be at 3900 MHz.
I want to mention that simply setting it to static max performance is not an ideal solution for me. There must be a reason why the OS Mode isn't working as intended. I have always believed that the OS Mode is best for virtualization without unnecessarily burning electricity.
I hope someone in the community has had similar experiences and found a solution.
I look forward to any help you can provide!
I have been primarily using HPE servers (G7,8,9) for years and have always had the Power Regulator set to OS Mode without any issues.
However, I recently switched to the DL325 Gen10 Plus and encountered some problems.
On the Gen10 host, I have configured the Workload Profile as "Virtualization - Power Efficient," which utilizes the OS mode power regulator accordingly. You can find an overview of the Workload Profiles here: Workload Profiles Overview
I have manually activated Turbo Boost.
The only difference between "Virtualization - Power Efficient" and "Virtualization - Max Performance" is the static max performance power regulator (which I want to avoid) and C-State 6 (I have not observed a core being turned off).
Here you can read what the exact workload profiles change in the bios settings: Workload Profiles BIOS Settings
You can check the available power settings here: Power Settings Overview
A pdf about HPE Power Management with RHEL and SUSE for Gen10 (intel) can be found here: Server Power Management PDF
Before putting the system into "production", I conducted some benchmarks and monitored the CPU frequency.
As you can see, the CPU is fully utilized, and the highest frequency of my AMD 7F52 CPU is achieved in the "Virtualization - Power Efficient" Workload Profile.
However, now the system is in "production", I've noticed that only about half of the CPU frequency is being used most of the time. The CPU usage averages around ~24%, unlike during the benchmark.
This leads to significant performance fluctuations in the VMs.
I can't quite explain this. I don't think it's a problem related to PVE 8 (my first system with PVE 8, the others were on version 7 or older).
Could the C-State 6 setting in the BIOS be responsible for this? I thought C-State 6 meant a core could be turned off, not that the cores would run at half frequency. Am I mistaken?
Now it gets really strange. When I check via cpupower, I see the following output:
Code:
analyzing CPU 0:
driver: acpi-cpufreq
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: 2.50 GHz - 3.50 GHz
available frequency steps: 3.50 GHz, 3.00 GHz, 2.50 GHz
available cpufreq governors: conservative ondemand userspace powersave performance schedutil
current policy: frequency should be within 2.50 GHz and 3.50 GHz.
The governor "performance" may decide which speed to use
within this range.
current CPU frequency: 3.50 GHz (asserted by call to hardware)
boost state support:
Supported: yes
Active: yes
Boost States: 0
Total States: 3
Pstate-P0: 3500MHz
Pstate-P1: 3000MHz
Pstate-P2: 2500MHzWhat is going on? Do I need to install additional packages for acpi-cpufreq usage? I thought they were already installed automatically, and I don't recall manually installing packages for that.
Another peculiar aspect is that the boost state is not being displayed. It should be at 3900 MHz.
I want to mention that simply setting it to static max performance is not an ideal solution for me. There must be a reason why the OS Mode isn't working as intended. I have always believed that the OS Mode is best for virtualization without unnecessarily burning electricity.
I hope someone in the community has had similar experiences and found a solution.
I look forward to any help you can provide!