AnimeSuki Forums

Register Forum Rules FAQ Members List Social Groups Search Today's Posts Mark Forums Read

Go Back   AnimeSuki Forum > Anime Related Topics > General Anime > Fansub Groups

Notices

Reply
 
Thread Tools
Old 2008-08-17, 23:33   Link #21
Meltingice
Klipsch
 
 
Join Date: Oct 2007
Location: Frontier
Quote:
Originally Posted by Prestwick View Post
Text...

TL-DR question: How easy would it be to release a common standard of encoding for those wanting to encode for devices beyond the PC?
Not too many encoders care about the part in bold.
Meltingice is offline   Reply With Quote
Old 2008-08-18, 00:07   Link #22
Prestwick
The Victorian
 
Join Date: Aug 2008
Location: UK
Send a message via MSN to Prestwick
Its a bit irrelevant now anyway as pointed out in the TL-DR answer straight after that.
Prestwick is offline   Reply With Quote
Old 2008-08-18, 14:16   Link #23
comatose
Senior Member
 
Join Date: Dec 2007
Quote:
Originally Posted by Daiz View Post
Only 8 refs and b-frames? That's stupid, especially with level 5.1... 16 or nothing!
True that. You might as well go 4.1 with those settings for PS3/X360.
comatose is offline   Reply With Quote
Old 2008-08-18, 19:01   Link #24
TheFluff
Excessively jovial fellow
 
 
Join Date: Dec 2005
Location: ISDB-T
Age: 37
Real Encoders always use 16 references and set --level 5.1 explicitly on all content regardless of actual level/profile just to annoy PS3/X360 owners and DXVA users.

(I'm such a nice guy you see)
__________________
| ffmpegsource
17:43:13 <~deculture> Also, TheFluff, you are so fucking slowpoke.jpg that people think we dropped the DVD's.
17:43:16 <~deculture> nice job, fag!

01:04:41 < Plorkyeran> it was annoying to typeset so it should be annoying to read
TheFluff is offline   Reply With Quote
Old 2008-08-18, 20:27   Link #25
lubczyk
Senior Member
 
Join Date: Jan 2005
What's the benefit of using 16 reference frames in contrast to just 5 or 3? Aren't the benefits of such high settings almost negligible?
lubczyk is offline   Reply With Quote
Old 2008-08-18, 21:45   Link #26
TheFluff
Excessively jovial fellow
 
 
Join Date: Dec 2005
Location: ISDB-T
Age: 37
Quote:
Originally Posted by lubczyk View Post
What's the benefit of using 16 reference frames in contrast to just 5 or 3? Aren't the benefits of such high settings almost negligible?
The benefit is that it annoys people with braindead hardware players, which usually choke on that many references because their decoding buffer isn't very big.

That aside with anime 16 refs can actually be useful, the improvements are small but they do exist. Empirical observations from my own encodes indicate that refs beyond 10 are used about 1-2% of the time.
__________________
| ffmpegsource
17:43:13 <~deculture> Also, TheFluff, you are so fucking slowpoke.jpg that people think we dropped the DVD's.
17:43:16 <~deculture> nice job, fag!

01:04:41 < Plorkyeran> it was annoying to typeset so it should be annoying to read
TheFluff is offline   Reply With Quote
Old 2008-08-18, 22:21   Link #27
Dark Shikari
x264 Developer
 
 
Join Date: Feb 2008
Quote:
Originally Posted by lubczyk View Post
What's the benefit of using 16 reference frames in contrast to just 5 or 3? Aren't the benefits of such high settings almost negligible?
For anime, in my experience, each ref beyond 5-6 can give up to 1% efficiency improvement--far more than non-anime.
Dark Shikari is offline   Reply With Quote
Old 2008-08-19, 07:55   Link #28
martino
makes no files now
 
 
Join Date: May 2006
Quote:
Originally Posted by TheFluff View Post
Real Encoders always use 16 references and set...
Awww... I'm fake. D:
__________________
"Light and shadow don't battle each other, because they're two sides of the same coin"
martino is offline   Reply With Quote
Old 2008-08-19, 12:07   Link #29
fireshark
CRF Believer
*Fansubber
 
 
Join Date: Jan 2007
Send a message via AIM to fireshark
NOOO! Forget about 4.1, my iPod only takes 3.0!

