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 17th February 2017, 15:40   #1001  |  Link
CruNcher
Registered User
 
CruNcher's Avatar
 
Join Date: Apr 2002
Location: Germany
Posts: 4,926
Nothing for the masses really in its overall price/performance balancing
Also well still need to understand how efficiently Ryzen will work together MultiGPU and how especially the ACE will interact between IGPU/CPU and Discrete sharing workloads and balancing them efficiently supporting each other, where Nvidia has 0 connection point with Intel and Nvidia mostly hacking up and workaround stuff
__________________
all my compares are riddles so please try to decipher them yourselves :)

It is about Time

Join the Revolution NOW before it is to Late !

http://forum.doom9.org/showthread.php?t=168004

Last edited by CruNcher; 17th February 2017 at 16:04.
CruNcher is offline   Reply With Quote
Old 17th February 2017, 15:48   #1002  |  Link
NikosD
Registered User
 
Join Date: Aug 2010
Location: Athens, Greece
Posts: 2,901
Intel's HEDT processors, those ultra expensive 6, 8, 10 core processors, will be the first victims of RyZen beasts.

With Intel's prices ~1700 $ and very expensive motherboards, those processors are dead on arrival of RyZen.

Besides very specific AVX/AVX2 performance, those Intel's CPUs are a completely meaningless choice.
__________________
Win 10 x64 (19042.572) - Core i5-2400 - Radeon RX 470 (20.10.1)
HEVC decoding benchmarks
H.264 DXVA Benchmarks for all
NikosD is offline   Reply With Quote
Old 17th February 2017, 16:35   #1003  |  Link
nevcairiel
Registered Developer
 
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,344
Quote:
Originally Posted by CruNcher View Post
will interact between IGPU/CPU and Discrete sharing workloads and balancing them efficiently supporting each other, where Nvidia has 0 connection point with Intel and Nvidia mostly hacking up and workaround stuff
Ryzen doesn't have a iGPU.

Ryzen is at a point between Intels consumer models and the HEDT platform. It offers 8 cores at a price somewhere in the HEDT range (ie. latest rumors put EU prices at ~600€ for the 1800X, which is slightly above a i7 6850k, or the 1700X for ~470€ which sits similar to a 6800k), but on the other hand it doesn't have some of the HEDT features like extra PCIe lanes for multi-GPU.
__________________
LAV Filters - open source ffmpeg based media splitter and decoders

Last edited by nevcairiel; 17th February 2017 at 16:41. Reason: typos
nevcairiel is offline   Reply With Quote
Old 17th February 2017, 17:30   #1004  |  Link
CruNcher
Registered User
 
CruNcher's Avatar
 
Join Date: Apr 2002
Location: Germany
Posts: 4,926
Raven Ridge (Ryzen APU) will have it based on most probably VEGA IP and run on every since then released AM4 Ryzen Mainboard
The basic Zen CPU is not that highlight but that combination combined with a Discrete HBM2 GPU and the workloads it allows Async efficiently powered by Vulkan
__________________
all my compares are riddles so please try to decipher them yourselves :)

It is about Time

Join the Revolution NOW before it is to Late !

http://forum.doom9.org/showthread.php?t=168004

Last edited by CruNcher; 17th February 2017 at 17:38.
CruNcher is offline   Reply With Quote
Old 17th February 2017, 17:57   #1005  |  Link
NikosD
Registered User
 
Join Date: Aug 2010
Location: Athens, Greece
Posts: 2,901
In order to not misinform anyone we must put the facts as they are.

The 8c/16t RyZen 1800X with 3.6GHz base clock and 4.0GHz turbo has a TDP of 95W and will cost ~600€

The 8c/16t Intel i7 6900K with 3.2GHz base clock and 3.7GHz turbo has a TDP of 140W and costs ~1200€

So, the Intel's equivalent for most workloads of RyZen 1800X has double price.

In other words, dead meat.
__________________
Win 10 x64 (19042.572) - Core i5-2400 - Radeon RX 470 (20.10.1)
HEVC decoding benchmarks
H.264 DXVA Benchmarks for all
NikosD is offline   Reply With Quote
Old 18th February 2017, 11:31   #1006  |  Link
leonccyiu
Registered User
 
Join Date: Feb 2015
Posts: 45
Quote:
Originally Posted by NikosD View Post
Intel's HEDT processors, those ultra expensive 6, 8, 10 core processors, will be the first victims of RyZen beasts.

With Intel's prices ~1700 $ and very expensive motherboards, those processors are dead on arrival of RyZen.

