1427933 - Edge has significantly better battery life on video playback

Status: NEW

(Core :: Audio/Video: Playback, enhancement, P2)

Description

[Jeff Muizelaar [:jrmuizel]]

"Microsoft Edge lasted 79% longer than Firefox, 29% longer than Chrome, and 40% longer than Opera in this test."

https://github.com/MicrosoftEdge/videotest/tree/master/2017-12

Comment 1

[Jeff Muizelaar [:jrmuizel]]

See also https://www.youtube.com/watch?v=9NC_4m1oMIY

Comment 2

[Blake Wu [:bwu][:blakewu]]

We tested Firefox 58 local playbback before. The gap is not so huge. We need to test a real site.

Comment 3

[Jeff Muizelaar [:jrmuizel]]

I'm confused. Are you saying that with the test here https://github.com/MicrosoftEdge/videotest/tree/master/2017-12 we don't see as a big a difference as Microsoft does?

Comment 4

[Blake Wu [:bwu][:blakewu]]

We only tested local playback (play a local file) on 58 and we didn't see a big difference compared Chrome. In Microsoft test case, they tested video streaming which we have a plan to test it as well.

Comment 5

[Tzuhao Kuo [:kaku]]

Microsoft Edge team's experiment was conducted by playing a Netflix video (as described on their page). So, there are three components that are related, networking, media playback, and graphics rendering.

Last year, we also found that playing video is quite power-consuming on Macbook, and we had bug 1400787 to track this issue. In bug 1400787, we conducted a series of experiments, and we were almost sure that it's not media playback, but the rendering pipeline which consumes power most. 

We will then conduct the same experiment in bug 1400787 again on Windows again to see which component is most worthy be optimized wrt power usage.

Comment 6

[Tzuhao Kuo [:kaku]]

From Microsoft's GitHub materials (see videotest/2017-12/index.html), it's worth to be noted that Firefox's GPU power usage is around 10X greater than Chrome and Edge. 

This is similar to the findings on bug 1400787.

Comment 7

[Jeff Muizelaar [:jrmuizel]]

(In reply to Tzuhao Kuo [:kaku] from comment #5)
> Microsoft Edge team's experiment was conducted by playing a Netflix video
> (as described on their page). So, there are three components that are
> related, networking, media playback, and graphics rendering.

The more recent experiments do not use Netflix (so that should avoid any DRM releated things)

Comment 8

[Benyamin Limanto]

Any update on this, it's really frustating tbh. Firefox need to outbeat the browser in the market, regain back the market that it's lost. Now many company doesn't want get their hand wet with FF... They will always force people use edge or chrome instead when there're error..

Comment 9

[Jeff Muizelaar [:jrmuizel]]

Video playback power usage should be improved now that we have DirectComposition support? Benyamin, what is the scenario and power numbers where you see FF worse than Edge?

Comment 10

[Benyamin Limanto]

It indeed better on Windows(last time could be 2 hours differences), but in the long run still can't outbeat CREdge, I don't know why. My Machine using i7-2640m, and have 12gb of ram. It has 1 hours differences(if streaming youtube), both use youtube CPU tamer, both enabled AVC/H264 as VP9 is bad for battery life, but that 1 hours differences, still quite huge.

Any ways to debug it? Because I want to keep FF, but this keep me fallback to edge for this problem. Thanks.

Comment 11

[Jeff Muizelaar [:jrmuizel]]

Benyamin, can you attach the graphics section of your about:support?

Comment 12

[Benyamin Limanto]

Here you go https://gist.github.com/benyaminl/55042ed8a2f1155a9428e83aaa28814d

Comment 13

[Jeff Muizelaar [:jrmuizel]]

Benyamin, can you install Intel Power Gadget and share the reported wattage in both scenarios?

Comment 14

[Benyamin Limanto]

Attachment: PwrData_2021-5-10-18-29-41 - CREdge.csv

Hello, This's CR Edge data, lower about half the firefox (wow, I never expect it the different it's that far, does power gadget broken?) 7W vs 12W average, it's too far :/

i7-2640m, Thinkpad X220, 12gb ram
Video tested
https://www.youtube.com/watch?v=8EvZI46cHsQ

All using H264 avc1(no vp9, when using vp9, the spike on CPU too high tbh)

tested with same addons enabled on both browser (so the result isn't too bias)

1. uBo
2. Youtube enhancement

Comment 15

[Benyamin Limanto]

Attachment: PwrData_2021-5-10-18-33-1 - Firefox 88.0.1.csv

This's Firefox Data. It's drawing power higher than CREdge.

The step of the testing :

1. Make sure only one of the browser run when test, no other apps is running (only one drive)
2. Start the log
3. Run https://www.youtube.com/watch?v=8EvZI46cHsQ video (I only click random video, I don't know other video, as only for testing, I only see the wattage the pwr0 shown)
4. Stop the result when the video done running

I never think this would be this far, youtube is really crazy... -_- (No wonder Sportify use gecko engine to run their proprietary music player, not geckoview, hem... I really wanted to know what secret recipe are... why it's so far...)

Comment 16

[Benyamin Limanto]

Anyway I compare it more with some cases, so I found something interesting tbh.

1. Firefox does better in battery in long run, if, if I browser static pages, or at least server side rendered page, with small js on user side (no cpu spike, whatsoever). BUT in term of rendering video, it does take more battery life than CREdge or other CR based browser
2. If I use CREdge to cassually browse same type of website, static website, news, or other, same as on firefox, for light browsing, seems webrender does won in this race in term of powerdraw (I seen it via power gadget, thank you for introducing me to this software, it's better than HWINFO to show/record the data). The reason is, when I type, or scroll, for some unkown reason, CR based browser spike the CPU till 99% of the CPU Clock, where firefox doesn't, it seems because the page rendered is rendered on the GPU, so it doesn not affect much cpu usage, so daily browsing, firefox won.
3. But when heavy webpage that's consist of video rendering, seems it's still un-optimized not like on chrome/CRedge, so this's a opportunities for FF to fill the gap (I think most of people in this pandemic, are, in fact seeing a lot of video, IF, IF firefox COULD outbeat CREdge, or Chromium based, then it will in theory start regaining market, it does take time, as video rendening, encoding, and decoding isn't as simple as anyone says so, I just hope FF COULD fill that GAP!)
4. WebRender proven to be a great things! I think this need to be exposed to the media (as anyone says, propaganda is the way to move product forward, so as for FF, and Moz, REGAIN THE MARKETSHARE BACK!). Keep it up! It's great! (Anyway it's great if the laptop is having iGPU, I've other laptop that doesn't have iGPU, and the batter life really bad with FF, so yeah, NVIDIA Sucks... a lot.... it's legacy asus laptop anyway)