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 > VP9 and AV1

Reply
 
Thread Tools Search this Thread Display Modes
Old 5th November 2019, 09:40   #1921  |  Link
nevcairiel
Registered Developer
 
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,344
Comparisons between LAV 0.74.1 and later nightly versions are flawed since the threading strategy changed in FFmpeg, which resulted in 0.74.1 using more frame threads then the later nightlies, making 0.74.1 artificially faster. As such, all your results are invalidated.
This is why you should use as little software as possible to do benchmarking (ie. go as close to the core as possible), as you never know what changes might interfer with your conclusions.

I've also once again changed the thread distribution in 0.74.1-30 from last night, and while its going to use more threads again now, similar to the old logic, its not going to be identical to 0.74.1 in all cases (because I added more tile threads on high core-count CPUs)
__________________
LAV Filters - open source ffmpeg based media splitter and decoders

Last edited by nevcairiel; 5th November 2019 at 09:48.
nevcairiel is offline   Reply With Quote
Old 7th November 2019, 23:56   #1922  |  Link
Mr_Khyron
Member
 
Mr_Khyron's Avatar
 
Join Date: Nov 2002
Posts: 203
AOMedia Research Symposium 2019 Videos
https://www.youtube.com/playlist?lis...wewtWKpxXky8iI
Mr_Khyron is offline   Reply With Quote
Old 8th November 2019, 14:49   #1923  |  Link
utack
Registered User
 
Join Date: Apr 2018
Posts: 63
Quote:
Originally Posted by marcomsousa View Post
Lesson learnt from WebP
There are only two I can think of
  • despite being technically more advanced you can still lose to a decades old legacy format when your encoder is terrible
  • it does not matter that your format is worse than the legacy competition, if you claim that it is better often enough others will start parroting it and adopt it

Seriously the only area where it might be a tiny bit better is for ultra-high compression where it does not start falling apart as badly as jpeg, for any sane (mid ot high) image quality range the vast array of jpeg encoders are doing a significantly better job of retaining detail
utack is offline   Reply With Quote
Old 8th November 2019, 20:04   #1924  |  Link
dapperdan
Registered User
 
Join Date: Aug 2009
Posts: 201
Webp had some other benefits over JPEG outside of compressing photographic images.


JPEG XL seems like WebP's successor in this regard. It's targeted at lots of pain points that would make it a good choice to replace JPEG (and PNG and GiF) on the web and in the browser even if it didn't beat JPEG on compression, though it claims that as well. And maybe the JPEG name will help, though that doesn't seem to have benefitted anyone but the original JPEG.

Not sure there's room for AVIF and JPEG XL but maybe they have subtly different niches.
dapperdan is offline   Reply With Quote
Old 8th November 2019, 20:09   #1925  |  Link
dapperdan
Registered User
 
Join Date: Aug 2009
Posts: 201
Ronald's slide showing 4 AV1 encoders all scaling well seems like an improvement from his slide at BIG Apple Video where only SVT seemed to be managing that, with Eve just behind and Rav1e and libaom trailing.

Not sure it it's a direct comparison to the earlier slide but if it is then things should be a lot better for AV1 when cores are available.
dapperdan is offline   Reply With Quote
Old 9th November 2019, 01:38   #1926  |  Link
soresu
Registered User
 
Join Date: May 2005
Location: Swansea, Wales, UK
Posts: 196
Quote:
Originally Posted by dapperdan View Post
Ronald's slide showing 4 AV1 encoders all scaling well seems like an improvement from his slide at BIG Apple Video where only SVT seemed to be managing that, with Eve just behind and Rav1e and libaom trailing.

Not sure it it's a direct comparison to the earlier slide but if it is then things should be a lot better for AV1 when cores are available.
An interesting point Ronald made implies that AV1 has an intrinsic parallel scaling limitation due to an oversight during the encoder development (12:20 in the video), something to do with superblock boundaries.

Hopefully a lesson learned for AV2 efforts going forward.
soresu is offline   Reply With Quote
Old 9th November 2019, 15:49   #1927  |  Link
marcomsousa
Registered User
 
Join Date: Jul 2018
Posts: 80
Rav1e release 0.1.0
First official release, published during the Video Dev Days 2019 in Tokyo.