Besides very specific AVX/AVX2 performance, those Intel's CPUs are a completely meaningless choice.
What about Ultra HD Blu-ray and Netflix in 4K? I read that they require Kaby Lake for copy protection, even Intel's HEDT platform doesn't work with either of those it's frustrating.
leonccyiu is offline   Reply With Quote
Old 18th February 2017, 13:41   #1007  |  Link
nevcairiel
Registered Developer
 
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,344
Quote:
Originally Posted by leonccyiu View Post
What about Ultra HD Blu-ray and Netflix in 4K? I read that they require Kaby Lake for copy protection, even Intel's HEDT platform doesn't work with either of those it's frustrating.
It requires the Kaby Lake GPU, which the HEDT CPUs obviously don't have, since they don't have any iGPU.

With any hope dedicated GPUs from NVIDIA and AMD can also certify for that soon.
__________________
LAV Filters - open source ffmpeg based media splitter and decoders
nevcairiel is offline   Reply With Quote
Old 1st March 2017, 16:31   #1008  |  Link
hajj_3
Registered User
 
Join Date: Mar 2004
Posts: 1,120
Netflix has given a talk about VP9 using low bitrates for static scenes: https://www.engadget.com/2017/03/01/...ad-connection/

They haven't made a blog post about it, at least not yet.
hajj_3 is offline   Reply With Quote
Old 2nd March 2017, 06:32   #1009  |  Link
CruNcher
Registered User
 
CruNcher's Avatar
 
Join Date: Apr 2002
Location: Germany
Posts: 4,926
Nothing new Really going 100 kbps on such a device when complexity is low the screen size allows it to hide most artifacts and if it's to blurry post sharpening will help it perceptually for the viewer but watching that 100 kbps stream then at 1080p or UHD and you will see all issues, HDMI out

And if you show something like this to your normal press guy audience of course they'll be impressed.

Though absolutely nothing new Mobile Hype

Per Pixel spatial temporal resolution overhead depending on the Viewing conditions see this for example the bigger the screen and resolution the more it will start to fall apart in motion

https://www.sendspace.com/file/050xe5

This is also why Metrics need to be taken into account with the actual Viewing condition and some already do that, especially for Mobile a low PSNR/SSIM can be sufficient enough your target surely doesn't need to be in the 50 or 0.99 range for that small display size of a Mobile device to be accepted as OK by the viewing audience if it's not showing significant prediction errors that distract

Quote:
As it turns out, Netflix seems to have cracked the code, and in a way that feels sort of obvious.

The thing to remember is that not all movies (or TV shows, for that matter) are created equal. Long, lingering static shots obviously aren't as complex as fight scenes, but to date, Netflix has been encoding those videos as though they were the same.
Geez no wonder trump want's to ban journalists, what for a snake oil seller capability that guy has explaining the difference between CBR and VBR + Psy optimization

Especialy how he doesn't let Netflix look bad for what they did in the Past and Sold but glorifies them to the Olymp with the "they cracked the code" geez crazy


But overall if you dig deeper you read out of what that show up actually was about

It is about the Integration of VMAF results as Metric for at least the Complexity Masking inside of their VP9 Encoder for their further encodings so instead of PSNR and SSIM or (enter something else) they gonna use their VMAF based (Deep Learned) approach as Encoder Psy tuning internally which makes sense, for what would they have developed it else.
__________________
all my compares are riddles so please try to decipher them yourselves :)

It is about Time

Join the Revolution NOW before it is to Late !

http://forum.doom9.org/showthread.php?t=168004

Last edited by CruNcher; 2nd March 2017 at 07:45.
CruNcher is offline   Reply With Quote
Old 2nd March 2017, 08:20   #1010  |  Link
LigH
German doom9/Gleitz SuMo
 
LigH's Avatar
 
Join Date: Oct 2001
Location: Germany, rural Altmark
Posts: 6,753
Don't miss the point that CBR (or "constricted VBR") is the rather obvious approach for delivering content over limited bandwidth. Less constricted VBR would only be applicable for rather huge buffers.
__________________

New German Gleitz board
MediaFire: x264 | x265 | VPx | AOM | Xvid
LigH is offline   Reply With Quote
Old 4th March 2017, 08:44   #1011  |  Link
Blue_MiSfit
Derek Prestegard IRL
 
Blue_MiSfit's Avatar
 
Join Date: Nov 2003
Location: Los Angeles
Posts: 5,988
Yep, switching away from CBR / small buffer+maxrate VBR to true (almost) unrestricted 2 pass VBR is a big advantage in the download scenario. Long, fully adaptive GOPs also help.

