Cant make nvenc work for handbrake docker

[aronmgv]

Hey,

If I may ask for your help here..

I have successfully configured nvidia-docker for my graphic card to be used in docker containers - so my Plex is doing HW transcoding in docker. However I am struggling to get handbrake doing encoding via GPU.

My GPU is Nvidia Quadro P2200. Per https://developer.nvidia.com/video-encode-decode-gpu-support-matrix it has nvenc chip..

So my setup is: Debian 10 Nvidia Quadro P2200 Installed packages:

• nvidia-detect
• nvidia-driver
• nvidia-cuda-toolkit
• nvidia-container-runtime
• nvidia-container-toolkit

docker-compose.yml:

version: "3.7"

  handbrake:
    container_name: plex.529.handbrake.90.91
    image: jlesage/handbrake
    restart: always
    ports:
      - 52990:5800
      - 52991:5900
    expose:
      - 5800
      - 5900
    volumes:
      - /lvm/raid/private/media:/plex/media
      - /lvm/raid/private/data:/plex/data
      - $PWD/handbrake/config:/config:rw
      - $PWD/handbrake/storage:/storage:ro
    environment:
      - TZ=Europe/Prague
      - USER_ID=1000
      - GROUP_ID=994
      - NVIDIA_VISIBLE_DEVICES=GPU-d705d61e-a848-a592-e014-d196cd66eec9
      - NVIDIA_DRIVER_CAPABILITIES=compute,video,utility
      - DISPLAY_WIDTH=1600
      - DISPLAY_HEIGHT=900

/etc/docker/daemon.json:

{
  "default-runtime": "nvidia",
  "runtimes": {
     "nvidia": {
       "path": "/usr/bin/nvidia-container-runtime",
       "runtimeArgs": []
       }
     }
}

Thanks, Michal

EDIT: I installed Handbrake directly on host to see if can choose nvenc profile, but same result - it is not listed there :/

[zocker-160]

@MacGyver27 first of all I am really really sorry for this super late reply, I didn't get any notification that an issue got opened :-((

the native Handbrake version that comes with the debian repo does not have NVENC support enabled at least I couldn't get it to work either. That's why the docker container is recommended

your docker compose setup above doesn't have NVENC enabled either, because you are using the wrong image: image: jlesage/handbrake doesn't have NVENC support, which is the reason why I created my repo over here

have you tried it with my docker image? It should work using my instructions

[zocker-160]

it should look like this

[aronmgv]

Sorry my bad, I pasted old docker-compose.yml file there. I was building from your github repo:

version: "3.7"

services:
  handbrake:
    container_name: handbrake.539
    build: ./handbrake-nvenc-docker
    restart: always
    ports:
      - 53980:5800
      - 53981:5900
    expose:
      - 5800
      - 5900
    volumes:
      - $PWD/config:/config:rw
      - $PWD/storage:/storage:ro
    environment:
      - TZ=Europe/Prague
      - USER_ID=1000
      - GROUP_ID=994
      - NVIDIA_VISIBLE_DEVICES=GPU-d705d61e-a848-a592-e014-d196cd66eec9
      - NVIDIA_DRIVER_CAPABILITIES=compute,video,utility
      - DISPLAY_WIDTH=1600
      - DISPLAY_HEIGHT=900

Using Cuda from the repo for Debian 10 - which is 9.2:

nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2018 NVIDIA Corporation
Built on Tue_Jun_12_23:07:04_CDT_2018
Cuda compilation tools, release 9.2, V9.2.148

And still cant see it 🤔 :

Thank you for your patience, Michal

[zocker-160]

@MacGyver27 do you have nvidia-docker2 installed?

does this command work for you? docker run --rm --runtime nvidia nvidia/cuda:9.0-base nvidia-smi

I sadly coudln't replicate your problem, it works for me using those files: I did change yours a bit docker-compose.yml:

version: "3.3"

services:
  handbrake:
    container_name: handbrake.539
    image: zocker160/handbrake-nvenc:latest
    restart: always
    ports:
      - 53980:5800
      - 53981:5900
    expose:
      - 5800
      - 5900
    volumes:
      - $PWD/test:/config:rw
      - $PWD/test:/storage:ro
    environment:
      - TZ=Europe/Berlin
      - USER_ID=1000
      - GROUP_ID=994
      - NVIDIA_VISIBLE_DEVICES=all
      - NVIDIA_DRIVER_CAPABILITIES=all
      - DISPLAY_WIDTH=1600
      - DISPLAY_HEIGHT=900

and /etc/docker/daemon.json:

