2054275 - Occasional stutter scrolling since 152.0.5?

Status: UNCONFIRMED

(Core :: Graphics: WebRender, defect, P2)

Description

[:hello]

Attachment: test.txt

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:152.0) Gecko/20100101 Firefox/152.0

Steps to reproduce:

What did you do? (steps to reproduce)

1.Open Firefox 152.0.5 on Windows 10.
2,Visit the following Reddit thread:
https://www.reddit.com/r/uBlockOrigin/comments/1pwha78/element_picker_quick_filter_creation_browsers/
3.Continuously scroll up and down through the page for several seconds,
4.Repeat if necessary. The issue is intermittent but occurs frequently on this page. I have also observed it on other websites. especially around posts containing embedded media like the thread linked

Actual results:

What happened? (actual results)

Scrolling intermittently stutters. The page briefly hitches, jumps, or appears to overshoot while scrolling. The issue seems to occur more frequently on pages containing embedded media (such as videos), like the Reddit thread linked in the reproduction steps, although I have also observed it on other websites that have similar imbedded media (multiple videos playing at the same time)

The issue occurs with both the mouse wheel and Firefox's auto-scroll, so it does not appear to be related to the mouse wheel hardware.

The problem does not occur in Microsoft Edge on the same system.

I have already tried:

Refresh Firefox (new profile)

Reinstall Firefox

Disable/uninstall uBlock Origin

Update/change NVIDIA graphics drivers

None of these changed the behavior.

Expected results:

What should have happened? (expected results)

Scrolling should remain consistently smooth without intermittent hitches, jumps, or overshooting, matching the behavior observed in Microsoft Edge on the same hardware.

I have attached about:support text and if specific logging is required, please let me know which logging preset modules you would like me to use, and I will provide the requested logs.

Comment 1

[BugBot [:suhaib / :marco]]

The Bugbug bot thinks this bug should belong to the 'Core::Graphics' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.

Comment 2

[:hello]

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

Performance profile attached

I used this thread as an example and auto scroll up and down all through out the logging session.
https://www.reddit.com/r/uBlockOrigin/comments/1pwha78/element_picker_quick_filter_creation_browsers/

Comment 3

[Lee Salzman [:lsalzman]]

Glenn, do you see anything in the profile here that looks like it might explain the stutter WebRender-wise? I am not seeing anything in particular that might explain it. The commenter mentions embedded media like video, so perhaps it could be related to variation in frame rendering times from compositing such?

Comment 4

[Glenn Watson [:gw]]

I can see a few long composite times in the render thread, ~50-80ms, which is probably what you're observing, especially when running at 144 Hz. Could you try change gfx.webrender.layer-compositor in about:config to false, restart the browser and see if that makes any difference?

Comment 5

[:hello]

(In reply to Glenn Watson [:gw] from comment #4)

I can see a few long composite times in the render thread, ~50-80ms, which is probably what you're observing, especially when running at 144 Hz. Could you try change gfx.webrender.layer-compositor in about:config to false, restart the browser and see if that makes any difference?

Hello,

I tried setting gfx.webrender.layer-compositor to false in about:config, but it made no difference.

While scrolling, videos autoplay, which uses the video decoder, so I wondered if that might be the issue.

What did make a noticeable difference was setting media.hardware-video-decoding.enabled to false. (I know this basically turns off hardware acceleration for videos but not UI)

Scrolling is now much smoother. It's not 100% smooth, but the improvement is very significant.

I'm not sure what's wrong with hardware acceleration. I've been encountering occasional odd quirks that consistently disappear when I disable hardware acceleration, and they usually seem to involve video playback. There was another issue I didn't mention here because I created a separate thread about it, and disabling hardware acceleration resolved that one as well.

Even tried refreshing the profile just incase of corruption or changing gpu driver.

Comment 6

[Glenn Watson [:gw]]

Adding on a couple of people who know more about hardware video decoding, it does seem specifically related to that.

Comment 7

[:hello]

(In reply to Glenn Watson [:gw] from comment #6)

Adding on a couple of people who know more about hardware video decoding, it does seem specifically related to that.

