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 14th December 2018, 22:14   #1321  |  Link
Motenai Yoda
Registered User
 
Motenai Yoda's Avatar
 
Join Date: Jan 2010
Posts: 709
the ones that don't care about gpu capabilities and still get a display without need a discrete card
__________________
powered by Google Translator
Motenai Yoda is offline   Reply With Quote
Old 14th December 2018, 22:19   #1322  |  Link
nevcairiel
Registered Developer
 
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,342
Quote:
Originally Posted by Nintendo Maniac 64 View Post
But lets be honest here - with AMD finally being a viable alternative again, who is really buying Intel for their graphics capabilities?
Gen11 is also supposed to be significantly faster. And Intel has among the best media capabilities today already, while AMD has the worst.
So for a small form factor media PC, there would be no competition for me.
__________________
LAV Filters - open source ffmpeg based media splitter and decoders

Last edited by nevcairiel; 14th December 2018 at 22:28.
nevcairiel is offline   Reply With Quote
Old 15th December 2018, 04:16   #1323  |  Link
Nintendo Maniac 64
Registered User
 
Nintendo Maniac 64's Avatar
 
Join Date: Nov 2009
Location: Northeast Ohio
Posts: 447
Quote:
Originally Posted by Motenai Yoda View Post
the ones that don't care about gpu capabilities and still get a display without need a discrete card
Uhhhh...
__________________
____HTPC____  | __Desktop PC__
2.93GHz Xeon x3470 (4c/8t Nehalem) | 4.5GHz 1.24v dual-core Haswell G3258
Radeon HD5870  | Intel iGPU      
2x2GB+2x1GB DDR3-1333 | 4x4GB DDR3-1600       
Nintendo Maniac 64 is offline   Reply With Quote
Old 17th December 2018, 13:43   #1324  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
you are aware that amd is still missing VP9 for hardware decoding even on vega.

while AMD is very competitive in the CPU market there GPU's are currently at an all time low. vega is using a lot of power needs a huge die and is really slow if you take size into consideration and the hardware decoder is pretty much worse than the nvidia cards that are getting 4 years old.
huhn is offline   Reply With Quote
Old 17th December 2018, 18:53   #1325  |  Link
Mr_Khyron
Member
 
Mr_Khyron's Avatar
 
Join Date: Nov 2002
Posts: 203
Quote:
Originally Posted by huhn View Post
you are aware that amd is still missing VP9 for hardware decoding even on vega.

while AMD is very competitive in the CPU market there GPU's are currently at an all time low. vega is using a lot of power needs a huge die and is really slow if you take size into consideration and the hardware decoder is pretty much worse than the nvidia cards that are getting 4 years old.
https://en.wikichip.org/w/images/a/a...whitepaper.pdf
on page 14
Quote:
Vega” can also decode the VP9 format at resolutions up to 3840x2160 using a hybrid approach where the video and shader engines collaborate to offload work from the CPU.
Mr_Khyron is offline   Reply With Quote
Old 17th December 2018, 19:56   #1326  |  Link
sneaker_ger
Registered User
 
Join Date: Dec 2002
Posts: 5,565
AFAIK the newer AMD ones like Ryzen 5 2500U (Raven Ridge/ Vega 8) now have VP9 10 bit ASIC decoding.

But yeah, they are late. Don't expect it to be different for AV1.

Last edited by sneaker_ger; 17th December 2018 at 19:59.
sneaker_ger is offline   Reply With Quote
Old 17th December 2018, 21:39   #1327  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
hybrid decoding has nothing todo with hardware decoding.

letting a CPU and GPU core do the work of an ASIC is simply not the same.

it would be nice if the newest APU have asic decoder but not that trust worth test i found told me the vega 11 is hybrid.
huhn is offline   Reply With Quote
Old 18th December 2018, 08:54   #1328  |  Link
alex1399
Registered User
 
Join Date: Jun 2018
Posts: 56
No sense, the amd hybrid decoding of vp9 just works as the way that intel do on the hybrid decoding of hevc in their 6th generation processor skylake. They ARE hardware decode.
alex1399 is offline   Reply With Quote
Old 18th December 2018, 19:33   #1329  |  Link
Blue_MiSfit
Derek Prestegard IRL
 
Blue_MiSfit's Avatar
 
Join Date: Nov 2003
Location: Los Angeles
Posts: 5,988
Keep things on topic, please.
Blue_MiSfit is offline   Reply With Quote
Old 18th December 2018, 22:31   #1330  |  Link
SmilingWolf
I am maddo saientisto!
 