>>1 No one is going to bother, since they have no interest in encoding for that platform. Either trust "press botan, get encode" or become an encoder to see what everyone is making a fuss about.
fireshark is offline   Reply With Quote
Old 2008-08-19, 18:43   Link #30
Left64Vegeta
Fansubber
*Fansubber
 
 
Join Date: Jun 2007
Quote:
Originally Posted by Dark Shikari View Post
For anime, in my experience, each ref beyond 5-6 can give up to 1% efficiency improvement--far more than non-anime.
I wonder... at what cost in cpu consumption of the final viewer's pc ?
Left64Vegeta is offline   Reply With Quote
Old 2008-08-19, 19:03   Link #31
DryFire
Panda Herder
 
Join Date: Dec 2005
Location: A bombed out building in Beruit.
Quote:
Originally Posted by Left64Vegeta View Post
I wonder... at what cost in cpu consumption of the final viewer's pc ?
Not enough for us (encoders) to care.
DryFire is offline   Reply With Quote
Old 2008-08-19, 19:10   Link #32
Left64Vegeta
Fansubber
*Fansubber
 
 
Join Date: Jun 2007
Quote:
Originally Posted by DryFire View Post
Not enough for us (encoders) to care.
There have been countless times, though, that I've heard complaints from single-core users, complaining about unacceptable playback on their machines. (dropped frames, audio/video desync etc). So, as an encoder, I am really interesting in sorting out this matter with real numbers and percentages.
Left64Vegeta is offline   Reply With Quote
Old 2008-08-19, 19:16   Link #33
jfs
Aegisub dev
 
 
Join Date: Sep 2004
Location: Stockholm, Sweden
Age: 39
I'm pretty sure that more reference frames doesn't increase CPU usage at decoding, only memory usage because it needs to keep more decoded frames in memory.

Actually, on a second though it might increase CPU usage because B-frames may refer to much later frames meaning that more later frames must be decoded before a frame can be displayed.
__________________

Aegisub developer [ Forum | Manual | Feature requests | Bug reports | IRC ]
Don't ask for: More VSFilter changes (I won't), karaoke effects, help in PM's
jfs is offline   Reply With Quote
Old 2008-08-19, 23:39   Link #34
DryFire
Panda Herder
 
Join Date: Dec 2005
Location: A bombed out building in Beruit.
From what I understand bitrate, CABAC and resolution have the biggest impact. If you're very curious try a crf encodes with different reference frames and see what the difference in decoding speed is with timecodec.
DryFire is offline   Reply With Quote
Old 2008-08-19, 23:41   Link #35
neothe0ne
Senior Member
 
 
Join Date: Mar 2008
Quote:
Originally Posted by Left64Vegeta View Post
There have been countless times, though, that I've heard complaints from single-core users, complaining about unacceptable playback on their machines. (dropped frames, audio/video desync etc). So, as an encoder, I am really interesting in sorting out this matter with real numbers and percentages.
I myself caught complaints from some users that Eclipse and gg's 720p h264 files lag when Menclave's 720p h264 files of the same episodes did not on anidb. I would also be curious as to what settings exactly result in higher CPU utilization when decoding.
__________________
neothe0ne is offline   Reply With Quote
Old 2008-08-19, 23:44   Link #36
Dark Shikari
x264 Developer
 
 
Join Date: Feb 2008
Quote:
Originally Posted by DryFire View Post
From what I understand bitrate, CABAC and resolution have the biggest impact. If you're very curious try a crf encodes with different reference frames and see what the difference in decoding speed is with timecodec.
The four biggies are resolution, CABAC, bitrate, and deblocking.

B-frames slow down decoding, but they also reduce bitrate at the same quality level, so they're a net win, since the bitrate affects decoding speed a lot too.

Multiple reference frames has nearly no impact on decoding speed (basically zero). It might even help because, again, it reduces bitrate at the same quality level.
Dark Shikari is offline   Reply With Quote
Old 2008-08-20, 01:07   Link #37
raziel666
Cogito, Ergo Sum
*Fansubber
 
 
Join Date: Jul 2006
Age: 43
Quote:
Originally Posted by Dark Shikari View Post
The four biggies are resolution, CABAC, bitrate, and deblocking.
I think psyRDO slows down decoding as well. It might not be noticeable in standard resolutions and/or framerates, but there was this 60fps ED which I encoded with both psyRDO enabled and disabled.