ABR encoding is typically done with fixed GOP so the work can be parallelized across many servers (aka split and stitch) while still maintaining GOP alignment. In download we don't care about GOP alignment so we can let the encoder really stretch its legs
Blue_MiSfit is offline   Reply With Quote
Old 4th March 2017, 17:23   #1012  |  Link
CruNcher
Registered User
 
CruNcher's Avatar
 
Join Date: Apr 2002
Location: Germany
Posts: 4,926
Yes but the core of this Presentation was their VMAF Research Investments
__________________
all my compares are riddles so please try to decipher them yourselves :)

It is about Time

Join the Revolution NOW before it is to Late !

http://forum.doom9.org/showthread.php?t=168004
CruNcher is offline   Reply With Quote
Old 14th March 2017, 19:28   #1013  |  Link
dapperdan
Registered User
 
Join Date: Aug 2009
Posts: 201
Some multithreaded encoding improvements for VP9 encoding:

https://groups.google.com/a/webmproj...el/oiHjgEdii2U
dapperdan is offline   Reply With Quote
Old 14th March 2017, 19:35   #1014  |  Link
dapperdan
Registered User
 
Join Date: Aug 2009
Posts: 201
Netflix (who also are namechecked in the above multi-threading email) talk a little about their use of VP9 for Android downloads here:

http://techblog.netflix.com/2017/03/...n-android.html

Most of it is about Android development, but there's a few parpagraphs on VP9 under the heading "Improving Video Quality".
dapperdan is offline   Reply With Quote
Old 4th April 2017, 16:21   #1015  |  Link
mandarinka
Registered User
 
mandarinka's Avatar
 
Join Date: Jan 2007
Posts: 729
https://groups.google.com/a/webmproj...ss/G_fuix013KM

Seems libvpx is getting improved threading for VP9 encoding. No idea if it is something similar to slices or WPP, or to actual frame threads. Anybody knows what the details are?
mandarinka is offline   Reply With Quote
Old 4th April 2017, 16:45   #1016  |  Link
Mystery Keeper
Beyond Kawaii
 
Mystery Keeper's Avatar
 
Join Date: Feb 2008
Location: Russia
Posts: 724
Quote:
Originally Posted by mandarinka View Post
https://groups.google.com/a/webmproj...ss/G_fuix013KM

Seems libvpx is getting improved threading for VP9 encoding. No idea if it is something similar to slices or WPP, or to actual frame threads. Anybody knows what the details are?
I don't. But I think their approach is dumb. What I want is this.
__________________
...desu!
Mystery Keeper is offline   Reply With Quote
Old 4th April 2017, 17:56   #1017  |  Link
mandarinka
Registered User
 
mandarinka's Avatar
 
Join Date: Jan 2007
Posts: 729
That method was actually provisionally used in x265, in 2013.
But I think it was only for purposes of early realtime encoding demos.

It was eventually removed, probably because of the huge downsides in memory consumption. And the fact that for offline encoding, it is more practical to just encode in parts?
mandarinka is offline   Reply With Quote
Old 4th April 2017, 18:06   #1018  |  Link
Mystery Keeper
Beyond Kawaii
 
Mystery Keeper's Avatar
 
Join Date: Feb 2008
Location: Russia
Posts: 724
Quote:
Originally Posted by mandarinka View Post
That method was actually provisionally used in x265, in 2013.
But I think it was only for purposes of early realtime encoding demos.

It was eventually removed, probably because of the huge downsides in memory consumption. And the fact that for offline encoding, it is more practical to just encode in parts?
That's not trivial to automate. I don't even know how I could merge the encoded cuts.
__________________
...desu!
Mystery Keeper is offline   Reply With Quote
Old 20th July 2017, 17:43   #1019  |  Link
Clare
Registered User
 
Join Date: Apr 2016
Posts: 61
I'm doing some tests over a selection of 30 clips encoded at different crf values in order to compare them.

On PSNR-HVS-M, x265 is slightly better than VP9:


But if we look at the VMAF score, VP9 is squeezing more quality per bit:


I was surprised, I thought x265 would have beat VP9 by a margin on VMAF, especially considering its psychovisual optimizations.

Last edited by Clare; 20th July 2017 at 18:27.
Clare is offline   Reply With Quote
Old 21st July 2017, 13:46   #1020  |  Link
dapperdan
Registered User
 
Join Date: Aug 2009
Posts: 201
Have VP9 (and AV1) been working on doing better on VMAF? With Netflix adopting them it wouldn't be that surprising if tweaks or bug-fixes that particularly affect VMAF got prioritized.

The question then is if it is "teaching to the test" and therefore not a valid quality increase, or if VMAF is good enough to direct them towards genuine improvements that agree with subjective viewers.
dapperdan is offline   Reply With Quote
Reply

Tags
google, ngov, vp8, vp9, vpx, webm

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 00:15.


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