Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Video Encoding > MPEG-4 AVC / H.264

Reply
 
Thread Tools Search this Thread Display Modes
Old 29th August 2013, 16:29   #1  |  Link
montreal.simo
Registered User
 
Join Date: Aug 2012
Posts: 17
Bandwidth measurements - influence of different parameters

Hello,
I'm trying to understand the influence of different x264 encoding parameters on the bandwidth consumption.
I tried different partition analysis, reference frames, me, subme... but I have a minor influence.
Are there some parameters which are interesting to test that affect the bandwidth?
Thx for help

___
A.M

Last edited by montreal.simo; 29th August 2013 at 21:24. Reason: typo
montreal.simo is offline   Reply With Quote
Old 29th August 2013, 22:22   #2  |  Link
sneaker_ger
Registered User
 
Join Date: Dec 2002
Posts: 5,570
"bandwidth" as in "bitrate"?
Many parameters have influence on the bitrate in constant quality modes like --cfr, but this kind of testing does not really make all that much sense since you can within certain limits arbitrarily control the bitrate through --crf or --bitrate (+ things like vbv).

To get a rough idea on the influence of the parameters you could take a look at the differences of the presets listed under x264.exe --fullhelp.

Maybe it would be better to ask what exactly you want to achieve, i.e. are you looking for some practical advice for a problem at hand or do you just want to do some kind of scientific tests with no real purpose?
sneaker_ger is offline   Reply With Quote
Old 30th August 2013, 18:20   #3  |  Link
montreal.simo
Registered User
 
Join Date: Aug 2012
Posts: 17
hello, thank you for your answer,
This is a technical work and I'm looking for some practical advice,
I'm using x264 for streaming.
In fact, I made latency ​​measurements for different encoding parameters. I am now looking to do the same thing for bandwidth.
the basic idea is simple, if we change encoding parameters, what's the impact or influence on the bandwidth. for example,for latency, the ultrafast presets and tune ZeroLatency reduce latency.
it is clear that the bitrate and vbv are important, given that the fixed rate is used. but if, for example instead of using dia motion estimation , we us hex or umh, then the bandwidth must change. So, my question is if there are any kind of encoding parameters that can directly affect the bandwidth at the output.

thx
montreal.simo is offline   Reply With Quote
Old 30th August 2013, 19:03   #4  |  Link
Groucho2004
 
Join Date: Mar 2006
Location: Barcelona
Posts: 5,042
Quote:
Originally Posted by montreal.simo View Post
hello, thank you for your answer,
This is a technical work and I'm looking for some practical advice,
I'm using x264 for streaming.
In fact, I made latency ​​measurements for different encoding parameters. I am now looking to do the same thing for bandwidth.
the basic idea is simple, if we change encoding parameters, what's the impact or influence on the bandwidth. for example,for latency, the ultrafast presets and tune ZeroLatency reduce latency.
it is clear that the bitrate and vbv are important, given that the fixed rate is used. but if, for example instead of using dia motion estimation , we us hex or umh, then the bandwidth must change. So, my question is if there are any kind of encoding parameters that can directly affect the bandwidth at the output.

thx
Bandwidth is a property of a data transmission system. I think you're confusing this with the bitrate of a video stream (or more general, a data stream).
Think of bandwidth as the diameter of a water hose and bitrate as the water pressure.
__________________
Groucho's Avisynth Stuff
Groucho2004 is offline   Reply With Quote
Old 30th August 2013, 19:07   #5  |  Link
Asmodian
Registered User
 
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,263
Quote:
Originally Posted by montreal.simo View Post
So, my question is if there are any kind of encoding parameters that can directly affect the bandwidth at the output.
--bitrate, --crf, or --qp are the only ones that directly affect the bitrate.

Quality is proportional to bitrate. I think of settings like motion estimation as changing the CPU time vs quality, not bitrate. Of course the relationship between settings, bitrate, and quality is really a complex multidimensional surface, not a 2D graph. However, it is very hard to get a good estimation of how much a setting affects size at the same quality because there are no good objective measurements of quality and changing a setting changes both the size and the quality. Also how much a particular setting helps/hurts can change quite a lot depending on the source.

Like sneaker_ger already suggested, I think you should look at the settings which change as you change presets, these are the settings with the strongest effect. As you move from ultrafast to veryslow presets you get diminishing returns (less quality gained per increase in encoding time / less size lost per increase in encoding time / less size lost at the same quality).

Or you can pick the slowest preset that gives you the latency you need and you will get the highest quality you can given the bandwidth you have available.
Asmodian is offline   Reply With Quote
Old 30th August 2013, 22:11   #6  |  Link
xooyoozoo
Registered User
 
Join Date: Dec 2012
Posts: 200
Basically what everyone else said, but if you want to measure things yourself and put it in terms relevant to your company's interests, you can use bdrate to interpolate average percentage size change. To put it in rough visual terms, if you wanted to compare the red and black line in this graph graph, you'd do this and then calculate average change per log bitrate. The JCT docs often have this calculation included in an Excel macro.

If you choose to use 2pass, you'd have to make sure the metric extremities of each encoder are reasonably close. You can choose whichever metric (PSNR/logSSIM/etc) you want as long as you are aware of the caveats for each and make sure x264 is tuned appropriately. And there's always the obvious caveat that no algorithmic method can 100% predict human vision. However, that doesn't mean one can't go past 0%.
xooyoozoo is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 20:30.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, vBulletin Solutions Inc.