{
    "default-runtime": "nvidia",
    "runtimes": {
        "nvidia": {
            "path": "nvidia-container-runtime",
            "runtimeArgs": []
        }
    }
}

[aronmgv]

Per the original github repo nvidia-docker2 is deprecated and it was replaced by nvidia-container-toolkit. In the first post I mention what I have installed:

• nvidia-detect
• nvidia-driver
• nvidia-cuda-toolkit
• nvidia-container-runtime
• nvidia-container-toolkit

Yes, that command works:

docker run --rm --runtime nvidia nvidia/cuda:9.0-base nvidia-smi
Fri Feb 14 12:38:42 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.74       Driver Version: 418.74       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Quadro P2200        On   | 00000000:0B:00.0  On |                  N/A |
| 51%   44C    P0    21W /  75W |     68MiB /  5055MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+

Now I copied your docker-compose.yml file and still the same :/ Any other ideas maybe?? :)

EDIT: Hmmm, why in that command output it says CUDA Version: 10.1 when on host I have Cuda compilation tools, release 9.2, V9.2.148 ?

[zocker-160]

Per the original github repo nvidia-docker2 is deprecated and it was replaced by nvidia-container-toolkit yes right, that's why I was asking, it might be problematic, when having both installed at the same time

have you copied my daemon.json as well? and then use sudo systemctl daemon-reload sudo systemctl restart docker to reload it

maybe you should try this command here, to see if the problem is docker-compose or not

docker run --rm -t \
    --name=handbrake \
    --gpus all \
    -p 53980:5800 \
    -p 53981:5900 \
    zocker160/handbrake-nvenc:latest

Hmmm, why in that command output it says CUDA Version: 10.1 when on host I have Cuda compilation tools, release 9.2, V9.2.148 ?

the one doesn't have anything to do with the other, because NVCC is only for compiling CUDA software (compilation tools) and the other is the CUDA version you have installed via the GPU driver

[aronmgv]

have you copied my daemon.json as well? and then use sudo systemctl daemon-reload sudo systemctl restart docker to reload it

Did it now and run it via your docker run command and still the same.. Here is the output from it (if it helps):

cat /etc/docker/daemon.json
{
    "default-runtime": "nvidia",
    "runtimes": {
        "nvidia": {
            "path": "nvidia-container-runtime",
            "runtimeArgs": []
        }
    }
}

