Feedbacks needed for Segmental Video Encoding

Discussing about advanced video encoding techniques including GPU encoding, distributive encoding and segmental encoding. Topics are not limited to MediaCoder.
Posts: 2
Joined: Mon Dec 02, 2013 7:15 pm

Re: Feedbacks needed for Segmental Video Encoding

Postby IRBabuin » Mon Dec 02, 2013 7:25 pm

I have a 4 (8 logical) core i7 2630QM and I see this unusual behaviour when encoding using x264/CPU...It does not load the cores at maximum and from time to time, the frequency scales back to 800MHz (from 2200-2600). I tried to load progressively the cores 1-2-3-4, only when using 1-2 cores I saw full load and the frequency stable at maximum. I tried SVE but I can't say that I saw much of an improvement. Encoding was done with the file on the sistem disk (SSD) and output file on the same disk. 6GB DDR3 RAM.

Is this normal for x264/xvid? To not load the cores properly?
Posts: 1
Joined: Sat Oct 25, 2014 7:44 am

Re: Feedbacks needed for Segmental Video Encoding

Postby MusicFreakS » Sat Oct 25, 2014 7:52 am

I'm also testing a with the SVE.

What i noticed is this

when i use Cocurrent segments 4 the encode will stop after about 1 minute and gives a error 14.
i use a 3 pass h264 hi10p. The cpu will go 100% and memory about 55-60%

i have a sandy bridge -e i7 3820 Cpu with 16 gb ddr3 memory.

is 4 segments too heavy for the pc then?
Site Admin
Site Admin
Posts: 4071
Joined: Mon May 15, 2006 7:43 pm
Location: Sydney

Re: Feedbacks needed for Segmental Video Encoding

Postby stanley » Sun Nov 02, 2014 11:58 am

3-pass may not work weel with SVE. Please try 1-pass or 2-pass instead.
When things work together, things work.
Posts: 16
Joined: Sat Aug 24, 2013 10:24 pm

Re: Feedbacks needed for Segmental Video Encoding

Postby nodiaque » Sun Nov 16, 2014 2:36 am

I don't know if it's my setting but I failed to see the problem on my computer (why would I use SVE).

I have a i7-920 @4.2ghz with 6g of ram. I encode raw blu-ray mkv to x264 using cpu only (no cuda, quicksync, opencli...). I have my own custom preset, using high10 and level 4.1 with 2-pass. When I start the job, all 8 logical core are at 100% and only stop once the encoding is over.

I see that SVE is suppose to help fill all core, but my cpu never sleep. Is it because my computer is too old and newer doesn't get filled as much? I often think about adding cuda and opencli, but when I read about these two, there's always a chance to have lower quality then pure CPU, and there's option you can't use with them (my encoding settings are quite tweaked for maximum compression/quality).
Posts: 14
Joined: Mon Mar 17, 2014 9:30 am

Re: Feedbacks needed for Segmental Video Encoding

Postby Valagard » Thu Jan 01, 2015 9:34 pm

I've never had SVE work correctly, no matter what version of Mediacoder I've used eventually the audio is always out of sync no matter how I encode it.

The video always works correctly, but the audio after 20+ min is incredibly out of sync, even though video works flawlessly

Here's a suggestion, if SVE is causing the audio to go all weird cause its been segmented, why not just do two passes? An audio and then video pass, and then truncate the files?

By doing this you end up avoiding all the issues by segmental encoding the audio
Posts: 1
Joined: Wed Oct 28, 2015 11:54 pm

Re: Feedbacks needed for Segmental Video Encoding

Postby oyabun » Thu Oct 29, 2015 12:54 am

Hi, I am running Mediacoder build 5782 on a 6-core (12 hyperthreaded) Xeon X5675@4.35GHz. Encoding Blade Runner Blu-ray.

Since core utilization is ~75% with 2-pass H265 and WPP enabled I tried using SVE with 2-pass H265. I first tested it on a 30-sec sequence, between 10:00 and 10:30. The original encode lasted 94.6secs, with 2 concurrent segments the time went down to 89.8secs, which is only a 5.5% improvement. Perhaps with 4 segments it would have been larger.

However the final video was spliced all wrong (incorrect sequence of segments - this could be an incompatibility of SVE and selecting a specific time bracket). Additionally, when using 4 segments, H265 encoding stops after a while with a message that "slice=P but 2pass stats say B". Last it is not possible to encode ACE->HE-AAC 5.1 audio with SVE, after a while a warning about Audio encoder thread blocked comes up and encoding ends. So basically SVE was unusable.

I think that in general SVE is not suited to 2-pass encoding, at lease in its current implementation. Even if it worked correctly, the quality penalty from using such small segments would be too large. Maybe there should be a way to split the video in 2,3 or 4 parts (or create 2, 3 or 4 decoding outputs from the same file), encode them in parallel as separate files and then merge the results.

From what I gather, Distributed Encoding is basically SVE over multiple PCs, therefore if I can't get SVE to work on a single PC, Distributed Encoding won't work either, correct?

Return to “GPU/Distributive/Segmental Encoding”

Who is online

Users browsing this forum: No registered users and 1 guest