GNOME's Wayland Session Shows Potential For Better Battery Life Than With X.Org

[wooque]

Benchmark tells to me that Xorg will have better battery life.
I see much lower minimal consumption in Xorg (8.8W Wayland vs 4.8W Xorg), which means system under Xorg achieves lower power states and does more aggressive power savings, which will more realistically translate to real world usage as people usually don't run full load benchmarks on battery but do much lighter tasks, such as web browsing where there are lot of idle time.

[ddriver]

Benchmark tells to me that Xorg will have better battery life.
I see much lower minimal consumption in Xorg (8.8W Wayland vs 4.8W Xorg), which means system under Xorg achieves lower power states and does more aggressive power savings, which will more realistically translate to real world usage as people usually don't run full load benchmarks on battery but do much lighter tasks, such as web browsing where there are lot of idle time.

Lol, that's just noise with no statistical significance. Wayland has better power efficiency, further demonstrated by the tangibly lower temps as well.

My guess is wayland has a higher minimum event loop cycle, so it doesn't get as low in the few best case scenarios, but is still nevertheless significantly better when there's actual work to be done.

[oiaohm]

Benchmark tells to me that Xorg will have better battery life.
I see much lower minimal consumption in Xorg (8.8W Wayland vs 4.8W Xorg), which means system under Xorg achieves lower power states and does more aggressive power savings, which will more realistically translate to real world usage as people usually don't run full load benchmarks on battery but do much lighter tasks, such as web browsing where there are lot of idle time.

Really I would say you are attempting to use data past what it can in fact tell you here. Same problem applies here that was benchmarks with quite bit of load all the time so these number don't line up to light task benchmarks. So it was kind of odd that Xorg was dropping down to 4.8 for split seconds. That could be linked to why Xorg was 3 Watt higher. Was X.org stalling for some reason to trigger the 4.8 watts?

High load benchmarks cannot really be used to draw lines that well for light load benchmarks.

The numbers say more investigation required to find out what those 4.8W Xorg are about. Are they are optimisation missing from the Wayland side or are they a side effect of some X11 Xorg issue causing processing stalls.

Remember the average was 3 watts lower. Yes something stalling and needing to be catch up on it self in data processing over and over again can end up being heavier in power usage due to pushing cpu/gpu clocks higher into less power effective speeds. This kind of fun issue gives a nice little false flag of power effectiveness when what you are looking at is a symptom of power inefficiency caused by invalid stalling.

So we need some light load benchmarks like 100 percent idle for 5 mins.

[CochainComplex]

Nice, this one does not escalade. It seems tide has turned. This is not meant in a provocative way - Rather a hint that Wayland might gain broader acceptance in the Linux world.

[Shiba]

Good! You will need that extra battery life when gnome-shell crashes bringing down the entire session and you have to redo all your unsaved work.

[oleid]

Benchmark tells to me that Xorg will have better battery life.
I see much lower minimal consumption in Xorg (8.8W Wayland vs 4.8W Xorg), which means system under Xorg achieves lower power states and does more aggressive power savings, which will more realistically translate to real world usage as people usually don't run full load benchmarks on battery but do much lighter tasks, such as web browsing where there are lot of idle time.

The single data points are not that important, the integral over time is.

[Volta]

On native Wayland session it should be even better.

[wooque]

Also hardware accelerated video decode does not work under Wayland at the moment in Chrome and Chromium based browsers, so they will have greater power consumption playing videos, which matters a lot on laptops.

[aufkrawall]

Also hardware accelerated video decode does not work under Wayland at the moment in Chrome and Chromium based browsers, so they will have greater power consumption playing videos, which matters a lot on laptops.

Doesn't work on Xorg anymore either.

Of course Wayland will be much more power efficient when you can directly scan out all kinds of single surfaces. This isn't really of interest for any dGPU power consumption, but of course ULP devices. However, there is still a shipton of other performance optimizations not picked up for Gnome. The shell seems to be some kind of CPU cycle eater that kills lowend SoCs.

[wooque]

Doesn't work on Xorg anymore either.

Works on my machine. I'm using Chromium on Arch Linux so they might be doing some special patching.