Quadro P400 Ubuntu 22.04 FFMpeg issues

lewiswalsh

I'm not sure what I'm doing wrong. ersatztv:latest-nvidia using docker compose. All working as expected, except hardware acceleration. I have nvenc chosen in FFMPeg Profiles, and there is tick next to hardware acceleration on the front page. Latest nvidia drivers install (565). nvidia-smi shows the following:

+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 560.35.03              Driver Version: 560.35.03      CUDA Version: 12.6     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  Quadro P400                    Off |   00000000:2A:00.0 Off |                  N/A |
| 34%   39C    P8             N/A /  N/A  |      14MiB /   2048MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI        PID   Type   Process name                              GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|    0   N/A  N/A      1600      G   /usr/lib/xorg/Xorg                              8MiB |
|    0   N/A  N/A      2281      G   /usr/bin/gnome-shell                            1MiB |
+-----------------------------------------------------------------------------------------+

However, in the troubleshooting section for Nvidia I see the following:

ffmpeg version 6.1 Copyright (c) 2000-2023 the FFmpeg developers
  built with gcc 9 (Ubuntu 9.4.0-1ubuntu1~20.04.2)
  configuration: --ld=clang++ --disable-debug --disable-doc --disable-ffplay --enable-ffprobe --enable-cuda --enable-cuda-llvm --enable-cuvid --enable-fontconfig --enable-gpl --enable-libaom --enable-libdav1d --enable-libass --enable-libfdk_aac --enable-libfreetype --enable-libkvazaar --enable-libmp3lame --enable-libnpp --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenjpeg --enable-libopus --enable-libsrt --enable-libtheora --enable-libv4l2 --enable-libvorbis --enable-libvpx --enable-libwebp --enable-libxml2 --enable-libx264 --enable-libx265 --enable-libxvid --enable-nonfree --enable-nvdec --enable-nvenc --enable-opencl --enable-openssl --enable-small --enable-stripping --enable-vaapi --enable-vdpau --enable-version3 --extra-cflags='-I/tmp/ffnvcodec/include/ffnvcodec -I/usr/local/cuda/include/' --extra-ldflags='-L/usr/local/cuda/lib64 -L/usr/local/cuda/lib32/' --extra-libs=-ldl --extra-libs=-lpthread
  libavutil      58. 29.100 / 58. 29.100
  libavcodec     60. 31.102 / 60. 31.102
  libavformat    60. 16.100 / 60. 16.100
  libavdevice    60.  3.100 / 60.  3.100
  libavfilter     9. 12.100 /  9. 12.100
  libswscale      7.  5.100 /  7.  5.100
  libswresample   4. 12.100 /  4. 12.100
  libpostproc    57.  3.100 / 57.  3.100
Input #0, lavfi, from 'nullsrc':
  Duration: N/A, start: 0.000000, bitrate: N/A
  Stream #0:0: Video: wrapped_avframe, yuv420p, 320x240 [SAR 1:1 DAR 4:3], 25 fps, 25 tbr, 25 tbn
Stream mapping:
  Stream #0:0 -> #0:0 (wrapped_avframe (native) -> h264 (h264_nvenc))
Press [q] to stop, [?] for help
[h264_nvenc @ 0x1b2bb800] Cannot load libcuda.so.1
[vost#0:0/h264_nvenc @ 0x1b2bb3c0] Error while opening encoder - maybe incorrect parameters such as bit_rate, rate, width or height.
Error while filtering: Operation not permitted
[out#0/null @ 0x1b2ba0c0] Nothing was written into output file, because at least one of its streams received no packets.
frame=    0 fps=0.0 q=0.0 Lsize=       0kB time=N/A bitrate=N/A speed=N/A    
Conversion failed!

jason

How are you starting the container? You will need to explicitly give the container access to the GPU, and it's a little different depending on how you run the container.

lewiswalsh

Thanks for your reply. I'm docker compose on ubuntu 20.04

jason

See the environment and deploy entries from the nvidia docker-compose here https://github.com/ErsatzTV/ErsatzTV/blob/main/docker/docker-compose.nvidia.yml - that might get you going.

