1530083 - vp9 hardware acceleration broken

Status: UNCONFIRMED

(Core :: Audio/Video: Playback, defect)

Description

[Tuti Tubunge]

Attachment: vp9.png

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:67.0) Gecko/20100101 Firefox/67.0

Actual results:

Windows 10 1809, RTX2070 414.19, FF67.0a1
Vp9 Hardware acceleration does not work on YouTube.

Comment 1

[Jean-Yves Avenard [:jya]]

(In reply to Tuti Tubunge from comment #0)

Created attachment 9046105 [details]
vp9.png

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:67.0) Gecko/20100101 Firefox/67.0

Actual results:

Windows 10 1809, RTX2070 414.19, FF67.0a1
Vp9 Hardware acceleration does not work on YouTube.

Please attach your about:support output.

Comment 2

[Jean-Yves Avenard [:jya]]

Also, please install the media devtools from https://addons.mozilla.org/en-US/firefox/addon/devtools-media-panel/

when playing the vp9 video, press Ctrl-Shift-I go to the Media-Webrtc tab and select the URL to the video element playing. Expand it and copy/paste here the result.

Thank you

Comment 3

[Jean-Yves Avenard [:jya]]

Are you still experiencing the issue with a current nightly?

Comment 4

[Anton]

I have this problem.
Firefox 69
Windows 10 LTSC

PC №1 - GPU Vega 11
Firefox not use HW Decoder in VP9: https://i.ibb.co/4sxYmRp/1.jpg
Vega 11 Decoder support VP9 (DXVA Checker): https://ibb.co/rZBY3rY
MPC-BE + VP9 = OK: https://ibb.co/m89VJBf

PC №2 - RTX2080
VP9 in Firefox not use NVENC: https://ibb.co/d2D9zcf][img]https://i.ibb.co/J7HLMdc/4.jpg
NVENC Turing have support VP9@8K: https://i.ibb.co/bsjPnqc/5.jpg

about:support - https://pastebin.com/U1FVnvnt

media.mediasource.webm.enabled = true
media.webm.enabled = true

Comment 5

[Asif Youssuff]

Attachment: media-webrtc-profossi

Another user reported this issue here: https://www.reddit.com/r/firefox/comments/hiwt69/vp9_encoded_videos_eg_youtube_stutter_firefox/

See also attached file about-support-profossi

>I'm getting really annoying stutter while playing videos with resolutions at or above 1080p, for example https://www.youtube.com/watch?v=5uIcHZY5B20

>Disabling hardware acceleration fixes the problem, even at 8k (but is not exactly an elegant solution). Using Edge fixes the problem. Has anyone else had such issues recently? I've only seen similar complaints from 2019. Do you know of better workarounds?

>I'm running Windows 10.0.18362 on a relatively fast PC - Ryzen 7 3700X, Radeon RX 5700XT with latest drivers, a single 144Hz 1440p monitor.

Comment 6

[Asif Youssuff]

Attachment: about-support-profossi

Comment 7

[Asif Youssuff]

Jean-Yves, need-infoing you since we now have a log from media-webrtc. See comments 5 and 6.

Comment 8

[Jean-Yves Avenard [:jya]]

forwarding the ni?

The original bug is probably the same as bug 1634991, probably a dup.

The other newly provided info is I believe unrelated and should have its own bug

Comment 9

[Bryce Seager van Dyk [:bryce] (he/him) - Not reading bugmail]

We've added markers to the Firefox profiler to help detect failures in Window's decoding pipeline. If you're encountering this issue it would be helpful if you could gather a profile using the 'Media' config on Firefox Nightly from after 2020-10-20, then share it and paste the link here.

Comment 10

[Bryce Seager van Dyk [:bryce] (he/him) - Not reading bugmail]

Removing NIs. If we get more info we may be able to action this. We have further markers and have done some reworking of the decoding architecture, so it would be useful to have more profiles (or to know that this is not longer causing problems). Another consideration is that since we rely on platforms implementations of hardware decoders, it's possible for this to also change if operating systems rework their decoders.

Comment 11

[Asif Youssuff]

Profile from: https://www.reddit.com/r/firefox/comments/mbu0gw/vp9_hardware_decode/

https://share.firefox.dev/3d0mMU2

i7-8750H, Intel UHD 630, FF 87

Comment 12

[BugBot [:suhaib / :marco/ :calixte]]

Clear a needinfo that is pending on an inactive user.

Inactive users most likely will not respond; if the missing information is essential and cannot be collected another way, the bug maybe should be closed as INCOMPLETE.

For more information, please visit auto_nag documentation.

Comment 13

[Artem S. Tashkinov]

I also experience the same problem under 1809, see bug 1813339