Features
  • Intra and inter frames
  • 64x64 superblocks
  • 4x4 to 64x64 RDO-selected square and 2:1/1:2 rectangular blocks
  • DC, H, V, Paeth, smooth, and a subset of directional prediction modes
  • DCT, (FLIP-)ADST and identity transforms (up to 64x64, 16x16 and 32x32 respectively)
  • 8-, 10- and 12-bit depth color
  • 4:2:0 (full support), 4:2:2 and 4:4:4 (limited) chroma sampling
  • 11 speed settings (0-10)
  • Near real-time encoding at high speed levels
  • Rate control (single-pass and two-pass)
  • Temporal RDO
  • Scene cut detection
  • CLI tool and C API

https://github.com/xiph/rav1e/releases/tag/0.1.0
__________________
AV1 win64 VS2019 builds
Last build here

Last edited by marcomsousa; 9th November 2019 at 16:17.
marcomsousa is offline   Reply With Quote
Old 9th November 2019, 16:41   #1928  |  Link
mzso
Registered User
 
Join Date: Oct 2009
Posts: 930
Hi!
Are there any AV1 videos on youtube besides the beta playlist? So far I haven't found any.
mzso is offline   Reply With Quote
Old 9th November 2019, 16:49   #1929  |  Link
marcomsousa
Registered User
 
Join Date: Jul 2018
Posts: 80
Quote:
Originally Posted by mzso View Post
Hi!
Are there any AV1 videos on youtube besides the beta playlist? So far I haven't found any.
Almost all top videos, but only in low resolutions.
__________________
AV1 win64 VS2019 builds
Last build here
marcomsousa is offline   Reply With Quote
Old 9th November 2019, 19:20   #1930  |  Link
mzso
Registered User
 
Join Date: Oct 2009
Posts: 930
Quote:
Originally Posted by marcomsousa View Post
Almost all top videos, but only in low resolutions.
Thanks. Well, I guess I won't be coming across many then. I don't watch stuff like that, and it looks like a few million views are far from enough. A 4+ billion Ed Sheeran song had it up to to 2160p, but a 2+ billion Taylor swift song only has it up to 720p. I managed to find some Wired videos with a couple million views, that have AV1 though.

It seems like Firefox's (well, Waterfox's to be accurate) AV1 decoding is quite poor. MPV's (after upgrading) and LAV's seem to be a lot better, no hangs or stutter.
mzso is offline   Reply With Quote
Old 10th November 2019, 09:29   #1931  |  Link
marcomsousa
Registered User
 
Join Date: Jul 2018
Posts: 80
HandBrake 1.3.0 Released

* Added support for reading AV1 via libdav1d

https://github.com/HandBrake/HandBrake/releases
__________________
AV1 win64 VS2019 builds
Last build here
marcomsousa is offline   Reply With Quote
Old 10th November 2019, 16:25   #1932  |  Link
dapperdan
Registered User
 
Join Date: Aug 2009
Posts: 201
Quote:
Originally Posted by mzso View Post
A 4+ billion Ed Sheeran song had it up to to 2160p, but a 2+ billion Taylor swift song only has it up to 720p. I managed to find some Wired videos with a couple million views, that have AV1 though.
A YouTube engineer made a comment about people using YouTube as a radio station, but their licence requiring video, so ultra low bitrate AV1 being a kind of workaround for contractual obligations.

So it's possible that YouTube is measuring the bitrate/resolution of these views and prioritising the ones that are viewed at high quality, not just the ones that are viewed a lot. Just a guess though.
dapperdan is offline   Reply With Quote
Old 11th November 2019, 07:58   #1933  |  Link
Beelzebubu
Registered User
 
Join Date: Feb 2003
Location: New York, NY (USA)
Posts: 109
Quote:
Originally Posted by soresu View Post
An interesting point Ronald made implies that AV1 has an intrinsic parallel scaling limitation due to an oversight during the encoder development (12:20 in the video), something to do with superblock boundaries.
The superblock boundary one to get superblock-row multithreading (an encoder-side-only version of partitions in vp8 or wavefront in hevc) is quite micro, because it is easily worked around by just ignoring the superblock edge's correctness and sacrifice your search' accuracy a tiny little bit. The frame multi-threading one is a bigger deal.
Beelzebubu is offline   Reply With Quote
Old 11th November 2019, 22:50   #1934  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 4,752
Quote:
Originally Posted by dapperdan View Post
A YouTube engineer made a comment about people using YouTube as a radio station, but their licence requiring video, so ultra low bitrate AV1 being a kind of workaround for contractual obligations.

