Pausing play causes cpu spin

[minusf]

Hello. Before I file an issue, I was wondering if anyone sees similar behaviour on macbook pro m5, macos 27.0.1, mpv 0.41.0.

I start listening to a bandcamp album thru yt-dlp (all latest packages from brew), e.g.:

mpv --no-video https://predawka.bandcamp.com/album/quasiorganic
client removed during hook handling
Playing: https://predawka.bandcamp.com/track/off-tray
 ● Audio  --aid=1  (mp3 2ch 44100 Hz 128 kbps)
File tags:
 Artist: predawka
 Album: quasiorganic
 Album_Artist: predawka
 Date: 20250725
 Title: off tray
 Track: 1
 Uploader: predawka
AO: [coreaudio] 44100Hz stereo 2ch floatp
A: 00:00:20 / 00:02:10 (16%) Cache: 108s/3MB

And then I pause it. Seemingly nothing wrong is happening, but if I come back in 30 minutes, mpv is using 5-6% of one core and the notebook temperature has gone up from 31C to 55C. The moment I quit the player, temperature starts decreasing back to normal levels. Reproducable every time.

mpv v0.41.0 Copyright © 2000-2025 mpv/MPlayer/mplayer2 projects
libplacebo version: v7.360.1
FFmpeg version: 9.0.2
FFmpeg library versions:
   libavcodec      63.1.102
   libavdevice     63.1.102
   libavfilter     12.1.102
   libavformat     63.1.102
   libavutil       61.1.102
   libswresample   7.1.102
   libswscale      10.1.102

[minusf]

I forgot to say: playing itself does not cause that kind of cpu usage, just when pausing.

[low-batt]

On this:

I was wondering if anyone sees similar behaviour on macbook pro m5, macos 27.0.1, mpv 0.41.0.

With the limited evidence and without a log file I'm not sure if the scenario I'm about to describe is what you are encountering or not. With mpv 0.41.0 under macOS 27 the following can happen…

First, the coreaudio output driver fails due to issue #18384. Also good to be aware that there is a Core Audio related crash than can occur, issue #18274.

When the Core Audio driver fails, mpv falls back to the avfoundation audio output driver. That driver is susceptible to the problem described in PR #18390 which results in a CPU spin. That problem is occurs when:

• AVFoundation output driver is used
• Playback is paused
• Audio output device changes

When that problem occurs it can be cleared by resuming playback and pausing again.

Check to see if you see mpv falling back to AVFoundation. Think about what might be changing the audio device. Typically this is caused by AirPods connecting. Finally, if resuming and pausing playback clears the CPU spin, then suspect the problem in PR #18390.

I believe mpv plans on fixing all of these problems for the 0.42.0 release.

[minusf]

Thank you for writing this up with all the pointers! Thanks for all your work on iina as well.

It very much looks like this is the case happening.

I went to read a bunch of issues you reported on and just want to give back a small tip, hopefully useful:

-while true; do ps -c -o 'comm,%cpu,cputime' | grep mpv | grep -v grep; sleep 5; done
+while true; do ps -c -o 'comm,%cpu,cputime' | grep '[m]pv' | sleep 5; done

[low-batt]

I always like sharing knowledge. Sometimes people will point out something you missed. Or sometimes people will give you a TIP on how to improve something. Always nice to simplify bash commands. There is a typo. Should be a ; before sleep, not |, like so:

while true; do ps -c -o 'comm,%cpu,cputime' | grep '[m]pv'; sleep 5; done

I will be sure to remember to use this concise form.

The bad news is that mpv 0.42.0 still exhibits the CPU spin. PR #18390 was moved from the v0.42.0 milestone to the v0.43.0 one and currently has not been merged.

[minusf]

In my understanding now that 0.42 fixed coreaudio usage, avfoundation issues would remain more obscure as it's used less in general.

[low-batt]

Yes, the coreaudio initialization failure that was causing mpv to fall back to using the avfoundation output driver has been fixed in mpv 0.42.0. However fans of Spatial Audio have to use the avfoundation driver, so it is still important to address the CPU spin. At least with the coreaudio fixes it no longer affects all users running macOS 27.

The CPU spin problem is actively being worked on.