Open Bug 1771637 Opened 2 years ago Updated 16 days ago

video blinking when play two video in two window on two screen with vastly different refresh rate (60Hz and 160Hz) at same time on Radeon RX 6900 XT on Adrenalin 22.5.2 Optional

Categories

(Core :: Graphics: WebRender, defect)

Firefox 102
Desktop
Windows 11
defect

Tracking

()

UNCONFIRMED

People

(Reporter: mmis1000, Unassigned)

References

(Blocks 2 open bugs)

Details

(Keywords: regression)

+++ This bug was initially created as a clone of Bug #1759328 +++

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:102.0) Gecko/20100101 Firefox/102.0

Steps to reproduce:

  1. plug two monitor with vastly different refresh rate (160hz, 60hz)
  2. open two Firefox window, one on each monitor.
  3. play some youtube video on both window (resolution or bit-rate does not seem to be matter, reproducible even with 144p video)
  4. show performance overlay with right click

Actual results:

The video on monitor with lower refresh rate (60hz) blinks in very high frequency.

Expected results:

Both video plays normally.

Note:

If you change the 160hz screen to 60, it works normally.
If you pause one of the videos, it works normally.
It breaks only if you plays both.
If you disable gfx.webrender.dcomp-video-overlay-win, it plays normally.

Note 2:

This symptom of this bug is basically the same as Bug #1759328,
but it have some difference

  1. happens on different driver version (Adrenalin 22.5.2 Optional)
  2. is far more reproducible (happen even without high gpu load).
  3. does not affect things except overlay on video

So it is opened as a new issue instead.

See Also: → 1759328

:mmis1000, since this bug is a regression, could you fill (if possible) the regressed_by field?
For more information, please visit auto_nag documentation.

Flags: needinfo?(mmis1000)

(In reply to Release mgmt bot [:suhaib / :marco/ :calixte] from comment #1)

:mmis1000, since this bug is a regression, could you fill (if possible) the regressed_by field?
For more information, please visit auto_nag documentation.

I think this is more a bug about the amd driver?(Even windows game bar can't capture the glitch, which is really weird) Firefox just triggered it by accident by using the overlay. About the regression range. I think it is regressed by the commit of 'gfx.webrender.dcomp-video-overlay-win' (see previous issue for detail). Should I try to force enable this in even earlier version to see what feature actually breaks it?

A further test make me realize, it's not the firefox dropping frame. It's the whole monitor skipping frames.

I made a vsync tester page that swaps blue/red rectangle in speed of 1 time/frame.

https://codepen.io/mmis1000/pen/qBxqgLr

The color looks like purple most of time when you put it in the high refresh rate main monitor in normal condition.
But as soon as I am trying to reproduce the issue above.
The monitor is no longer v-synced at all. frames are skipping. And the color in the v-sync test page no longer looks like purple.
And this happens even you put the v-sync test page in another browser.
Which indicates this is a system-wide problem that triggered accidentally by Firefox.

The radeon 22.11.2 has a patch note about

  • During video playback and window switching, an intermittent driver timeout or black screen may occur on Radeon™ RX 6000 series GPUs using some 240Hz refresh rate displays or high refresh rate primary display plus low refresh rate secondary display configurations.

But it appears not fixing this issue. The symptom above is reproducible in my end. So I guess it fixed something else.

Flags: needinfo?(mmis1000)
You need to log in before you can comment on or make changes to this bug.