SmilingWolf's Avatar
 
Join Date: Aug 2018
Posts: 95
Status report, redux
"rav1e is doing well" edition

1st edition: http://forum.doom9.org/showthread.ph...49#post1852449
2nd edition: http://forum.doom9.org/showthread.ph...87#post1857587
Whatever paragraph I don't repeat here can be assumed to be the same as in the aforementioned posts

First of all: graphs! Click to enlarge
Y axis: chosen metric
X axis: bits per pixel

720p:


1080p:


Encoders improvement over time:
720p:


1080p:


BD rates for 720p:
Code:
Codecs ladder:              |  x264 relative:
x264 -> rav1e               |  x264 -> rav1e
        RATE (%) DSNR (dB)  |          RATE (%) DSNR (dB)
 MSSSIM -10.7345 0.541324   |   MSSSIM -10.7345 0.541324
PSNRHVS -15.3271 1.07245    |  PSNRHVS -15.3271 1.07245
  HVMAF -7.40703 2.07138    |    HVMAF -7.40703 2.07138
----------------------------|-----------------------------
rav1e -> vp9                |  x264 -> vp9
        RATE (%) DSNR (dB)  |          RATE (%) DSNR (dB)
 MSSSIM -15.1057 0.68453    |   MSSSIM -21.81   1.08927
PSNRHVS -11.2436 0.654976   |  PSNRHVS -22.9586 1.51188
  HVMAF -19.2883 2.57019    |    HVMAF -24.1102 3.58993
----------------------------|-----------------------------
vp9 -> x265                 |  x264 -> x265
        RATE (%) DSNR (dB)  |          RATE (%) DSNR (dB)
 MSSSIM -4.25723 0.169151   |   MSSSIM -25.6195 1.21115
PSNRHVS -8.19042 0.41409    |  PSNRHVS -29.8289 1.83058
  HVMAF -10.6714 0.708441   |    HVMAF -31.2046 4.45371
----------------------------|-----------------------------
x265 -> av1                 |  x264 -> av1
        RATE (%) DSNR (dB)  |          RATE (%) DSNR (dB)
 MSSSIM -18.9088 0.7852     |   MSSSIM -38.0511 1.97653
PSNRHVS -15.3123 0.761791   |  PSNRHVS -38.6659 2.56119
  HVMAF -18.0023 1.0489     |    HVMAF -44.0411 4.3982
BD rates for 1080p:
Code:
Codecs ladder:                |  x264 relative:
x264 -> rav1e                 |  x264 -> rav1e
        RATE (%) DSNR (dB)    |          RATE (%) DSNR (dB)
 MSSSIM -20.0683 1.00011      |   MSSSIM -20.0683 1.00011
PSNRHVS -21.9935 1.47903      |  PSNRHVS -21.9935 1.47903
  HVMAF -18.4773 3.96202      |    HVMAF -18.4773 3.96202
------------------------------|-----------------------------
rav1e -> vp9                  |  x264 -> vp9
        RATE (%) DSNR (dB)    |          RATE (%) DSNR (dB)
 MSSSIM -18.2653 0.729489     |   MSSSIM -31.1754 1.49143
PSNRHVS -14.922  0.755605     |  PSNRHVS -30.1275 1.87845
  HVMAF -20.1645 2.31195      |    HVMAF -32.4505 4.72978
------------------------------|-----------------------------
vp9 -> x265                   |  x264 -> x265
        RATE (%) DSNR (dB)    |          RATE (%) DSNR (dB)
 MSSSIM 5.13717 -0.177855     |   MSSSIM -28.9956 1.18206
PSNRHVS -0.096748 -0.0123981  |  PSNRHVS -31.474  1.63676
  HVMAF -3.78107 0.0881882    |    HVMAF -34.6185 4.22357
------------------------------|-----------------------------
x265 -> av1                   |  x264 -> av1
        RATE (%) DSNR (dB)    |          RATE (%) DSNR (dB)
 MSSSIM -26.486  0.938124     |   MSSSIM -45.4959 2.12535
PSNRHVS -21.7431 0.905916     |  PSNRHVS -43.4792 2.56047
  HVMAF -22.7091 1.17861      |    HVMAF -48.0404 4.69582
Encoders:
x264 157-2935-545de2f
x265 2.9-4-471726d3a046
rav1e 0.1.0-977-64b9f501
libaom 1.0.0-908-g3a607f7b0
libvpx 1.7.0-1352-gea57f9acd