Here is another profiler log from a site I visit that also exhibits scroll hitching, although not as severely. In this case, the site does not have any media playing in the background.(this site is a manhua site so there's not even media playing in the background)

For this test, I reset gfx.webrender.layer-compositor to its default value to simulate Firefox's default configuration.
https://share.firefox.dev/4pmQAyV

In this case, setting media.hardware-video-decoding.enabled to false only helps slightly.The hitch as I scroll up seems only lessened not completely removed. Scrolling is still not as smooth as in Edge, or as smooth as Firefox was for me 1–2 weeks ago.

For comparison, scrolling on Edge is completely smooth in both directions as of this moment on both sites.

To reproduce the behavior, I pressed the middle mouse button and used auto-scroll by moving the mouse up and down, letting the page scroll on its own. Every time I reached the upper part of the page, I observed a small hitch.

I also notice that the scroll hitching occurs at the same time as the stutter I described in the main report. I suspect the two issues are closely related.

So far:
Issue is not older then 2 weeks
Edge scrolling is smooth as butter using the same sites
Having media playing in the background tends to make the scrolling stutter worst but it can also sometimes happen in sites that don't show media when scrolling.

Comment 8

[:hello]

To further clarify maybe I can help pinpointing the issue.

From what I can remember this issue happened between the current firefox 152.0.6 and 152.0.4 .I don't know if maybe something was sneaked in inbetween these 2 updates.Which add ups to everything being fine around 2 weeks ago scrolling wise.

Comment 9

[Glenn Watson [:gw]]

If you're able to reproduce reliably, running mozregression (https://mozilla.github.io/mozregression/) and reporting the result here would be very helpful - it can often track the regression down to a specific commit.

Comment 10

[:hello]

(In reply to Glenn Watson [:gw] from comment #9)

If you're able to reproduce reliably, running mozregression (https://mozilla.github.io/mozregression/) and reporting the result here would be very helpful - it can often track the regression down to a specific commit.

Question,

I never used mozregression.

If I download it and run it from date A to date B that wont effect my regular firefox install right?don't want to lose booksmarks/history of my current firefox profile.

Comment 11

[:hello]

(In reply to Glenn Watson [:gw] from comment #9)

If you're able to reproduce reliably, running mozregression (https://mozilla.github.io/mozregression/) and reporting the result here would be very helpful - it can often track the regression down to a specific commit.

Ok Glenn,

I managed to figure out how to use

chose regression
left every other settings default
Choose from the 30th of june to now

https://www.reddit.com/r/uBlockOrigin/comments/1pwha78/element_picker_quick_filter_creation_browsers/
Test scroll.Held middle mouse button for auto scroll and went up and down.There was a first stutter with the occasional frame pacing issue.
Felt marginally better then now but Edge is just smooth

2nd test site https://asurascans.com/
Test scroll.Held middle mouse button to auto scroll up.As the scroll goes down it's smooth.As the scroll goes up again near the end there's a minor hitch every time.It can be subtle.
Again,Edge is unnefected.No such stutter with any scrolling movement.

Now I'm a bit confused.I was sure I wasn't effected by this issue 2+ weeks ago.
Question now is why is it happening on firefox and not chrome?
Mozzregression gives you a clean new profile with nothing so its not even a matter of me having a wonky profile or a driver issue or else edge would be effected aswell.

Comment 12

[:hello]

Question now is why is it happening on firefox and not chrome?
Sorry I meant Edge.I used edge to compare.

Comment 13

[Glenn Watson [:gw]]

The output of mozregression, if it works, should give you a link you can paste here with a range of possible changes that could have caused it, which we can then isolate further.

Comment 14

[:hello]

(In reply to Glenn Watson [:gw] from comment #13)

The output of mozregression, if it works, should give you a link you can paste here with a range of possible changes that could have caused it, which we can then isolate further.

Hello Glenn,

I honestly have no idea what to say.

Oddly enough, I'm noticing erratic scrolling performance on the site examples I mentioned, even as far back as June 16, 2026. I'm sure this wasn't an issue that far back. I was convinced it was fine even two weeks ago.

