F5F Stay Refreshed Hardware Desktop Power usage during standby differs between Linux and Windows systems; the variation isn't clear.

Power usage during standby differs between Linux and Windows systems; the variation isn't clear.

Power usage during standby differs between Linux and Windows systems; the variation isn't clear.

B
black_energy_
Member
137
08-18-2018, 04:41 AM
#1
I was evaluating some new components I purchased (an HP ProDesk 600 G4 Mini PC equipped with an i5 8500T, 16 GB RAM, and a 256 GB Micron SATA SSD) and noticed the system idles significantly lower in Windows 10 21H2 compared to Xubuntu 24.04. After a recent installation, letting it rest for 20 to 30 minutes, I observed that Windows consumed between 3.7 to 6.9 watts during idle, whereas Xubuntu used 7.8 to 11.8 watts. This measurement was taken with a Kill A Watt P3 P4400. On Xubuntu, adjusting CPU frequencies to the minimum of 800 MHz didn’t reduce the idle power much. I ran `sudo cpufreq-set -c N -u 800MHz` for all six cores, varying N from 0 to 5. Windows operated on the balanced power plan, and its minimum processor state was 0%. It’s puzzling that Linux tends to draw more power at idle than Windows. Could you suggest any areas to investigate further?
B
black_energy_
08-18-2018, 04:41 AM #1

I was evaluating some new components I purchased (an HP ProDesk 600 G4 Mini PC equipped with an i5 8500T, 16 GB RAM, and a 256 GB Micron SATA SSD) and noticed the system idles significantly lower in Windows 10 21H2 compared to Xubuntu 24.04. After a recent installation, letting it rest for 20 to 30 minutes, I observed that Windows consumed between 3.7 to 6.9 watts during idle, whereas Xubuntu used 7.8 to 11.8 watts. This measurement was taken with a Kill A Watt P3 P4400. On Xubuntu, adjusting CPU frequencies to the minimum of 800 MHz didn’t reduce the idle power much. I ran `sudo cpufreq-set -c N -u 800MHz` for all six cores, varying N from 0 to 5. Windows operated on the balanced power plan, and its minimum processor state was 0%. It’s puzzling that Linux tends to draw more power at idle than Windows. Could you suggest any areas to investigate further?

T
207
08-18-2018, 04:41 AM
#2
Check if Xubuntu runs many processes at once. Look at Task Manager or htop to see what's active. If Windows shares similar GPU usage, it might not be a big concern. It could just mean you're using more resources when idle in Xubuntu.
T
ToastaStroodel
08-18-2018, 04:41 AM #2

Check if Xubuntu runs many processes at once. Look at Task Manager or htop to see what's active. If Windows shares similar GPU usage, it might not be a big concern. It could just mean you're using more resources when idle in Xubuntu.

B
beth04
Member
156
08-18-2018, 04:41 AM
#3
It might be related to driver or BIOS support for ASPM and LAN. Are the C-states matching?
B
beth04
08-18-2018, 04:41 AM #3

It might be related to driver or BIOS support for ASPM and LAN. Are the C-states matching?

P
pedro_tkf
Senior Member
643
08-18-2018, 04:41 AM
#4
You can verify by following the steps provided or consulting a reliable source.
P
pedro_tkf
08-18-2018, 04:41 AM #4

You can verify by following the steps provided or consulting a reliable source.

F
Fordtuff18
Member
141
08-18-2018, 04:41 AM
#5
Reflects the Linux community's focus on collaboration and shared goals, prioritizing energy efficiency when needed.
F
Fordtuff18
08-18-2018, 04:41 AM #5

Reflects the Linux community's focus on collaboration and shared goals, prioritizing energy efficiency when needed.

M
mentalminion
Member
57
08-18-2018, 04:41 AM
#6
Variations in idle power usage usually stem from two OSs failing to utilize identical low-power modes. A poorly crafted driver in either system can block the CPU from accessing certain C states. On Windows, ThrottleStop helps identify which C states are active. HWiNFO provides similar data. I recall a Linux utility that displays C state occupancy percentages as well.
M
mentalminion
08-18-2018, 04:41 AM #6

Variations in idle power usage usually stem from two OSs failing to utilize identical low-power modes. A poorly crafted driver in either system can block the CPU from accessing certain C states. On Windows, ThrottleStop helps identify which C states are active. HWiNFO provides similar data. I recall a Linux utility that displays C state occupancy percentages as well.

S
Santonite
Member
208
08-18-2018, 04:41 AM
#7
Power usage details for an OpenBSD desktop with four fans, including two with LED lights. Overall setup draws 49.5 watts. Main parts include Intel 12700KF, G.SKILL Ripjaws, Sapphire RX 7600, ASRock B760M, and several fans. The system consumes around 10.55 kW when powered on. Display power is about 7 watts. Total consumption is tracked via a meter. I plan to compare Windows 11, FreeBSD, and Linux on this machine later to see which uses less energy. Factors like desktop environment, background processes, and Linux distribution tuning will affect results. Some distros may use significantly less power than others.
S
Santonite
08-18-2018, 04:41 AM #7

Power usage details for an OpenBSD desktop with four fans, including two with LED lights. Overall setup draws 49.5 watts. Main parts include Intel 12700KF, G.SKILL Ripjaws, Sapphire RX 7600, ASRock B760M, and several fans. The system consumes around 10.55 kW when powered on. Display power is about 7 watts. Total consumption is tracked via a meter. I plan to compare Windows 11, FreeBSD, and Linux on this machine later to see which uses less energy. Factors like desktop environment, background processes, and Linux distribution tuning will affect results. Some distros may use significantly less power than others.