Cmdlines:
x264 --preset veryslow --tune ssim --crf 16 -o test.x264.crf16.264 orig.i420.y4m
x265 --preset veryslow --tune ssim --crf 16 -o test.x265.crf16.hevc orig.i420.y4m
rav1e --low_latency false -o test.rav1e.cq80.ivf --quantizer 80 -s 2 --tune psnr orig.i420.y4m
aomenc --frame-parallel=0 --tile-columns=3 --auto-alt-ref=1 --cpu-used=4 --tune=psnr --passes=2 --threads=2 --end-usage=q --cq-level=20 --test-decode=fatal -o test.av1.cq20.webm orig.i420.y4m
vpxenc --codec=vp9 --frame-parallel=0 --tile-columns=2 --good --cpu-used=0 --tune=psnr --passes=2 --threads=2 --end-usage=q --cq-level=20 --test-decode=fatal --ivf -o test.vp9.cq20.ivf orig.i420.y4m

Quality settings:
x264: CRF 16-24 step 1, and 24-34 step 2
x265: CRF 16-24 step 1, and 24-34 step 2
rav1e: CQ 80-160 step 16
aomenc: CQ 20-40 step 4
vpxenc: CQ 20-48 step 4
VMAF: model used: nflxall_vmafv4, pooling: harmonic_mean

Notes:
Revisiting the sequences from the previous report:
Code:
F.Y.C, x264 -> rav1e
        RATE (%) DSNR (dB)
 MSSSIM -21.8223 1.2258
PSNRHVS -28.408 2.29056
  HVMAF -18.7682 3.89782

PresageFlowerFight, x264 -> rav1e
        RATE (%) DSNR (dB)
 MSSSIM -33.8353 1.95641
PSNRHVS -33.5774 2.47118
  HVMAF -29.2079 7.17318

PresageFlowerWalk, x264 -> rav1e
        RATE (%) DSNR (dB)
 MSSSIM -6.98635 0.333085
PSNRHVS -3.78743 0.244148
  HVMAF 2.23865 -0.150589
MAJOR improvements in static/very low motion scenes thanks to adaptive keyframe selection, marginal improvements in the others.
SmilingWolf is offline   Reply With Quote
Old 19th December 2018, 00:36   #1331  |  Link
Phanton_13
Registered User
 
Join Date: May 2002
Posts: 95
SmilingWolf, can you also indicates the encoding speed, it don't need to be in fps as it can be in relation of a specific encoder, is more to see the variation in the speed of aoemc and rav1e due to optimizations or other improvements.
Phanton_13 is offline   Reply With Quote
Old 19th December 2018, 05:26   #1332  |  Link
utack
Registered User
 
Join Date: Apr 2018
Posts: 63
The cpu-used heuristics in libaom seems to be poorly tuned in lossless mode.
Tested with a digitally animated gif
hitting 50% size increase at 1/4 encode time saved compared to cpu-used=0, and double the size at 60% the encode time of cpu-used=0
each dot presets one cpu-used value

Last edited by utack; 19th December 2018 at 05:29.
utack is offline   Reply With Quote
Old 19th December 2018, 07:31   #1333  |  Link
Nintendo Maniac 64
Registered User
 
Nintendo Maniac 64's Avatar
 
Join Date: Nov 2009
Location: Northeast Ohio
Posts: 447
Quote:
Originally Posted by utack View Post
digitally animated gif
If you ever want to test higher-quality animation examples (read: not limited to 256 colors), then perhaps try an animated PNG.
__________________
____HTPC____  | __Desktop PC__
2.93GHz Xeon x3470 (4c/8t Nehalem) | 4.5GHz 1.24v dual-core Haswell G3258
Radeon HD5870  | Intel iGPU      
2x2GB+2x1GB DDR3-1333 | 4x4GB DDR3-1600       
Nintendo Maniac 64 is offline   Reply With Quote
Old 19th December 2018, 11:06   #1334  |  Link
SmilingWolf
I am maddo saientisto!
 