sudo systemctl daemon-reload
sudo systemctl restart docker

docker run --rm -t \
>     --name=handbrake \
>     --gpus all \
>     -p 53980:5800 \
>     -p 53981:5900 \
>     zocker160/handbrake-nvenc:latest
[s6-init] making user provided files available at /var/run/s6/etc...exited 0.
[s6-init] ensuring user provided files have correct perms...exited 0.
[fix-attrs.d] applying ownership & permissions fixes...
[fix-attrs.d] done.
[cont-init.d] executing container initialization scripts...
[cont-init.d] 00-app-niceness.sh: executing...
[cont-init.d] 00-app-niceness.sh: exited 0.
[cont-init.d] 00-app-script.sh: executing...
[cont-init.d] 00-app-script.sh: exited 0.
[cont-init.d] 00-app-user-map.sh: executing...
[cont-init.d] 00-app-user-map.sh: exited 0.
[cont-init.d] 00-clean-logmonitor-states.sh: executing...
[cont-init.d] 00-clean-logmonitor-states.sh: exited 0.
[cont-init.d] 00-clean-tmp-dir.sh: executing...
[cont-init.d] 00-clean-tmp-dir.sh: exited 0.
[cont-init.d] 00-set-app-deps.sh: executing...
[cont-init.d] 00-set-app-deps.sh: exited 0.
[cont-init.d] 00-set-home.sh: executing...
[cont-init.d] 00-set-home.sh: exited 0.
[cont-init.d] 00-take-config-ownership.sh: executing...
[cont-init.d] 00-take-config-ownership.sh: exited 0.
[cont-init.d] 00-xdg-runtime-dir.sh: executing...
[cont-init.d] 00-xdg-runtime-dir.sh: exited 0.
[cont-init.d] 10-certs.sh: executing...
[cont-init.d] 10-certs.sh: exited 0.
[cont-init.d] 10-cjk-font.sh: executing...
[cont-init.d] 10-cjk-font.sh: exited 0.
[cont-init.d] 10-nginx.sh: executing...
[cont-init.d] 10-nginx.sh: exited 0.
[cont-init.d] 10-vnc-password.sh: executing...
[cont-init.d] 10-vnc-password.sh: exited 0.
[cont-init.d] 10-web-index.sh: executing...
[cont-init.d] 10-web-index.sh: exited 0.
[cont-init.d] 95-check-optical-drive.sh: executing...
[cont-init.d] 95-check-optical-drive.sh: looking for usable optical drives...
[cont-init.d] 95-check-optical-drive.sh: no usable optical drive found.
[cont-init.d] 95-check-optical-drive.sh: exited 0.
[cont-init.d] 95-check-qsv.sh: executing...
[cont-init.d] 95-check-qsv.sh: Processor: AMD Ryzen 9 3950X 16-Core Processor
[cont-init.d] 95-check-qsv.sh: Intel Quick Sync Video not supported: not an Intel processor.
[cont-init.d] 95-check-qsv.sh: exited 0.
[cont-init.d] handbrake.sh: executing...
[cont-init.d] handbrake.sh: core dump file location: core
[cont-init.d] handbrake.sh: core dump file size:  (blocks)
[cont-init.d] handbrake.sh: exited 0.
[cont-init.d] done.
[services.d] starting services
[services.d] starting s6-fdholderd...
[services.d] starting xvfb...
[xvfb] starting...
[services.d] starting nginx...
[services.d] starting openbox...
[nginx] starting...
[openbox] starting...
[services.d] starting logmonitor...
[services.d] starting autovideoconverter...
[logmonitor] no file to monitor: disabling service...
[services.d] starting certsmonitor...
[services.d] starting x11vnc...
[services.d] starting statusmonitor...
[services.d] starting app...
[statusmonitor] no file to monitor: disabling service...
[certsmonitor] disabling service: secure connection not enabled.
[x11vnc] starting...
[app] starting HandBrake...
14/02/2020 13:30:47 passing arg to libvncserver: -rfbport
14/02/2020 13:30:47 passing arg to libvncserver: 5900
14/02/2020 13:30:47 passing arg to libvncserver: -rfbportv6
14/02/2020 13:30:47 passing arg to libvncserver: -1
14/02/2020 13:30:47 passing arg to libvncserver: -httpportv6
14/02/2020 13:30:47 passing arg to libvncserver: -1
14/02/2020 13:30:47 passing arg to libvncserver: -desktop
14/02/2020 13:30:47 passing arg to libvncserver: HandBrake
14/02/2020 13:30:47 x11vnc version: 0.9.14 lastmod: 2015-11-14  pid: 913
14/02/2020 13:30:47 Using X display :0
14/02/2020 13:30:47 rootwin: 0x298 reswin: 0x400001 dpy: 0x6d7b0c50
14/02/2020 13:30:47
