Docker Nvidia Hardware acceleration GPU can be seen but isnt being utilised

[ck3mp 0]

Hi,

Been using emby a while and love it. On my Archlinux server I host it inside a docker container. However I have just added a Nvidia 1050 GPU to help solve some issue I am having with on the fly stream transcoding.

Emby can see the GPU(s) as they are present here in the transcode options.

 

But when watching a film...

 

This is my docker-compose file:

version: '3.3'
services:
  embyserver:
    container_name: embyserver
    runtime: nvidia
    volumes:
      - /opt/embyserver:/config
      - /media/raid_2:/media
    ports:
      - '8096:8096'
      - '8920:8920'
    environment:
      - UID=1000
      - GID=1000
      - GIDLIST=986,989
    image: 'emby/embyserver:latest'
    restart: unless-stopped

I have ran basic tests with nvidia-smi inside containers and the GPU can be accessed from those so I'm not sure its my docker setup?

services:
  test:
    image: nvidia/cuda
    command: nvidia-smi
    runtime: nvidia

 

I am also unable to access the logs so I am now a bit sumped where to turn!

 

Thanks in advance for any advice... happy to share logs if anyone can help me solve that issue!!

 

 

[ck3mp 0]

More information... I seem to have HW accel working fine in an FFMPEG container... these commands both trigger GPU usage under nvidia-smi

 

Source file:
Complete name                            : /media/test.mkv
Format                                   : Matroska
Format version                           : Version 4
File size                                : 31.2 GiB
Duration                                 : 3 h 1 min
Overall bit rate mode                    : Variable
Overall bit rate                         : 24.7 Mb/s
Encoded date                             : UTC 2019-08-04 00:42:56
Writing application                      : mkvmerge v35.0.0 ('All The Love In The World') 64-bit
Writing library                          : libebml v1.3.9 + libmatroska v1.5.2

Video
ID                                       : 1
Format                                   : HEVC
Format/Info                              : High Efficiency Video Coding
Format profile                           : Main 10@L5.1@High
Codec ID                                 : V_MPEGH/ISO/HEVC
Duration                                 : 3 h 1 min
Bit rate                                 : 19.2 Mb/s
Width                                    : 3 840 pixels
Height                                   : 1 608 pixels
Display aspect ratio                     : 2.40:1
Frame rate mode                          : Constant
Frame rate                               : 23.976 (24000/1001) FPS
Color space                              : YUV
Chroma subsampling                       : 4:2:0 (Type 1)
Bit depth                                : 10 bits
Bits/(Pixel*Frame)                       : 0.129
Stream size                              : 24.3 GiB (78%)
Title                                    : FFlipped @ HDChina
Default                                  : Yes
Forced                                   : No
Color range                              : Limited
Color primaries                          : BT.2020
Transfer characteristics                 : PQ
Matrix coefficients                      : BT.2020 non-constant

 

Encode to H264 - Works
docker run --runtime=nvidia -v /media:/media jrottenberg/ffmpeg:4.1-nvidia -i /media/test.mkv -c:v h264_nvenc -vf format=yuv420p test.mkv

Decode with CUDA - Works
docker run --runtime=nvidia -v /media:/media jrottenberg/ffmpeg:4.1-nvidia -hwaccel cuda -i /media/test.mkv test.mkv

 

[Luke 40131]

Hi, I would suggest taking a look at this:

 

[ck3mp 0]

As per your comment in that thread moving to the Beta has solved my issue...

Thanks,