lewiswalsh

Thanks but I'm still stumped. I've tried all sorts of things but still get this:

ffmpeg version 6.1 Copyright (c) 2000-2023 the FFmpeg developers
  built with gcc 9 (Ubuntu 9.4.0-1ubuntu1~20.04.2)
  configuration: --ld=clang++ --disable-debug --disable-doc --disable-ffplay --enable-ffprobe --enable-cuda --enable-cuda-llvm --enable-cuvid --enable-fontconfig --enable-gpl --enable-libaom --enable-libdav1d --enable-libass --enable-libfdk_aac --enable-libfreetype --enable-libkvazaar --enable-libmp3lame --enable-libnpp --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenjpeg --enable-libopus --enable-libsrt --enable-libtheora --enable-libv4l2 --enable-libvorbis --enable-libvpx --enable-libwebp --enable-libxml2 --enable-libx264 --enable-libx265 --enable-libxvid --enable-nonfree --enable-nvdec --enable-nvenc --enable-opencl --enable-openssl --enable-small --enable-stripping --enable-vaapi --enable-vdpau --enable-version3 --extra-cflags='-I/tmp/ffnvcodec/include/ffnvcodec -I/usr/local/cuda/include/' --extra-ldflags='-L/usr/local/cuda/lib64 -L/usr/local/cuda/lib32/' --extra-libs=-ldl --extra-libs=-lpthread
  libavutil      58. 29.100 / 58. 29.100
  libavcodec     60. 31.102 / 60. 31.102
  libavformat    60. 16.100 / 60. 16.100
  libavdevice    60.  3.100 / 60.  3.100
  libavfilter     9. 12.100 /  9. 12.100
  libswscale      7.  5.100 /  7.  5.100
  libswresample   4. 12.100 /  4. 12.100
  libpostproc    57.  3.100 / 57.  3.100
Input #0, lavfi, from 'nullsrc':
  Duration: N/A, start: 0.000000, bitrate: N/A
  Stream #0:0: Video: wrapped_avframe, yuv420p, 320x240 [SAR 1:1 DAR 4:3], 25 fps, 25 tbr, 25 tbn
Stream mapping:
  Stream #0:0 -> #0:0 (wrapped_avframe (native) -> h264 (h264_nvenc))
Press [q] to stop, [?] for help
[h264_nvenc @ 0x17963800] Cannot load libcuda.so.1
[vost#0:0/h264_nvenc @ 0x179633c0] Error while opening encoder - maybe incorrect parameters such as bit_rate, rate, width or height.
Error while filtering: Operation not permitted
[out#0/null @ 0x179620c0] Nothing was written into output file, because at least one of its streams received no packets.
frame=    0 fps=0.0 q=0.0 Lsize=       0kB time=N/A bitrate=N/A speed=N/A    
Conversion failed!

My docker-compose:

services:
  ersatztv:
    privileged: true
    container_name: ersatztv
    environment:
      - TZ=Europe/London
      - NVIDIA_VISIBLE_DEVICES=all
      - NVIDIA_DRIVER_CAPABILITIES=compute,utility,video
    ports:
      - '8409:8409'
    volumes:
      - '/local/general/.dockerdata/ersatztv:/root/.local/share/ersatztv'
      - '/local/media/Media:/mnt/media:ro'
      - '/local/media/YouTube-Archive:/mnt/yt:ro'
    restart: unless-stopped
    image: 'jasongdove/ersatztv:latest-nvidia'
    devices:
      - /dev/dri:/dev/dri

nvidia-smi:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.256.02   Driver Version: 470.256.02   CUDA Version: 11.4     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Quadro P400         Off  | 00000000:2A:00.0 Off |                  N/A |
| 34%   40C    P8    N/A /  N/A |     12MiB /  1999MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A      1627      G   /usr/lib/xorg/Xorg                  8MiB |
|    0   N/A  N/A      2200      G   /usr/bin/gnome-shell                1MiB |
+-----------------------------------------------------------------------------+

ixia

I had a similar issue setting up with a P4. After your cuda toolkit and potential datacenter driver install, and after your graphics driver installation (which in my case was horribly outdated to 470 and had to be manually purged), you need to also install the Nvidia Container Toolkit (not documented in the datacenter setup) to allow the vGPUs into the container:

https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html

Assuming you have run your installs correctly, you should be rewarded with an nvidia-smi from the host that shows the vGPU in use for ffmpeg.

Hope that helps.

*EDIT: I see yours also defaulted to 470. You NEED to purge this as even with the correct container toolkit in place, 470 will crash ffmpeg regularly.