So it's possible that YouTube is measuring the bitrate/resolution of these views and prioritising the ones that are viewed at high quality, not just the ones that are viewed a lot. Just a guess though.
A lot of the "TV for radio" clips have super static backgrounds or just scrolling lyrics. So they should encode down super small with more advanced codecs and long GOPs.
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 18th November 2019, 23:09   #1935  |  Link
Adonisds
Registered User
 
Join Date: Sep 2018
Posts: 14
When do you think Stadia using AV1 will be available?
Adonisds is offline   Reply With Quote
Old 19th November 2019, 00:00   #1936  |  Link
marcomsousa
Registered User
 
Join Date: Jul 2018
Posts: 80
Quote:
Originally Posted by Adonisds View Post
When do you think Stadia using AV1 will be available?
Stadia is use cheap client cpu, so AV1 will be in next year for high-end cpu, and the follow year to cheap devices.. So 2 to 3 year for sure.

Marco Sousa
__________________
AV1 win64 VS2019 builds
Last build here
marcomsousa is offline   Reply With Quote
Old 19th November 2019, 08:25   #1937  |  Link
soresu
Registered User
 
Join Date: May 2005
Location: Swansea, Wales, UK
Posts: 196
Quote:
Originally Posted by marcomsousa View Post
Stadia is use cheap client cpu, so AV1 will be in next year for high-end cpu, and the follow year to cheap devices.. So 2 to 3 year for sure.

Marco Sousa
There's nothing to suggest AV1 will be in either Samsung or Qualcomm's flagship SoC next year, Qualcomm still has yet to even join AOM when last I checked.

On the other hand the likes of Amlogic have plans to have AV1 in lower end chips for 2020, so you appear to be wrong on both counts.
soresu is offline   Reply With Quote
Old 19th November 2019, 18:18   #1938  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 4,752
Quote:
Originally Posted by marcomsousa View Post
Stadia is use cheap client cpu, so AV1 will be in next year for high-end cpu, and the follow year to cheap devices.. So 2 to 3 year for sure.
On the decode side, maybe in a few years. But Does Stadia have a path to a good 4Kp60 encoder with sufficient efficiency and cost?

There is a lot of untapped potential in using how the game is rendered to drive encoder optimization that could help here (the game knows your motion vectors!). Not sure if anyone's researching that with AV1 currently.
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 19th November 2019, 18:55   #1939  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 4,752
Quote:
Originally Posted by soresu View Post
There's nothing to suggest AV1 will be in either Samsung or Qualcomm's flagship SoC next year, Qualcomm still has yet to even join AOM when last I checked.

On the other hand the likes of Amlogic have plans to have AV1 in lower end chips for 2020, so you appear to be wrong on both counts.
So far, all the announced AV1 chips have been for living room devices, and nothing for mobile, correct?

Living room has a lot more breathing room for power consumption, thermal management, and thus process. The mobile chips are where every extra transistor costs the most.

Of course, having a working HW decoder at all is a huge milestone, even if it might take a bit for those to migrate into mobile SoCs.
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 20th November 2019, 03:47   #1940  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
i wouldn't wait for them to announce a hardware decoder they may just add them or evne ship them already without a word and without any way to access them.

as an example the "first" nvidia card with an HEVC main10 decoder was the 960 this card has a VP9 profile 0 decoder too it was not possible to access it for about a year or so.

the 960 was release in 22.01.2015
VP9 was finalised 17.06.2013
that's just 17 months
i dare to say that AV1 has far more attraction then VP9

AV1 is 19 month old so if someone really wanted to they could have added it and it wouldn't be the first time they made it public which may sound odd at first but if it can't be used anyway it may just be a better this way.

Quote:
There is a lot of untapped potential in using how the game is rendered to drive encoder optimization that could help here (the game knows your motion vectors!). Not sure if anyone's researching that with AV1 currently.
using multiply frames for encoding is a very bad thing in this case because latency is very important in this task there is a reason the currently released stadia is at best a very poor joke they want money for.

this only counts for lookahead which should be completely avoided if possible.
and how would you even access these information from the game in the first place.
huhn 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 00:28.


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