I dunno why, but on Firefox the scrolling performance always feels random. Sometimes sites like the ones I mentioned have issues when media is playing, and sometimes they don't. Even the manhua site can have the issue despite not having any media. At the same time, some sites are completely fine.

For example, Today I opened the problem page:
https://www.reddit.com/r/uBlockOrigin/comments/1pwha78/element_picker_quick_filter_creation_browsers/

...then switch to another tab, switch back, and scroll up and down, it's suddenly fine with no stuttering or hitching. Then, for no apparent reason, the scrolling becomes hitchy again.Rinse and repeat. I can't find any pattern or consistent way to reproduce it.

Either the issue was always there and I only just noticed it, and it's one of those things that once you see it, you can't unsee it, or I've simply run out of ideas.

All I know is that on Edge the scrolling performance is consistently smoother.There are no problem sites that give me erratic scrolling performance. Maybe these sites are just better optimized for Chrome/Edge?May be the act of scrolling itself is somehow more optimized?

Comment 15

[:hello]

The only consistent microstutter I can reproduce on Firefox, but not on Edge (where scrolling is consistently smooth), is when scrolling up on this site: https://asurascans.com/
I can even reproduce it as far back as the June 16 build in mozregression.
Again, I'm out of ideas. Either this behavior was always there and I only recently started noticing it on Firefox, or something else changed around the time I made this thread.

Comment 16

[Alastor Wu [:alwu]]

Hello, could you help to use about:logging to capture a media profile? (instruction here) Thanks!

Comment 17

[:hello]

(In reply to Alastor Wu [:alwu] from comment #16)

Hello, could you help to use about:logging to capture a media profile? (instruction here) Thanks!

https://share.firefox.dev/4f3Yiuk
First log for: https://asurascans.com/

Micro stutter every time I kept scrolling up at a certain point.Scrolling down no issue.This one can be subtle and sometimes not subtle.
Not present on edge at all.It's like the upper part of the site has something that is causing the micro hitch to happen.

https://share.firefox.dev/4ps18wz
Second log for scrolling up and down here:
https://www.reddit.com/r/uBlockOrigin/comments/1pwha78/element_picker_quick_filter_creation_browsers/

This one is an odd one.Sometimes its fine.Then I swap tab back and forth and its lagging scrolling or the other way around.As I load the page I start scrolling up and down and it starts with a hitch and then it smooths out somewhat but there's still framepacing issues when scrolling up and down with the occasional micro hitch.Again,sometimes it just works and is smooth until I look in another tab for a second and go back to the offending tab.
Again,this is not present on edge.

Could it be something up with hardware acceleration?I know hardware acceleration is also used for stuff like scrolling and not just playing media.Maybe it's the difference between how firefox uses hardware acceleration and edge?

Comment 18

[:hello]

https://share.firefox.dev/4ps18wz
Second log for scrolling up and down here:

Ignore this log please.I think I made a mistake and logged networking on this one
Here is the correct log https://share.firefox.dev/3THcAsk logging media.
All I said about this one still stands.Swapping tabs then back to the offending site then scrolling seems to trigger an initial hitch when scrolling and alot of the times the frame pacing is impacted aswell with occasional bouts of smooth scrolling out of nowhere.

Comment 19

[:hello]

(In reply to Alastor Wu [:alwu] from comment #16)

Hello, could you help to use about:logging to capture a media profile? (instruction here) Thanks!

I have provided the logs alastor

Comment 20

[:hello]

(In reply to :hello from comment #19)

(In reply to Alastor Wu [:alwu] from comment #16)

Hello, could you help to use about:logging to capture a media profile? (instruction here) Thanks!

I have provided the logs alastor

TEST
My email got de-activated and i re-activated it

Comment 21

[:hello]

(In reply to Glenn Watson [:gw] from comment #6)

Adding on a couple of people who know more about hardware video decoding, it does seem specifically related to that.

Hello Glenn,

Any possibility to look at the latest logs I posted?The more I experience the issue the more im noticing that if a video or clip is running at the same time it tends to make the scrolling perform worst.There is a clear difference in scrolling performance when there's no media playing playing on the same page im scrolling on.

