Commit "ab4421e" degraded rendering quality of vo_gpu_next

[Svetloslav]

mpv Information

Build of 2025-03-01 vs build of 2025-02-25 by zhongfly

Other Information

- Windows version: Windows 11 Enterprise 24h2
- GPU model, driver and version: NVIDIA 3035 Studio 572.16 (latest studio)
- Source of mpv: zhongfly's last build on 20250301 (including latest changes to mpv) 
- Latest known working version: zhongfly build of 20250225
- Issue started after the following happened: after the update to the 20250301 build

Reproduction Steps

Play a 4k HDR mkv with vo_gpu_net enabled, configured to render only on GPU and with further rendering filters added in the 20250225 and then in the 20250301 builds. On the same hardware NVIDIA 3050, rendering has degraded, it is now choppy with 20250301 build, whereas with 20250225 build is smooth. Can be reproduced by simply switching the mpv components (portable) back and forth between the indicated builds of zhongfly

Expected Behavior

4k HDR mkv video on a Samsung mini led and NVIDIA 3050 smooth playback

Actual Behavior

Same video is choppy. Same hardware, same configuration of mpv.

Log File

log.log

Sample Files

No response

I carefully read all instruction and confirm that I did the following:

• I tested with the latest mpv version to validate that the issue is not already fixed.
• I provided all required information including system and mpv version.
• I produced the log file with the exact same set of files, parameters, and conditions used in "Reproduction Steps", with the addition of --log-file=output.txt.
• I produced the log file while the behaviors described in "Actual Behavior" were actively observed.
• I attached the full, untruncated log file.
• I attached the backtrace in the case of a crash.

[Dudemanguy]

I think you meant that you experience performance issues not rendering quality. But anyways that commit nor anything recent in mpv should have anything to do with performance.

[zhongfly]

Assuming there is indeed a difference in performance, it's not necessarily related to mpv commits, and could be caused by updates to downstream dependency libraries.
Because with every build, most dependency libraries use the latest commits pulled by git.

You can read the Packages Version section of the job summary via the Build Details link provided in my release notes.All dependent repositories that use git version control are listed with the commit sha used at build time.

[llyyr]

Your log file url is invalid btw, can you reupload it? Though as others have said, none of the recent changes in mpv itself are likely to cause this

[Svetloslav]

Assuming there is indeed a difference in performance, it's not necessarily related to mpv commits, and could be caused by updates to downstream dependency libraries. Because with every build, most dependency libraries use the latest commits pulled by git.

You can read the Packages Version section of the job summary via the Build Details link provided in my release notes.All dependent repositories that use git version control are listed with the commit sha used at build time.

Thank you for pointing that out. I will stick for now with the older build and will eventually give a try to a future build and see if the issue has been resolved.

[kasper93]

Could you tell which builds exactly are you comparing? There are like 6 zhongfly builds on 20250225

[Svetloslav]

Could you tell which builds exactly are you comparing? There are like 6 zhongfly builds on 20250225

Of course, I always take the last one for a specific day, since it always includes the latest commits.

[kasper93]

I don't see anything significant that would cause this. Could you try to narrow down the range of builds? You can also use ones on our CI. Each merge commit has attached build if you click on the tick mark and summary.

EDIT:

Could you test those two builds and tell how they work?

A: https://github.com/mpv-player/mpv/actions/runs/13546934850/artifacts/2699233949
B: https://github.com/mpv-player/mpv/actions/runs/13531680724/artifacts/2699243531