14/02/2020 13:30:47 ------------------ USEFUL INFORMATION ------------------
[services.d] done.
[autovideoconverter] starting...
[autovideoconverter] Processing watch folder '/watch'...
14/02/2020 13:30:47 X DAMAGE available on display, using it for polling hints.
14/02/2020 13:30:47   To disable this behavior use: '-noxdamage'
14/02/2020 13:30:47
14/02/2020 13:30:47   Most compositing window managers like 'compiz' or 'beryl'
14/02/2020 13:30:47   cause X DAMAGE to fail, and so you may not see any screen
14/02/2020 13:30:47   updates via VNC.  Either disable 'compiz' (recommended) or
14/02/2020 13:30:47   supply the x11vnc '-noxdamage' command line option.
14/02/2020 13:30:47 X COMPOSITE available on display, using it for window polling.
14/02/2020 13:30:47   To disable this behavior use: '-noxcomposite'
14/02/2020 13:30:47
14/02/2020 13:30:47 Wireframing: -wireframe mode is in effect for window moves.
14/02/2020 13:30:47   If this yields undesired behavior (poor response, painting
14/02/2020 13:30:47   errors, etc) it may be disabled:
14/02/2020 13:30:47    - use '-nowf' to disable wireframing completely.
14/02/2020 13:30:47    - use '-nowcr' to disable the Copy Rectangle after the
14/02/2020 13:30:47      moved window is released in the new position.
14/02/2020 13:30:47   Also see the -help entry for tuning parameters.
14/02/2020 13:30:47   You can press 3 Alt_L's (Left "Alt" key) in a row to
14/02/2020 13:30:47   repaint the screen, also see the -fixscreen option for
14/02/2020 13:30:47   periodic repaints.
14/02/2020 13:30:47 GrabServer control via XTEST.
[autovideoconverter] Watch folder '/watch' processing terminated.
14/02/2020 13:30:47
14/02/2020 13:30:47 Scroll Detection: -scrollcopyrect mode is in effect to
14/02/2020 13:30:47   use RECORD extension to try to detect scrolling windows
14/02/2020 13:30:47   (induced by either user keystroke or mouse input).
14/02/2020 13:30:47   If this yields undesired behavior (poor response, painting
14/02/2020 13:30:47   errors, etc) it may be disabled via: '-noscr'
14/02/2020 13:30:47   Also see the -help entry for tuning parameters.
14/02/2020 13:30:47   You can press 3 Alt_L's (Left "Alt" key) in a row to
14/02/2020 13:30:47   repaint the screen, also see the -fixscreen option for
14/02/2020 13:30:47   periodic repaints.
14/02/2020 13:30:47
14/02/2020 13:30:47 XKEYBOARD: number of keysyms per keycode 7 is greater
14/02/2020 13:30:47   than 4 and 51 keysyms are mapped above 4.
14/02/2020 13:30:47   Automatically switching to -xkb mode.
14/02/2020 13:30:47   If this makes the key mapping worse you can
14/02/2020 13:30:47   disable it with the "-noxkb" option.
14/02/2020 13:30:47   Also, remember "-remap DEAD" for accenting characters.
14/02/2020 13:30:47
14/02/2020 13:30:47 X FBPM extension not supported.
14/02/2020 13:30:47 X display is not capable of DPMS.
14/02/2020 13:30:47 --------------------------------------------------------
14/02/2020 13:30:47
14/02/2020 13:30:47 Default visual ID: 0x21
14/02/2020 13:30:47 Read initial data from X display into framebuffer.
14/02/2020 13:30:47 initialize_screen: fb_depth/fb_bpp/fb_Bpl 24/32/5120
14/02/2020 13:30:47
14/02/2020 13:30:47 X display :0 is 32bpp depth=24 true color
14/02/2020 13:30:47
14/02/2020 13:30:47 Listening for VNC connections on TCP port 5900
14/02/2020 13:30:47
14/02/2020 13:30:47 Xinerama is present and active (e.g. multi-head).
14/02/2020 13:30:47 Xinerama: number of sub-screens: 1
14/02/2020 13:30:47 Xinerama: no blackouts needed (only one sub-screen)
14/02/2020 13:30:47
14/02/2020 13:30:47 fb read rate: 2834 MB/sec
14/02/2020 13:30:47 fast read: reset -wait  ms to: 10
14/02/2020 13:30:47 fast read: reset -defer ms to: 10
14/02/2020 13:30:47 The X server says there are 10 mouse buttons.
14/02/2020 13:30:47 screen setup finished.
14/02/2020 13:30:47