Thank you.

Comment 22

[:hello]

Here is another media log scrolling through facebook:
https://share.firefox.dev/4g9ATrM
You can kinda tell there's some frame pacing issues when videos are playing as you scroll down.Not super obvious but they are there.

Scrolling through the steam store:
https://share.firefox.dev/3RugHaP
Scrolling through the steam store.Occasional frame pacing issues as you scroll up and down aswell.Again,Not super obvious but it definitely feels like a regression scrolling performance wise vs to how it was before.

Comment 23

[Glenn Watson [:gw]]

I won't have time to look at this any time soon sorry - I'm also no expert on media playback, especially on windows. alwu any ideas?

Comment 24

[Alastor Wu [:alwu]]

Thanks for the profiles. From the profile, the frame drops seems lining up with Firefox creating a DirectComposition video swap chain, which on your setup lands in a slow GPU driver call.

Could you try two runs on whichever site reproduces the stutter most reliably for you? Change one pref at a time in about:config and RESTART after each — they only apply at startup.

Run 1: gfx.webrender.dcomp-video-vp-scaling-win = false
Run 2: (reset run 1 to true) gfx.webrender.dcomp-video-hw-overlay-win = false

Please reset both to true afterwards.

For each run, tell me if scrolling is better / worse / same — "no difference" is just as useful, so please report it either way. It's be perfect if you could also grab a profile per run, use the Graphics preset this time in about:logging. Thanks!

Comment 25

[:hello]

Attachment: Testing big read

Comment 26

[:hello]

(In reply to Alastor Wu [:alwu] from comment #24)

Thanks for the profiles. From the profile, the frame drops seems lining up with Firefox creating a DirectComposition video swap chain, which on your setup lands in a slow GPU driver call.

Could you try two runs on whichever site reproduces the stutter most reliably for you? Change one pref at a time in about:config and RESTART after each — they only apply at startup.

Run 1: gfx.webrender.dcomp-video-vp-scaling-win = false
Run 2: (reset run 1 to true) gfx.webrender.dcomp-video-hw-overlay-win = false

Please reset both to true afterwards.

For each run, tell me if scrolling is better / worse / same — "no difference" is just as useful, so please report it either way. It's be perfect if you could also grab a profile per run, use the Graphics preset this time in about:logging. Thanks!

Sorry about the big wall of text Alwu,I tested both settings on 3 different sites each so its a big read but there you go.

Minor correction on the wall of text.
On test 2 I say:

(this is referring to the reddit thread with 3 videos playing)
"Logging while the 3 videos where paused felt smoother but in this case there was still a subtle hitch from the start and smoothed out.
TEST 2 with videos paused felt smoother"

What I meant to say was TEST 1 with video paused was smoother then TEST 2 with videos paused.

Comment 27

[Ali]

I'm also experiencing this bug.

It can be reproduced while scrolling this page:
https://www.rnz.co.nz/life/wellbeing/nz-researchers-on-cusp-of-new-obesity-treatment

Here is the problematic change according to mozregression:

2026-08-12T01:28:50.058000: INFO : platform_version: 152.0a1
2026-08-12T01:29:04.274000: INFO : Narrowed integration regression window from [9ad9e8f0, 084a3a63] (3 builds) to [9ad9e8f0, db3d691b] (2 builds) (~1 steps left)

2026-08-12T01:29:06.634000: DEBUG : Found commit message:
Bug 2037368 - Add support for tiled images via quad shaders. r=gw

Differential Revision: https://phabricator.services.mozilla.com/D298783

Comment 28

[Ali]

I also experience the video scrolling bug, but I believe that has been around much longer than 152.

Myself and many others have posted about it here, where I narrowed it down to these builds in mozregression:
Last working build: 2022-03-17
First bad build: 2022-03-18

'gfx.webrender.dcomp-video-vp-scaling-win = false' does not resolve the issue for me, but
'gfx.webrender.dcomp-video-hw-overlay-win = true' does. (including on OP's reddit example)

Comment 29

[Ali]

Apologies, I mistyped. I meant hw-overlay-win set to false fixes the issue.