SmilingWolf's Avatar
 
Join Date: Aug 2018
Posts: 95
Quote:
Originally Posted by Phanton_13 View Post
SmilingWolf, can you also indicates the encoding speed, it don't need to be in fps as it can be in relation of a specific encoder, is more to see the variation in the speed of aoemc and rav1e due to optimizations or other improvements.
I use my PC under various loads while encoding, so I can't reliably measure times, which is why I haven't included anything time specific so far.
A quick test on the F.Y.C clip however has displayed an improvement, on average, of 12% over the last month (1.0.0-1058-g8547359cf vs 1.0.0-908-g3a607f7b0).
At the same time, quality as measured by MS-SSIM and PSNR-HVS-M has gone down ~2.5%, circa to the levels of 3 months ago (1.0.0-577-g8ae39302e).

As for rav1e, the times have only grown worse because there have been many new additions but almost no early breakout (or none at all?) strategies have been implemented yet.
And as of right now, some ASM optimizations are disabled on Windows, so I guess I'm basically running on Rust code. So not much of a speedup on that front either (yet).

Last edited by SmilingWolf; 19th December 2018 at 11:23.
SmilingWolf is offline   Reply With Quote
Old 20th December 2018, 14:52   #1335  |  Link
Phanton_13
Registered User
 
Join Date: May 2002
Posts: 95
Thanks SmilingWolf for the info.
Phanton_13 is offline   Reply With Quote
Old 22nd December 2018, 11:31   #1336  |  Link
sneaker_ger
Registered User
 
Join Date: Dec 2002
Posts: 5,565
Simply crashes on my system even with a simple "ffmpeg" (no parameters). Windows 7 x64, i5-2500K (SSE4.2 and AVX).
sneaker_ger is offline   Reply With Quote
Old 22nd December 2018, 17:47   #1337  |  Link
sneaker_ger
Registered User
 
Join Date: Dec 2002
Posts: 5,565
Still crashes.
sneaker_ger is offline   Reply With Quote
Old 22nd December 2018, 18:40   #1338  |  Link
lvqcl
Registered User
 
Join Date: Aug 2015
Posts: 293
My CPU doesn't support SSE4.2, only 4.1, but I still tried to run it.
ffmpeg.exe crashes at shlx instruction, which is part of BMI2 (Haswell/Excavator).
lvqcl is offline   Reply With Quote
Old 22nd December 2018, 19:11   #1339  |  Link
nevcairiel
Registered Developer
 
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,342
Its usually not a good idea to build such a restricted binary, nor does it really give a meaningful speed enhancement. Just use a better build.
I often intentionally restrict what I allow compilers to do, because even GCC 8 is still terrible at using advanced instruction set instructions, and it can and will cause issues left and right.

This is especially true in a large and old code base like ffmpeg, which has code that was written 15 years ago, and code that as written just now, code that was painstakingly hand-optimized, and whatnot which can result in quite varying stress on the compiler.
__________________
LAV Filters - open source ffmpeg based media splitter and decoders

Last edited by nevcairiel; 22nd December 2018 at 19:14.
nevcairiel is offline   Reply With Quote
Old 22nd December 2018, 22:01   #1340  |  Link
Wolfberry
Helenium(Easter)
 
Wolfberry's Avatar
 
Join Date: Aug 2017
Location: Hsinchu, Taiwan
Posts: 99
I think the crash comes from that I configured gcc to have skylake by default, forget to use the --cpu switch, will rebuild them later.

Thanks nevcairiel.
__________________
Monochrome Anomaly

Last edited by Wolfberry; 23rd December 2018 at 01:25.
Wolfberry 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 16:14.


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