[Cin] VAAPI supported Encoding formats
Andrew Randrianasulu
randrianasulu at gmail.com
Sun Nov 3 00:09:33 CET 2024
On Sat, Nov 2, 2024 at 8:15 PM Terje J. Hanssen via Cin <
cin at lists.cinelerra-gg.org> wrote:
>
> Den 01.11.2024 19:34, skrev Phyllis Smith:
>
> Terje, thank you for this comprehensive list of vaapi supported encodings.
>
> On Thu, Oct 31, 2024 at 5:38 PM Terje J. Hanssen via Cin <
> cin at lists.cinelerra-gg.org> wrote:
>
>> In hope it may be of some use for Cingg next, I make a summary and
>> supplement to the guide in
>> https://trac.ffmpeg.org/wiki/Hardware/VAAPI#Encoding
>> And instead of adding a tail to already too long recent threads, I start
>> a new thread here.
>>
>> Two additional references are used
>> Developer Reference: Media Capabilities Supported by Intel Hardware
>>
>> https://www.intel.com/content/www/us/en/docs/onevpl/developer-reference-media-intel-hardware/1-1/features-and-formats.html
>> Supported Encoding Input Format and Max Resolution
>>
>> https://github.com/intel/media-driver/blob/master/docs/media_features.md#supported-encoding-input-format-and-max-resolution
>>
>> to clarify the Encode Features for my own Intel Arc A-series Discrete
>> Graphics
>>
>> HEVC (fixed-function hardware + shader-based)
>> 8-bit, 4:2:0 (NV12)
>> 10-bit, 4:2:0 (P010)
>> 8-bit, 4:2:2 (YUY2, Windows only)
>> 10-bit, 4:2:2 (Y210)
>>
>> AV1
>> 8-bit, 4:2:0 (NV12)
>> 10-bit, 4:2:0 (P010)
>>
>>
>> A short reminder:
>> 8-bit vs 10-bit color depths is 256 vs 1024 color levels of each RGB
>> color, or 16.7 mill. vs 1.0 bill colors.
>> 4:2:0 vs 4:2:2 color sub-sampling is keeping 1/4 vs 1/2 of all 4:4:4
>> colors or the precision of color compression.
>>
>> https://fujifilm-x.com/en-us/series/the-filmmakers-handbook/8-bit-or-10-bit-video-color-explained/
>>
>>
>> An important point to notice is that the VAAPI encoding 'formats' are
>> based on those in the parentheses () above.
>>
>> Three simplified ffmpeg code lines with default bit-rates and without
>> audio are used in the following three examples.
>> Resulting fps encoding counts or speeds are the same in all cases; I
>> don't know why.
>>
>>
>> The input file used:
>>
>> ffprobe -hide_banner cfhd01.mkv
>> Input #0, matroska,webm, from 'cfhd01.mkv':
>> Metadata:
>> ENCODER : Lavf61.1.100
>> Duration: 00:01:11.24, start: 0.000000, bitrate: 393480 kb/s
>> Stream #0:0(eng): Video: cfhd (CFHD / 0x44484643), yuv422p10le(tv,
>> bt709, top coded first (swapped)), 1920x1080, SAR 1:1 DAR 16:9, 25 fps, 25
>> tbr, 1k tbn (default)
>> Metadata:
>> HANDLER_NAME : Apple Video Media Handler
>> VENDOR_ID : appl
>> ENCODER : Lavc61.3.100 cfhd
>> DURATION : 00:01:11.240000000
>>
>>
>> 1)
>> 8-bit yuv420p hevc vaapi encoding using -vf 'format=nv12,hwupload'
>> ------------------------------------------------------------------
>> ffmpeg -hide_banner -vaapi_device /dev/dri/renderD128 -i cfhd01.mkv -vf
>> 'format=nv12,hwupload' -c:v av1_vaapi cfhd01_av1_vaapi_format_nv12.mp4
>> ............
>> frame= 1781 fps= 46 q=-0.0 Lsize= 104132KiB time=00:01:11.24
>> bitrate=11974.3kbits/s speed=1.83x
>>
>>
>> ffprobe -hide_banner cfhd01_av1_vaapi_format_nv12.mp4
>> Input #0, mov,mp4,m4a,3gp,3g2,mj2, from
>> 'cfhd01_av1_vaapi_format_nv12.mp4':
>> Metadata:
>> major_brand : isom
>> minor_version : 512
>> compatible_brands: isomav01iso2mp41
>> encoder : Lavf61.7.100
>> Duration: 00:01:11.24, start: 0.000000, bitrate: 11974 kb/s
>> Stream #0:0[0x1](eng): Video: av1 (libdav1d) (Main) (av01 /
>> 0x31307661), yuv420p(tv, bt709, progressive), 1920x1080, 11973 kb/s, SAR
>> 1:1 DAR 16:9, 25 fps, 25 tbr, 12800 tbn (default)
>> Metadata:
>> handler_name : Apple Video Media Handler
>> vendor_id : [0][0][0][0]
>> encoder : Lavc61.19.100 av1_vaapi
>>
>>
>> 2)
>> 10-bit yuv420p10le hevc vaapi encoding using -vf 'format=p010,hwupload'
>> -----------------------------------------------------------------------
>> ffmpeg -hide_banner -vaapi_device /dev/dri/renderD128 -i cfhd01.mkv -vf
>> 'format=p010,hwupload' -c:v av1_vaapi cfhd01_av1_vaapi_format_p010.mp4
>> .............
