VLC 2.0 has 20% higher CPU usage?

[JohnMcPayne]

Yea, thank you very much!

Maybe this would be a nice feature for a future version:
Some kind of battery saving mode or something like that.

Something where you just have to make one or two clicks to activate it.

[utemplates]

Ok, another great solution seems to be to use the linear profile and not the main one for the resampler.

Can you kindly give a step-by-step guide of how to do that? Is this a setting that I adjust in the preferences?

[Jean-Baptiste Kempf]

Yes, this is.

[utemplates]

Well, what about that guide I browser around the VLC preferences menu and can you please confirm if this is what I am supposed to actually change:
Under advanced preference menu (accessible by clicking "Show All" button), then "Audio" and then "Miscellaneous" and then "SRC resampler", I can see that the "Sample rate converter type is currently set on "Sinc function (medium quality)". Is it that one that I am supposed to change to "Linear (fastest)"? And is there anything else I will need to adjust to execute your solution?

I basically have no idea if I am even on the correct menu so a reply with some details, at least with more than three words would be appreciated I meant that without sarcasm.

[JohnMcPayne]

Yes, you found the right menu. Just use "Linear".

[utemplates]

Excellent cheers John. Will this decrease much audio quality?

[JohnMcPayne]

I didn't notice much difference though I'm not sure if there is noticeable difference at all.

The one thing that might happen is that audio stutters or tiny bits are left out as much as I understand.

But if you are on a Mac you only have 3 choices:
- Play your videos with Quicktime 10 and Perian, which is the best one, because you have hardware acceleration, but there will be a lot of files that even Perian isn't able to play. And you won't have subtitles. CPU load will be around 7-9% playing a 720p MKV x264.

- Play your videos with any other video software like VLC. Here VLC seems to be the best choice with 25-40% CPU load, depending on what settings you use. Compared to that MPlayerX has a CPU usage of 60% and more which is ridiculous.

- Play your videos with XBMC, the only video software on Mac which uses hardware acceleration, but the side effect that it isn't a stand alone video software, but a media center all in one solution and it is kind of sluggish compared to the other ways. CPU load will be at 15%.

I just tried out a nightly build and it works fine and it runs on 30% CPU usage without the need of tinkering around in the settings.

[Jean-Baptiste Kempf]

I just tried out a nightly build and it works fine and it runs on 30% CPU usage without the need of tinkering around in the settings.

Which one did you try?

[utemplates]

Well I'm currently using the latest version of Movist which uses way less CPU than the v2 VLC version so that is an alternative for anyone that is looking for one.

[JohnMcPayne]

I just tried out a nightly build and it works fine and it runs on 30% CPU usage without the need of tinkering around in the settings.

Which one did you try?

vlc-2.1.0-x86_64-20120224-0114.zip 24-Feb-2012 01:14 27M

This one.
Still looking for a nightly build containing hardware acceleration, though. ;P


@utemplates:
movist isn't a good alternative as the free version is buggy and I don't want to pay for a version I cannot test. Besides that, there are limitations to software decoding which means that I don't believe a program can at the moment be much more energy efficient than VLC without hardware acceleration. I think movist just uses lesser quality which means lesser CPU load.

[Jean-Baptiste Kempf]

I just tried out a nightly build and it works fine and it runs on 30% CPU usage without the need of tinkering around in the settings.

Which one did you try?

vlc-2.1.0-x86_64-20120224-0114.zip 24-Feb-2012 01:14 27M

This one.

This is like just removing libsamplerate.

[utemplates]

@utemplates:
movist isn't a good alternative as the free version is buggy and I don't want to pay for a version I cannot test. Besides that, there are limitations to software decoding which means that I don't believe a program can at the moment be much more energy efficient than VLC without hardware acceleration. I think movist just uses lesser quality which means lesser CPU load.

Yes the free version is quite buggy and old. Although I have the very latest paid version, it has a great couple of bugs but the playback is excellent and instant. Is there any way that I can test quality to see differences between VLC and Movist? As it stands currently, Movist is less CPU heavy I assure you.

[Jean-Baptiste Kempf]

We will find a solution for 2.0.1 to have less CPU usage.

[RTouris]

Alright, did a little research and here's my findings so far with regards to version 2.0...

My testbed for the tests conducted below is a MacBookPro 2.2 2.33Ghz C2D with a Radeon X1600/256MB running OSX 10.6.8 with all the current updated applied to the system. The test cases consist of:

a. a local .avi FILE with an XVid video stream of resolution 624x352, framerate 23.97fps, 4:2:0 YUV with a stereo mp3 audio stream of 48Khz of 128kbps bitrate and
b. an internet ~700kbps .xspf STREAM with an MJPEG video strem of resolution 640x480, framerate 4fps, 4:2:2 YUV with a monaural IMA WAV ADPCM audio stream of 8Khz of 16kbps bitrate...

The results appear to be consistent and reproducible with other files of similar specs and are presented in precedence of VLC 2.0 default first (Sinc function (medium quality) / Zero Order Hold - Linear (which produce the exact same result) / followed by VLC 1.1.12 / followed by the QT 10.0-Perian bundle (quicktime.app + QTkit server add-up), windowed mode first, followed by fullscreen mode within brackets.

Without further ado...

case a. 31-34% [26-34%] / 16-18% [14-16%] / 13-15% [13-15%] / 28-30% [28-30%]
case b. 7-14% audio stutter [7-12% audio stutter] / 8-10% [7-8%] / 7-9% [7-9%] / no results for the QT-perian bunlde N/A

under case b. using the default settings for vlc 2.0 (sinc function - medium quality) the app would fail to load the required stream properly off the server which would produce a "spike" effect in the download timeline and the corresponding audio stutter. Reverting to either Zero Order Hold or Linear seems to remedy this behaviour.

It is obvious from the above that there is a certain amount of debugging that needs to be done with the current code of VLC 2.0 for OSX...The thing that strikes me most above all is that for some peculiar reason VLC windowed appears to drain more CPU cycled compared to fullscreen mode, which to be honest i don't quite understand (one would expect it to be the other way around since the app would have to account for extrapolating instructions to present the small video stream on a larger scale window - i suppose this is handled by core graphics in OSX?).

The laptop was plugged into an ac outlet during the tests. It is worth noticing that having CoolBook installed i was able to monitor the current CPU frequency (throttling level set to high) and the CPU was practically "idling" in the lowest possible setting of 996Mhz with a few occurrences where it would throttle up to the next setting (1494Mhz) and then back again...

For any further comments do not hesitate to post here, i will try to do my best and report back promptly.

[Jean-Baptiste Kempf]

For any further comments do not hesitate to post here, i will try to do my best and report back promptly.

I love you!

[nsmlv]

My computer's OS is Vista and I am expericing high CPU usage up to 70% when streaming my desktop to TV.
Would you please inform me which files I can delete in order to enhance the performance?