Proxmox powersaving question.

C

ComfySofa

Active Member
Feb 7, 2021
121
4
58
57
Hi there, wonder i could have some peoples input please.

Ive just upgraded one of my proxmox boxes to a Dell Optiplex 7090 USFF. Spec is as follows

i9-11900
2 internal ssd's 1x 4tb lexar and 1x 2tb Samsung
2x external spinning discs on USB
1x intenal USBc module
64gb memory
I usually use this guide for setting up the powersaving...

https://technologiehub.at/project-posts/tutorial/guide-for-proxmox-powersaving/

I kinda just applied it and not really looked any further but ive gone from an old optiplex i7 7th gen usff to an 11th gen one i was interested to see how much i saved (going newer) or used for the same reason.

i applied the powersaving profile and its locked the cpu at about 1.2ghz ish...i thought the powersave was supposed to ramp up the cpu under load? but it doesnt seem like it is?

Obviously if i apply the performance profile the cpu ramps up close to the max.

Is that correct...in that the powersave from what i can see throttles the cpu and it doesnt ramp up or should it?

Thanks for any replies.
 
There is no single `powersave`. Different drivers implement it differently. Some throttle up, some don't. To reason about it you need to know which one is used. (Note CPUs are usually supported by several different drivers.) Use e.g. `cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver` (or `cpupower frequency-info` if you installed it) to show it.
 
Last edited:
This is a homelab experience, in enterprise no one cares about consumption, really!

In the end, but hardware that is known for their power saving capacity like mobile targeted devices or even explicitly low power devices / cpus. The savings are EXTREMELY different. I also tried to build a powerful machine with good savings, but a low power machine under heavy load is still 4x less power hungry than the beefy machine in idle mode.

For me, I split my workload so that things that needs 24/7 are running on a ODROID H3 with 64 GB of RAM and two enterprise SSDs in a mirrored SSD setup with an M2. optane of 16 GB running 8-9W all day long. If I need more power, I just start different beefier machines on demand and I'm very happy with this setup.
 
  • Like
Reactions: UdoB
Well in the UK we have the highest electricity bills in the world...yes the world. My monthly consumption is about 370GBP for an average house...so for a machine running 24/7 powersaving is everything.

And yes like you i have a high powered box that i only power up when i use it.
 
Last edited:
tschmidt said:
There is no single `powersave`. Different drivers implement it differently. Some throttle up, some don't. To reason about it you need to know which one is used. (Note CPUs are usually supported by several different drivers.) Use e.g. `cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver` (or `cpupower frequency-info` if you installed it) to show it.
Click to expand...
Yes i know about that command im currently manually toggling between performance and powersave...i was hoping one profile would do both... doesnt look like the case though.
 
To clarify, are you happy with the power usage of the 7090 but are just concerned that the CPU doesn't ramp up when the load increases?

I have looked at similar guides but have always avoided installing anything at all into my PVE hosts.

For my hosts, I only use "powersave" for "/sys/devices/system/cpu/cpu*/cpufreq/scaling_governor" and that has been OK so far.
 
Actually, there are states in the US that pay far more overall for electricity, due to extreme summer heat, such as Arizona (ranked #1, with average summer bills of $1,000USD). Or certain areas of California where summer heat combined with peak usage rates can lead to bills upwards of $700USD per month.
 
Hi there, thanks very much for the link. Yes, ive installed powertop - as i said at the start of the thread...i followed the webpage...im not great with proxmox and always learning. Also, yes im more than happy with the power consumption i dont mind if its being ran hard it maxes out, but was hoping that when ive finished using it and the 24/7 vms are just running its drops back down...thats it...nothing more than that really.
 
If you want help, you need to answer the question(s)... tschmidt asked you multiple times for the scaling_driver: [1]...

To make it short:
  • Check the scaling_driver. On your i9-11900 it should be intel_pstate.
  • Check the scaling_available_governors. With intel_pstate there should only be performance and powersave. In that case, powersave should provide dynamic frequency scaling.
    If there are more than those two governors available, most likely another scaling_driver is in use and most likely powersave will pin the frequency to the minimum. In that case, you can use e.g. the schedutil governor: [2] (if listed as available, of course).
Commands are here:
Neobin

[TUTORIAL] Post in thread 'Fix always high CPU frequency in proxmox host.'

I wanted an overview; at least for myself. :cool:


Check currently used driver:
Bash: 
cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_driver

Check currently used governor:
Bash: 
cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

Check current clock speeds (once):
Bash: 
cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_cur_freq

Check current clock speeds (continuous):
Bash: 
watch -n 1 cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_cur_freq

Check available governors:
Bash: 
cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_available_governors

Set governor, for example...
  • Like

[1] https://wiki.archlinux.org/title/CPU_frequency_scaling#Scaling_drivers
[2] https://wiki.archlinux.org/title/CPU_frequency_scaling#Scaling_governors
 
You must log in or register to reply here.
Top Bottom
Back