>> frame= 1781 fps= 46 q=-0.0 Lsize= 111924KiB time=00:01:11.24
>> bitrate=12870.3kbits/s speed=1.82x
>>
>>
>> ffprobe -hide_banner cfhd01_av1_vaapi_format_p010.mp4
>> Input #0, mov,mp4,m4a,3gp,3g2,mj2, from
>> 'cfhd01_av1_vaapi_format_p010.mp4':
>> Metadata:
>> major_brand : isom
>> minor_version : 512
>> compatible_brands: isomav01iso2mp41
>> encoder : Lavf61.7.100
>> Duration: 00:01:11.24, start: 0.000000, bitrate: 12870 kb/s
>> Stream #0:0[0x1](eng): Video: av1 (libdav1d) (Main) (av01 /
>> 0x31307661), yuv420p10le(tv, bt709, progressive), 1920x1080, 12869 kb/s,
>> SAR 1:1 DAR 16:9, 25 fps, 25 tbr, 12800 tbn (default)
>> Metadata:
>> handler_name : Apple Video Media Handler
>> vendor_id : [0][0][0][0]
>> encoder : Lavc61.19.100 av1_vaapi
>>
>> 3)
>> 10-bit yuv422p10le hevc_vaapi encoding using -vf 'format=y210,hwupload'
>> -----------------------------------------------------------------------
>> ffmpeg -hide_banner -vaapi_device /dev/dri/renderD128 -i cfhd01.mkv -vf
>> 'format=y210,hwupload' -c:v hevc_vaapi cfhd01_hevc_vaapi_format_y210.mp4
>> frame= 1781 fps= 46 q=-0.0 Lsize= 105482KiB time=00:01:11.20
>> bitrate=12136.3kbits/s speed=1.83x
>>
>>
>> ffprobe -hide_banner cfhd01_hevc_vaapi_format_y210.mp4
>> Input #0, mov,mp4,m4a,3gp,3g2,mj2, from
>> 'cfhd01_hevc_vaapi_format_y210.mp4':
>> Metadata:
>> major_brand : isom
>> minor_version : 512
>> compatible_brands: isomiso2mp41
>> encoder : Lavf61.7.100
>> Duration: 00:01:11.24, start: 0.000000, bitrate: 12129 kb/s
>> Stream #0:0[0x1](eng): Video: hevc (Rext) (hev1 / 0x31766568),
>> yuv422p10le(tv, bt709, progressive), 1920x1080 [SAR 1:1 DAR 16:9], 12127
>> kb/s, 25 fps, 25 tbr, 12800 tbn (default)
>> Metadata:
>> handler_name : Apple Video Media Handler
>> vendor_id : [0][0][0][0]
>> encoder : Lavc61.19.100 hevc_vaapi
>>
>>
>>
>
> It was also interesting to standardize that the corresponding ffmpeg
> syntax above works for both vaapi/qsv and hevc/av1:
>
> - replace vaapi with qsv in 1) 2) and 3) above
> - and if applicable: replace hevc with av1 in 1) and 2) above
>
>
> The default bitrates used do vary a lot as seen on listed output file
> sizes:
>
> 19M cfhd01_hevc_qsv_format_nv12.mp4
> 92M cfhd01_hevc_vaapi_format_nv12.mp4
>
> 224M cfhd01_hevc_qsv_format_p010.mp4
> 90M cfhd01_hevc_vaapi_format_p010.mp4
>
> 270M cfhd01_hevc_qsv_format_y210.mp4
> 104M cfhd01_hevc_vaapi_format_y210.mp4
> ----------------
>
> 9,1M cfhd01_av1_qsv_format_nv12.mp4
> 102M cfhd01_av1_vaapi_format_nv12.mp4
>
> 9,1M cfhd01_av1_qsv_format_p010.mp4
> 110M cfhd01_av1_vaapi_format_p010.mp4
>
>
>
yes, last four lines seem say that qsv and vaapi AV1 encoders use * wildly*
different defaults for bitrate ....
>
>
>
> --
> Cin mailing list
> Cin at lists.cinelerra-gg.org
> https://lists.cinelerra-gg.org/mailman/listinfo/cin
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cinelerra-gg.org/pipermail/cin/attachments/20241103/f3bf3647/attachment-0001.htm>
More information about the Cin
mailing list