1738851 - Flickering while scrolling on High-Resolution display on Windows

Status: RESOLVED INCOMPLETE

(Core :: Graphics: WebRender, defect)

Description

[Jens Stutte [:jstutte]]

Attachment: FF96config.txt

On Windows 10, Lenovo X1 Carbon with external display Dell 27" 3840x2160 resolution scrolling and mouse movement causes BitBlock flickering at unexpected positions. End-Result once stable is correct.

See the config for details. I have a screen recording of this, too.

Comment 1

[Jens Stutte [:jstutte]]

With Google docs, frequent flickering is constant even without user interaction. On the same laptop on the built-in screen everything works as expected (even while the other monitor is active).

Comment 2

[Jens Stutte [:jstutte]]

It happens also with the FF93 release version.

Comment 3

[Jens Stutte [:jstutte]]

Some more data point: I tried to capture a profile, but if I start profiling it does not happen. What's more, even after stopping the recording it does not happen. Once I restart the browser it's there again. So my workaround for now is to start & stop profiling recording after browser start.

Comment 4

[Jim Mathies [:jimm]]

A screen recording (even from a second device like your phone) would be helpful.

Comment 5

[Jens Stutte [:jstutte]]

Here: https://youtu.be/1kl6CDHduX0 (HD still processing). It's quite visible on the second site in the right ad.

Comment 6

[Jens Stutte [:jstutte]]

I am now at home with a different setup which works like a charme ("small" full HD external monitor, otherwise identical) , further tests with the affected HW are possible next tuesday (for me).

Comment 7

[Timothy Nikkel (:tnikkel)]

Perhaps a regression range would be useful?

Comment 8

[Jens Stutte [:jstutte]]

Attachment: flicker_changeset

Changeset that probably contains the regressing bug.

Comment 9

[Jens Stutte [:jstutte]]

Using mozregression (no profile, no pref changes), the last central build that does work well here is 20200219095352. The build 20200220093508 clearly shows the same problem.

Comment 10

[Jens Stutte [:jstutte]]

What stands out in the changeset is:

changeset:   514593:66c1130c7a8b
user:        Andrew Osmond <[e-mail]>
date:        Tue Feb 18 13:56:13 2020 +0000
summary:     Bug 1614102 - Allow WebRender on all Intel Coffee Lake GT2+ GPUs. r=jrmuize
...
changeset:   514638:6745db621915
user:        Andrew Osmond <[e-mail]>
date:        Wed Feb 19 19:05:58 2020 +0000
summary:     Bug 1615015 - Ship WebRender on nightly to all explicitly allowed NVIDIA/AMD/Intel, regardless of battery/screen. r=jrmuizel

I have here:

Manufacturer		GenuineIntel
Name			Intel Core i7 8550U
Codename		Kaby Lake-R
Specification		Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
Package (platform ID)	Socket 1356 FCBGA (0x7)
CPUID			6.E.A
Extended CPUID		6.8E
Core Stepping		Y0
...
Northbridge		Intel Kaby Lake rev. 08
Southbridge		Intel Coffee Lake-U/Y PCH rev. 21
Bus Specification	PCI-Express 3.0 (8.0 GT/s)
Graphic Interface	PCI-Express

My best guess would be then bug 1614102.

Comment 11

[Jens Stutte [:jstutte]]

(In reply to Jens Stutte [:jstutte] from comment #3)

Some more data point: I tried to capture a profile, but if I start profiling it does not happen. What's more, even after stopping the recording it does not happen. Once I restart the browser it's there again. So my workaround for now is to start & stop profiling recording after browser start.

Please let me know if I can do something that documents the change here triggered by profile start&stop (if it is not just known to be webrender=off).

Comment 12

[Jens Stutte [:jstutte]]

I happened to forget my laptop at home so I picked a loaner, always Lenovo X1 Carbon but slightly older:

Manufacturer		GenuineIntel
Name			Intel Core i7 7600U
Codename		Kaby Lake-R
Specification		Intel(R) Core(TM) i7-7600U CPU @ 2.80GHz
Package (platform ID)	Socket 1515 FCBGA (0x7)
CPUID			6.E.9
Extended CPUID		6.8E
Core Stepping		B0
...
Northbridge		Intel Kaby Lake rev. 02
Southbridge		Intel Coffee Lake-U/Y PCH rev. 21
Bus Specification	PCI-Express 3.0 (8.0 GT/s)
Graphic Interface	PCI-Express

With this laptop it works without problems, though gfx.webrender.all was set to false out of the box, but switching it on does not trigger the problem on this machine, as well.

Comment 13

[Jim Mathies [:jimm]]

"adapterDescription": "Intel(R) UHD Graphics 620",
"adapterVendorID": "0x8086",
"adapterDeviceID": "0x5917",
"adapterSubsysID": "225c17aa",
"adapterRAM": 0,
"adapterDrivers": "igdumdim64 igd10iumd64 igd10iumd64 igd12umd64 igdumdim32 igd10iumd32 igd10iumd32 igd12umd32",
"driverVendor": "",
"driverVersion": "27.20.100.9749",
"driverDate": "6-24-2021",

Comment 14

[Darkspirit]

You have two displays with different refresh rates. Can the bug be fixed by setting gfx.webrender.compositor to false and restarting Firefox?

Comment 15

[Jens Stutte [:jstutte]]

(In reply to Darkspirit from comment #14)

You have two displays with different refresh rates. Can the bug be fixed by setting gfx.webrender.compositor to false and restarting Firefox?

Yes, setting gfx.webrender.compositor to false seems to help. And yes, I have different refresh rates here. The internal monitor cannot be changed from 60Hz, for the external I have only the choice between 29 and 30 Hz, with both settings I see the same flickering.

Comment 16

[Bob Hood [:bhood]]

Jens, it has been quite some time since this was reported. Can you test with the current release and let us know whether or not this behavior remains?

Comment 17

[Jens Stutte [:jstutte]]

Unfortunately I do not have access to the exact setup anymore - the docking stations in the office have been replaced (and I got a new laptop, though I have kept the old one for private use). I have not seen this with other setups, so for my personal use it works now (but that might not tell much about "is this bug fixed", of course). If nobody else experiences this, I think it is fine to be closed now.

Comment 18

[Bob Hood [:bhood]]

Thanks for getting back, Jens. I'll flag it as INCOMPLETE then since we didn't actually find a solution.