Measuring power consumption of Lenovo Mini PCs was on my bingo card. Finally got the time and motivation, all thanks to “I should probably update to Proxmox 9 atp”.
A bit of context, I run my homelab on a Lenovo M720q with i5-8500T and wanted to install some app, can’t exactly remember what. When I ran the helper script, it failed with an error indicating i have to upgrade to PVE9. So, all this led to me upgrade to PVE9 and measuring the power consumption in the process. Basically, I recorded the power consumption over 5min for each test case (row in the table) and took the average.
Initial Setup:
- Lenovo M720q Tiny
- Crucial BX500 500GB SATA SSD
- 32GB (2x16GB) 3200mhz DDR4 RAM
- Tp-Link Tapo P110 Smart Plug
- 1G Uplink Ethernet
- PVE 9.0.3
For initial testing, I used of combination of default bios settings, tweaked settings, with or without display attached.
| BIOS | Display | Ethernet | Idle Power (w) | Comments |
|---|---|---|---|---|
| Stock/Default | y | n | 8w | |
| Stock/Default | n | y | 6w | Jumps to 7w constantly |
| Tweaked | y | n | 8w | |
| Tweaked | n | y | 6w |
Note:
-
This is just proxmox, no lxc/apps are installed.
-
Following are the bios settings i used
Default/Stock - OS optimized defaults - y - Load default settings - y With Tweaks - Devices > USB Setup > Rear USB Ports - disabled - Devices > Audio Setup > Onboard Audio Controller - disabled - Devices > Network Setup > PXE * - disabled - Advanced > CPU Setup > Turbo Mode - disabled - Power > Enhanced Power Saving Mode - disabled - Power > Smart Power On - disabled - Security > Require POP on System boot - no
Changing BIOS settings didn’t helped much as you can see, I guess it doesn’t affect that much until unless those features/ports are actively being used.
Now i wanted to test the power consumption with few LXCs/VMs. Unfortunately, i got hit with linux file permission shitshow when i was trying to use PBS to restore my services, long story short, i was able to fix it and restore all my stuff which i thought i lost, but atp the setup was no longer ideal for testing. So, reinstalled PVE which got updated to 9.1.
I then changed couple of things from last time:
Crucial BX500 500GB SATA SSDCrucial P3 PCIE3.0 500GBPVE 9.0.3PVE 9.1
Bonus: I also added i3-8100 (65w) to the mix.
Here are the test results:
| Particular | i5-8500T (35w) | i3-8100 (65w) | ||||
|---|---|---|---|---|---|---|
| Idle (Average) | Lowest | Peak | Idle (Average) | Lowest | Peak | |
| w/ just display connected | 5w | 4w | 6w | 6w | 5w | 6w |
| w/o display | 4w | 4w | 6w | 5w | 4w | 7w |
| w/ ethernet connected (no display) | 5w | 5w | 6w | 4w | 3w | 5w |
| w/ Powersave CPU Profile | 4w | 4w | 5w | 4w | 3w | 6w |
| Services + Windows Video Playback | 15w | 13w | 17w | 14w | 13w | 30w |
| Services (No Video Playback) | 8w | 7w | 10w | 8w | 7w | 11w |
| Services (Except Windows) | 5w | 5w | 8w | 5w | 5w | 8w |
Notes:
- Every test case inherits the previous test case setup.
- Services i installed are caddy, adguard, vaultwarden, cloudflared, homeassistant, Windows 11 VM (debloated)
- I re-installed Proxmox and all the services (from backup) when the CPU was changed from i5-8500T to i3-8100
- Windows is mad unreliable with mad power spikes. Don’t why there were major spikes with i3-8100. Cause of high CPU TDP? Windows updating in background? who knows.
Takeaways:
- PVE9 is definitely superior in power consumption compared to PVE8, thanks to debian 13. Right now, my fully setup homelab with 1x PVE node and NAS machine (HP SFF with 2x 4TB) taking b/w 35-40w, which was 40-45w with PVE8.
- NVME is better than SATA SSD.
- Don’t attach or keep attached any display.
These results are just my personal testing, feel free challenge or share your experience as well.
Cheers.
to the effort put in! 
.