Comment 30

[media-triaging-bot]

Analysis

Two separate problems look conflated in this bug.

1. Scroll hitching with no media on the page (comment 7, comment 15, and :ali's rnz.co.nz repro in comment 27). The mozregression run in comment 27 narrowed to two integration builds and named bug 2037368 ("Support tiled images with the quad shader code path", Core :: Graphics: WebRender). That change landed 2026-05-13 with a target milestone of 152 Branch, i.e. during the 152 Nightly cycle -- so it was already present in 152.0 (released 2026-06-16) and therefore in 152.0.4 and 152.0.5 as well. The bisection looks sound, but it makes this a 151 -> 152 regression, not a 152.0.4 -> 152.0.6 dot-release regression, so the premise in the summary should be corrected. That is also consistent with comment 14 / comment 15, where the reporter found the hitch reproducing as far back as the 2026-06-16 builds.

2. Scroll stutter that is worse when video is on the page (the original report). Per comment 28 this one goes back to 2022-03-18 builds and is resolved for :ali by gfx.webrender.dcomp-video-hw-overlay-win=false -- that is bug 1954406, which is already open.

From the A/V side this does not look like a decode problem. In the media-preset profile from comment 18 (reddit, Firefox 152 / Win10 / NVIDIA):

• The media pipeline is clean: 15x WMFVDecoderCreation Success, no NS_ERROR_*, no DXVA failures, and MediaDecoderStateMachine #17 uses only 105 ms of CPU across a 41.5 s capture.
• Essentially all the cost is in the GPU process: Renderer 4.29 s and WRRenderBackend#1 2.40 s, inside Renderer::render_impl -> draw_frame (71% of Renderer samples) / composite_simple (45%), with 35% of Renderer samples in the NVIDIA user-mode driver (nvwgf2umx.dll) under DrawElementsInstanced.
• The long frames are Composite #1 intervals of 13-28 ms, and Renderer::draw_tile_list sits on the hot path -- the same code area bug 2037368 touched.
• DCSurfaceVideo markers fire at ~253/s, with DCSurfaceVideo::PresentVideo, DCLayerTree::AddSurface and DCLayerTree::MaybeCommit all present, matching the DirectComposition video swap-chain cost described in comment 24.

This also explains why media.hardware-video-decoding.enabled=false helped in comment 5 without this being a decoder bug: software-decoded frames are not promoted to a DirectComposition video overlay, so that pref removes the DCSurfaceVideo swap-chain work from the compositor rather than changing anything in the decode path. It is a compositor-side side effect, not a media workaround.

Moving to Graphics: WebRender and recording the regressor relation for the no-media half -- both candidate causes (bug 2037368 and bug 1954406) live there.

Notes for whoever picks this up

• :hello's per-site A/B results for the two prefs requested in comment 24 are in attachment 9623148 [details], with a correction in comment 26 -- worth reading before asking for a re-run.
• The needinfo on :sotaro from 2026-07-14 is still open, and is still the right person for the DirectComposition video-overlay half.
• Several prefs have been toggled over the life of this bug (gfx.webrender.layer-compositor, media.hardware-video-decoding.enabled, gfx.webrender.dcomp-video-vp-scaling-win, gfx.webrender.dcomp-video-hw-overlay-win). Any further before/after comparison should start from a confirmed all-default about:config.

Next Steps

• :nical -- could you sanity-check whether bug 2037368 could plausibly cause intermittent long composites while scrolling? The bisection in comment 27 points at it, and Renderer::draw_tile_list is on the hot path in the profile above.
• Splitting the video-overlay half out (or duping it forward to bug 1954406) would make both halves tractable; as filed they have different regression windows and different workaround prefs.

Comment 31

[[:Propheticus]]

Could bug 2027413 also be related? Mentions hardware acceleration + video watching + scrolling and freezing symptoms.
The entire browser UI freezing until switching focus away and back could be a different issue than a simple scroll stutter. But perhaps it's a more extreme version of the symptom caused by the same hardware accelerated video decoding issue?