This is an old revision of the document!

ffmpeg - Hardware transcoding using GPUs - AMD GPU

ffmpeg -i "input.mkv" -vaapi_device /dev/dri/renderD128 -vcodec hevc_vaapi -vf format='nv12|vaapi,hwupload' -map 0 -c:a copy -c:s copy -qp 23 "output.mkv"

NOTE:

-vaapi_device /dev/dri/renderD128, which is a shortcut to -init_hw_device vaapi=vaapi0:/dev/dri/renderD128 -filter_hw_device vaapi0, initializes a new hardware device of type vaapi.

-vcodec hevc_vaapi

-vf 'format=nv12,hwupload' creates the filtergraph and uses it to filter the stream.
- The encoders only accept input as VAAPI surfaces.
- If the input is in normal memory, it will need to be uploaded before giving the frames to the encoder; in the ffmpeg utility, the hwupload filter can be used for this.
- It will upload to a surface with the same layout as the software frame, so it may be necessary to add a format filter immediately before to get the input into the right format (hardware generally wants the nv12 layout).
- The hwupload filter also requires a device to upload to, which needs to be defined before the filter graph is created, which is done with -vaapi_device.

-qp 23: The video quality.
- The lower the value, the better will be the video quality, but also the files will be larger.
- The default value is 25 which is fairly good.
- For high-definition videos a value between 21-23 works best and produces good results.

NOTE: To figure out for which codecs GPU hardware video encoding acceleration can be used vainfo command provides some information:

vainfo

returns:

vainfo
libva info: VA-API version 1.16.0
libva info: User environment variable requested driver 'radeonsi'
libva info: Trying to open /opt/amdgpu/lib/x86_64-linux-gnu/dri/radeonsi_drv_video.so
libva info: Found init function __vaDriverInit_1_16
libva info: va_openDriver() returns 0
vainfo: VA-API version: 1.16 (libva 2.12.0)
vainfo: Driver version: Mesa Gallium driver 23.3.0-devel for Radeon RX 7900 XTX (navi31, LLVM 17.0.4, DRM 3.54, 6.5.0-15-generic)
vainfo: Supported profile and entrypoints
      VAProfileH264ConstrainedBaseline:	VAEntrypointVLD
      VAProfileH264ConstrainedBaseline:	VAEntrypointEncSlice
      VAProfileH264Main               :	VAEntrypointVLD
      VAProfileH264Main               :	VAEntrypointEncSlice
      VAProfileH264High               :	VAEntrypointVLD
      VAProfileH264High               :	VAEntrypointEncSlice
      VAProfileHEVCMain               :	VAEntrypointVLD
      VAProfileHEVCMain               :	VAEntrypointEncSlice
      VAProfileHEVCMain10             :	VAEntrypointVLD
      VAProfileHEVCMain10             :	VAEntrypointEncSlice
      VAProfileJPEGBaseline           :	VAEntrypointVLD
      VAProfileVP9Profile0            :	VAEntrypointVLD
      VAProfileVP9Profile2            :	VAEntrypointVLD
      VAProfileAV1Profile0            :	VAEntrypointVLD
      VAProfileAV1Profile0            :	VAEntrypointEncSlice
      VAProfileNone                   :	VAEntrypointVideoProc

For all entries with a value of VAEntrypointEncSlice hardware encoding can be used.

AMD GPUs

h264_amf
hevc_amf

Decoding

-hwaccel d3d11va

Encoding

For video resolution <= 4K

-c:v libx264
 
or
 
-c:v h264_amf

For video resolution > 4K

-c:v libx265 
 
or
 
-c:v hevc_amf

Examples

#1 - A corrupt 1920×1080 video:

ffmpeg -y -vsync 0 -i "input.mp4" -c:a copy -c:v h264_amf -pix_fmt yuv420p "output.mp4"

#2 - a 7680×3840 H.265 video with incompatible profile for VHT - a stutter playback 7260×3630 60fps 10bit video

ffmpeg -y -vsync 0 -i "input.mp4" -c:a copy -c:v hevc_amf -pix_fmt yuv420p "output.mp4"

#3 - a 7680×7680 video not hardware decoded because is H.264 encoded

ffmpeg -y -vsync 0 -i "input.mp4" -c:a copy -c:v hevc_amf -pix_fmt yuv420p "output.mp4"

Wiki

Table of Contents