The result is that the psyRDO disabled plays everywhere smoothly, while the psyRDO enabled stutters even in a friend's 2-month Core2Duo (can't remember the model atm).
__________________
Brighter Death Now
エルゴ プラクシー
死の代理人
Ergo Proxy
The Agent Of Death
raziel666 is offline   Reply With Quote
Old 2008-08-20, 01:08   Link #38
Left64Vegeta
Fansubber
*Fansubber
 
 
Join Date: Jun 2007
Quote:
Originally Posted by neothe0ne View Post
I myself caught complaints from some users that Eclipse and gg's 720p h264 files lag when Menclave's 720p h264 files of the same episodes did not on anidb. I would also be curious as to what settings exactly result in higher CPU utilization when decoding.
There is a nice program out there, called MediaInfo, that can read the encoding settings of a x264 video. It would be nice if someone that has both releases could show us too.

Quote:
Originally Posted by Dark Shikari View Post
The four biggies are resolution, CABAC, bitrate, and deblocking.
In another post, you were saying about a crappy encoding, which had no CABAC and no deblocking. So I guess, deactivating those two is not an option. As for the resolution, it is understandable, that bigger resolutions will result in a lot more partitions. However, I think that downscaling a nice raw source is neither an option.

So, what is left is the bitrate. Surely, when the bitrate increases, so does the complexity of the content it describes. In other words, more bitrate describes a content with more details. Absolutely true. But then again, we can' t lower it too much, can we?

Quote:
Originally Posted by Dark Shikari View Post
B-frames slow down decoding, but they also reduce bitrate at the same quality level, so they're a net win, since the bitrate affects decoding speed a lot too.

Multiple reference frames has nearly no impact on decoding speed (basically zero). It might even help because, again, it reduces bitrate at the same quality level.
I am pretty confused about how a reference works. Especially when it comes to a combined matter of both b-frames and references.

I haven't really studied how x264 exactly works, but as far as I know, in a traditional mpeg there are I-frames and B-frames. Now, when a b-frame is both dependent to I-frames and B-frames, I guess we have to do the extra work of decoding the depencies first. And if I am right, the concept of the "b-pyramid" refers to the fact that these dependencies can be dependent on other frames, too.

References, especially mixed-references, come to make this thing even more complex. It sounds to me like we are not only dealing with the individual frames and their decoding, but we are trying to deal with every partition of every frame.

Let's say that a b-frame needs another 16 b-frames to be decoded. And let's also say that each of these 16 b-frames, has 16 references for its partitions. I guess with mixed ref, the reference can be a b-frame as well, which can also have further references and dependencies.

So, how can it be that more reference frames provide us with a more lightweight encoding, instead of making it a lot heavier?

Last edited by Left64Vegeta; 2008-08-20 at 02:37.
Left64Vegeta is offline   Reply With Quote
Old 2008-08-20, 07:42   Link #39
DryFire
Panda Herder
 
Join Date: Dec 2005
Location: A bombed out building in Beruit.
Quote:
Originally Posted by raziel666 View Post
I think psyRDO slows down decoding as well.
PsyRDO is completely encoder-side. It may be giving the ED more bitrate.
DryFire is offline   Reply With Quote
Old 2008-08-20, 08:56   Link #40
Dark Shikari
x264 Developer
 
 
Join Date: Feb 2008
Quote:
Originally Posted by raziel666 View Post
I think psyRDO slows down decoding as well. It might not be noticeable in standard resolutions and/or framerates, but there was this 60fps ED which I encoded with both psyRDO enabled and disabled.

The result is that the psyRDO disabled plays everywhere smoothly, while the psyRDO enabled stutters even in a friend's 2-month Core2Duo (can't remember the model atm).
No, Psy-RD has absolutely no effect on decoding speed; its solely an analysis option.
Quote:
Originally Posted by Left64Vegeta View Post
There is a nice program out there, called MediaInfo, that can read the encoding settings of a x264 video. It would be nice if someone that has both releases could show us too.
strings filename | less is more complete and easier...

Quote:
Originally Posted by Left64Vegeta View Post
Let's say that a b-frame needs another 16 b-frames to be decoded. And let's also say that each of these 16 b-frames, has 16 references for its partitions. I guess with mixed ref, the reference can be a b-frame as well, which can also have further references and dependencies.
The entire point of reference frames is that they're already decoded, so you're just keeping it in memory and grabbing pixels from it whenever needed.
Dark Shikari is offline   Reply With Quote
Reply

Tags
best practice, common standards, encoding

Thread Tools

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 -5. The time now is 07:52.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.
We use Silk.