Transcoding Performance vs. That Other Server...

[pünktchen 1258]

FYI:

http://slhck.info/video/2017/02/24/crf-guide.html

http://goughlui.com/2016/08/27/video-compression-testing-x264-vs-x265-crf-in-handbrake-0-10-5/

[BAlGaInTl 279]

FYI:

http://slhck.info/video/2017/02/24/crf-guide.html

http://goughlui.com/2016/08/27/video-compression-testing-x264-vs-x265-crf-in-handbrake-0-10-5/

 

Lots of good info there, but a lot to wrap my head around too.

 

Still nobody seems to address CRF value and cpu strain/usage.  Most of the time, they are throwing more than enough power at it.

 

I'll see if I can't come up with something to test that on my system.  Could be useful information for folks using lower power systems.

[BAlGaInTl 279]

I've done some tweaking...

 

I've learned quite a bit already.  Mostly, I've learned that ffmpeg employs some sort of witchcraft with 0s and 1s.  No really... that's the only thing that can explain it.

 

 

That being said... 19 crf and superfast helps a lot.  I'm not sure what is magic about 19.  It's the same number Plex seemed to use.  I tried 18 and 20 with mixed results.

 

The streaming is more acceptable now.  Still not quite as smooth as Plex, but I will keep working at it.

 

I also found that I get MUCH better perfomance in the Android App than the Web App.  About 85 fps in Android and borderline 60 fps in the web.  Which is fine for me as I will just install apps everywhere.  

Edited March 9, 2017 by BAlGaInTl

[concord 0]

That being said... 19 crf and superfast helps a lot.  I'm not sure what is magic about 19.  It's the same number Plex seemed to use.  I tried 18 and 20 with mixed results.

 

The streaming is more acceptable now.  Still not quite as smooth as Plex, but I will keep working at it.

 

I also found that I get MUCH better perfomance in the Android App than the Web App.  About 85 fps in Android and borderline 60 fps in the web.  Which is fine for me as I will just install apps everywhere.  

 

Thanks for the info, changing to 19 helps like you stated.  Turning on hardware acceleration didn't help, guess my ivy-bridge chip is too old.  

 

Still, there's motion blur, especially on sports.  I'm testing this with a Roku 3.  I recorded part of an ACC basketball game, then played with the default settings.  I tried lower the H.264 option to slow, slower, etc but that made it worse, so I selected superfast and it improved, then went to ultrafast.  Taking your suggestion to set cfr to 19 does improve playback but it's not quite there to my satisfaction.

 

At first, my experimental (Fedora 24)  system was a HP N54L (my original server for KODI RasbPi and Chromebox clients ).  Tried to watch live tv like MASH with default settings and it was washed out along with motion blur.  When trying to watch a live ACC game (1080i), it would buffer also.   I then moved my drives to my Plex server (i5-2400S) and no more buffering.  When using Plex DVR and then playing back, their modified ffmpeg seems to do a better job.  Still, both systems seem to max out cpu usage around 30-40  (both 4 cores, so it ranged between 250-350 total) for 1080i playback.  Maybe going with a faster/newer CPU would help.  

[BAlGaInTl 279]

Thanks for the info, changing to 19 helps like you stated.  Turning on hardware acceleration didn't help, guess my ivy-bridge chip is too old.  

 

Still, there's motion blur, especially on sports.  I'm testing this with a Roku 3.  I recorded part of an ACC basketball game, then played with the default settings.  I tried lower the H.264 option to slow, slower, etc but that made it worse, so I selected superfast and it improved, then went to ultrafast.  Taking your suggestion to set cfr to 19 does improve playback but it's not quite there to my satisfaction.

 

At first, my experimental (Fedora 24)  system was a HP N54L (my original server for KODI RasbPi and Chromebox clients ).  Tried to watch live tv like MASH with default settings and it was washed out along with motion blur.  When trying to watch a live ACC game (1080i), it would buffer also.   I then moved my drives to my Plex server (i5-2400S) and no more buffering.  When using Plex DVR and then playing back, their modified ffmpeg seems to do a better job.  Still, both systems seem to max out cpu usage around 30-40  (both 4 cores, so it ranged between 250-350 total) for 1080i playback.  Maybe going with a faster/newer CPU would help.  

 

Glad my experimentation could help you too.

 

I have the option of going to faster hardware, but I really don't want to.  I think the benefits of my power sipping server outweigh the transcoding problems for me.  Since it has 8 true cores, it serves files and multitasks like a beast.   

[concord 0]

Glad my experimentation could help you too.

 

I have the option of going to faster hardware, but I really don't want to.  I think the benefits of my power sipping server outweigh the transcoding problems for me.  Since it has 8 true cores, it serves files and multitasks like a beast. 

 

I hope you updated your BMC firmware on your ASRock C2750D4i, heard they were dropping like flies after 15-18 months of use.  Something about the EEPROM being written to too much and wearing it out.

 

In any case, I hope to experiment more this weekend with Emby. I like Emby's client with the standard guide and support for the older HDHomeRun tuners.   Was playing with the Roku beta client, it's good but hope that the guide paging improves.    In the mean time I purchased an i-7-7700, motherboard, ram.  It's more than your 20w, but still 65w is not bad.  The Atom 4 core in the HP N54L just couldn't handle 1080i transcoding, but sounds like your  8 core processor can.