A Look At The Windows 10 vs. Linux Power Consumption On A Dell XPS 13 Laptop

[FrankL]

From the diagram it looks like the measurement sampling rate is too low to make a meaningful comparison

in fact, I'd suggest that measuring power consumption is not very interesting, as it doesn't provide the full picture.
A meaningful comparison should measure energy consumption used to perform a set number of tasks.

e.g. If a CPU needs to go into a higher power state to finish the task quicker but use less energy as a result (i.e. more efficient), then this will be reflected in a lower total energy consumption.
When only looking at power consumption, a CPU that spends more time in slower states will always look better, regardless of whether it's more efficient to perform the task at hand.

[mlau]

Interesting. To be honest I expected Win10 to perform worse.

I expected it to do much better, i.e. use less power than linux distros.

[heliosh]

A meaningful comparison should measure energy consumption used to perform a set number of tasks.

Yes, but idle power consumption should also be taken into account, since this is the state in which desktops/laptops are most of the time.
It arguably has the biggest influence on battery runtime of a typical laptop.

[Hi-Angel]

I expected it to do much better, i.e. use less power than linux distros.

Why? Linux kernel gets contributions from a lot more companies than Windows. Then, Linux GPU drivers are very good too. Now, if you remember that a lot of stuff in Windows is using .net, whilst in GNU/Linux usually C/C++ — it's rational to assume that Windows should perform worse.

[Tomin]

Why? Linux kernel gets contributions from a lot more companies than Windows. Then, Linux GPU drivers are very good too. Now, if you remember that a lot of stuff in Windows is using .net, whilst in GNU/Linux usually C/C++ — it's rational to assume that Windows should perform worse.

And Linux software uses Python ja JavaScript.

It's nice that Linux has got this far. I think mostly the missing or disabled power saving features have hindered battery life. A comparison of battery life in realistic scenarios would be a better measure. You rarely need to compare laptop power use when plugged in.

[JeansenVaars]

Windows 10 is still winning, minimum values are smallest there (probably on IDLE), and maximum values are still the minimum of all. This means the average may count because larger spikes on Windows 10?

Something I noticed when running on battery in Windows, is that fans spin off much more frequently than Linux. I wonder if that is because Linux throttles down CPU frequencies much harder than Windows.

[MaCroX95]

Why? Linux kernel gets contributions from a lot more companies than Windows. Then, Linux GPU drivers are very good too. Now, if you remember that a lot of stuff in Windows is using .net, whilst in GNU/Linux usually C/C++ — it's rational to assume that Windows should perform worse.

Well when comparing battery life between laptops on Windows and Linux my experience is still that Windows has at least 30-40% battery life, ofc most of that goes to all the power saving options that Windows has that mainstream distros don't bother with since they aren't really counting on being used for main laptop OSs. Would like to see some big improvements there though...

[ramnasko]

Michael I'm a bit curios about the min values, which all happen in the boot process. Which seams fine, but only Windows got a real advantage here. Why is this? Does this might happen because Windows was tested first and it was the only real cold boot? If so can you spin up the machine before the first test so all test get a warm cpu to start with? Otherwise it would be interesting how windows achieve nearly a third of power usage during boot.

[Michael]

Michael I'm a bit curios about the min values, which all happen in the boot process. Which seams fine, but only Windows got a real advantage here. Why is this? Does this might happen because Windows was tested first and it was the only real cold boot? If so can you spin up the machine before the first test so all test get a warm cpu to start with? Otherwise it would be interesting how windows achieve nearly a third of power usage during boot.

They were all tested from a cold boot / hard off before hand. When in the cold off position, Windows seems to only pull ~1 Watt from the full AC adapter rather than ~2 Watts on the fully charged battery when the other Linux distributions were powered off. But as for why, I don't have a solid answer, presumably Windows doing something different with its UEFI shutdown/configuration or so?

[Veerappan]

Michael I'm a bit curios about the min values, which all happen in the boot process. Which seams fine, but only Windows got a real advantage here. Why is this? Does this might happen because Windows was tested first and it was the only real cold boot? If so can you spin up the machine before the first test so all test get a warm cpu to start with? Otherwise it would be interesting how windows achieve nearly a third of power usage during boot.

Is it possible that the boot process in Linux is just starting more stuff in parallel and therefore has a higher minimum power draw while booting (because it's trying to get to the login screen faster)?

Could also just be that the kernel decides to enable power-saving features later on than windows does, leading to higher draw during boot.