MSU MPEG-4 AVC/H.264 2012 Video Codec Comparison - CALL FOR CODECS

[Dyomich]

Dear video codec developers,

Moscow State University Graphics & Multimedia Laboratory
starts next eighth annual H.264 video codecs comparison.

New this year features:

• GPU codecs will be analyzed:Nvidia, AMD, and Ivy Bridge based encoders.
  This year we want to make GPU-encoders analysis more
  interesting and complex. For this analysis we plan to use top GPUs,
  e.g. Nvidia and AMD and also to analyze Intel Ivy Bridge (Intel HD
  Graphics 4000) based encoder. Developers can send us
  GPU-accelerated or GPU-based encoder with or without CPU-based
  encoder!
  
• We plan to include new codecs that did not participate in previous
  comparison by choosing presets for them ourself.
  For that task we will use option analysis. So we will be glad to
  have a direct contact with codec developers. The main benefit of
  direct participation for developers is that developer could help us
  to choose correct settings for better results and additionally
  developers receive Enterprise version of comparison free of charge.
  
• Videoconference encoding analysis will be more complex comparing
  to previous comparison.
  
• New video sequences including professional video sequences for video codec analysis

Important Dates
February, 17 - Deadline for receipt of a H.264 codec with presets
February, 28 - Deadline for settling technical problems with codec
April, 24 - Report's draft will be sent to all participants
May, 5 - Deadline for reception of comments to the draft
May, 15 - Comparison report release

The full text of Call for Codecs is available at
http://compression.ru/video/codec_co...codecs_12.html

Variants of Participation
There are two variants for companies to participate in our comparison:

1. Participation for free. All results of your codec will be
   published, except special cases of measurements
   problems due to codec instability.
   
2. Private participation. A special report will be prepared only
   for your company. This report contains:

   • Your codec results and all material from the free version
   • Special additional analysis of your codec

If you are interested in the private participation,
please contact us for details.

Useful Links

• Call for codecs-2011
  http://compression.ru/video/codec_co...codecs_12.html
  
• Seventh Annual MSU MPEG-4 AVC/H.264 Video Codec Comparison
  http://www.compression.ru/video/code...ison/h264_2011
  
• Options Analysis of MPEG-4 AVC/H.264 Codec x264
  http://compression.ru/video/codec_co...sis_08_en.html

-----
Best regards,
Dr. Dmitriy Kulikov,
Moscow State University (MSU)
Graphics&Media Lab
Videocodec Testing Team
videocodec-testing@graphics.cs.msu.ru

[CruNcher]

Finaly the GPU (Asic,fixed,dynamic function) compare
I guess you will take Mainconcepts GPU ENcoder and Elemental's also into account, though i guess Arcsoft with their own GPU Encoder (Research they had 3 different Encoder implemented 1 was entirely Cuda based though bad quality most probably Mobile optimized just for pure speed, The other was the Pure NVCuvenc (Nvidia implementation) and then they had another one based on X264 implemented with partly GPU support) isn't on the list yet you should ask them if they want to participate

[IgorC]

MSU have published the report
http://compression.ru/video/codec_comparison/h264_2012/

[CruNcher]

Without reading it i guess Nvidia won on the Quality Encoder Part of the DSP (Kepler) and Intel comes second (Ivy Bridge), both yet not able to reach x264 @ even superfast (on a newer Generation 4 Core Intel CPU (SB 32nm or Ivy 22nm ) with some optimizations (@ least subme 1/2) though most probably both DSP drawing less power overall for the given result (letting Nvidia win the Overall Balance of consumption,speed and quality) (Related from my own testing of their Encoder base throughout Development)

Nvidia didn't participated with their new DSP Encoder Core (in Kepler GPUs) only through Mainconcepts Cuda (Shader) Encoder (on the older 580 GTX) ?

also AMD seems not interested to show any of their Improvements (that's very surprising) ?

So in the end nothing really new from this test I've would have expected some more to be honest though i guess everyone is in full H.265 preperation

But @ least getting results from the newer Nvidia DSP Core (Nvidia could have put you under NDA) instead of Mainconcepts Shader Core would have been nice, you get these results practically already everywhere else on the web now.

[iwod]

I am actually surprised by the Quality of Intel QuickSync for the speed it offers, but even then it is still far off to x264.

[CruNcher]

The Speed is not the interesting thing the difference in Power Consumption is

i made a small test on that though these are in the wild shots not real scientific but i came to the following though this is also based on Nvidias old Shader Encoder not the new Kepler one and i came to the following result @ comparable quality on a 4 Core Intel SB system.

all Decoding CPU all overall system Power Consumption (out of the wall measurement) all using CABAC High Level 3.1

Quicksync (SB I5-2400 not S) = 110W
x264 (with some preset optimizations) = 130W (though im confident it could be lowered and would stilll look as good with some deeper source changes)
Nvidia 460 GTX (full/partial encoding) = 180W (full and partial encoding differ in the final visual result)

all reaching same Speed of around AVG 180 fps @ 720p 30 fps 3 mbit

though obviously more interesting it becomes the Higher the Bitrate gets and CABAC drives on