The VNC desktop is:      88624f48ccb5:0
PORT=5900

******************************************************************************
Have you tried the x11vnc '-ncache' VNC client-side pixel caching feature yet?

The scheme stores pixel data offscreen on the VNC viewer side for faster
retrieval.  It should work with any VNC viewer.  Try it by running:

    x11vnc -ncache 10 ...

One can also add -ncache_cr for smooth 'copyrect' window motion.
More info: http://www.karlrunge.com/x11vnc/faq.html#faq-client-caching

(ghb:943): GLib-GIO-CRITICAL **: 13:30:47.975: g_dbus_proxy_new_sync: assertion 'G_IS_DBUS_CONNECTION (connection)' failed
14/02/2020 13:31:02 Got connection from client 127.0.0.1
14/02/2020 13:31:02   other clients:
14/02/2020 13:31:02 Got 'ws' WebSockets handshake
14/02/2020 13:31:02 Got protocol: binary
14/02/2020 13:31:02   - webSocketsHandshake: using binary/raw encoding
14/02/2020 13:31:02   - WebSockets client version hybi-13
14/02/2020 13:31:02 Disabled X server key autorepeat.
14/02/2020 13:31:02   to force back on run: 'xset r on' (3 times)
14/02/2020 13:31:02 incr accepted_client=1 for 127.0.0.1:60328  sock=10
14/02/2020 13:31:02 Client Protocol Version 3.8
14/02/2020 13:31:02 Protocol version sent 3.8, using 3.8
14/02/2020 13:31:02 rfbProcessClientSecurityType: executing handler for type 1
14/02/2020 13:31:02 rfbProcessClientSecurityType: returning securityResult for client rfb version >= 3.8
14/02/2020 13:31:02 Pixel format for client 127.0.0.1:
14/02/2020 13:31:02   32 bpp, depth 24, little endian
14/02/2020 13:31:02   true colour: max r 255 g 255 b 255, shift r 16 g 8 b 0
14/02/2020 13:31:02 no translation needed
14/02/2020 13:31:02 Enabling NewFBSize protocol extension for client 127.0.0.1
14/02/2020 13:31:02 Enabling full-color cursor updates for client 127.0.0.1
14/02/2020 13:31:02 Using image quality level 6 for client 127.0.0.1
14/02/2020 13:31:02 Using JPEG subsampling 0, Q79 for client 127.0.0.1
14/02/2020 13:31:02 Using compression level 9 for client 127.0.0.1
14/02/2020 13:31:02 Enabling LastRect protocol extension for client 127.0.0.1
14/02/2020 13:31:02 rfbProcessClientNormalMessage: ignoring unsupported encoding type Enc(0xFFFFFECC)
14/02/2020 13:31:02 Using tight encoding for client 127.0.0.1
14/02/2020 13:31:03 client_set_net: 127.0.0.1  0.0002
14/02/2020 13:31:03 created   xdamage object: 0x40002c
14/02/2020 13:31:04 copy_tiles: allocating first_line at size 41
14/02/2020 13:31:05 client 1 network rate 2498.4 KB/sec (48152.3 eff KB/sec)
14/02/2020 13:31:05 client 1 latency:  0.8 ms
14/02/2020 13:31:05 dt1: 0.0023, dt2: 0.0095 dt3: 0.0008 bytes: 28408
14/02/2020 13:31:05 link_rate: LR_LAN - 1 ms, 2498 KB/s
14/02/2020 13:31:11 created selwin: 0x40002d
14/02/2020 13:31:11 called initialize_xfixes()

[aronmgv]

Checked the Plex container and the HW transcoding is somehow working:

Does this prove anything?

[zocker-160]

this is really strange.......

have you tried running NVENC natively on the machine, e.g. using ffmpeg? with this command for example: ffmpeg -i <inputfile> -c:v nvenc <outputfile>

and then check with nvidia-settings or nvtop if the NVENC chip is being used

[aronmgv]

Firstly thanks for spending time with me here on this - I really appreciate it.

Couldnt install ffmpeg because of some broken dependencies, so took me some time to resolve it. So anyway ffmpeg now says: Unknown encoder 'nvenc':

ffmpeg -i input_media.mkv -c:v nvenc test1.mkv
ffmpeg version 4.1.4-1~deb10u1 Copyright (c) 2000-2019 the FFmpeg developers
  built with gcc 8 (Debian 8.3.0-6)
  configuration: --prefix=/usr --extra-version='1~deb10u1' --toolchain=hardened --libdir=/usr/lib/x86_64-linux-gnu --incdir=/usr/include/x86_64-linux-gnu --arch=amd64 --enable-gpl --disable-stripping --enable-avresample --disable-filter=resample --enable-avisynth --enable-gnutls --enable-ladspa --enable-libaom --enable-libass --enable-libbluray --enable-libbs2b --enable-libcaca --enable-libcdio --enable-libcodec2 --enable-libflite --enable-libfontconfig --enable-libfreetype --enable-libfribidi --enable-libgme --enable-libgsm --enable-libjack --enable-libmp3lame --enable-libmysofa --enable-libopenjpeg --enable-libopenmpt --enable-libopus --enable-libpulse --enable-librsvg --enable-librubberband --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libspeex --enable-libssh --enable-libtheora --enable-libtwolame --enable-libvidstab --enable-libvorbis --enable-libvpx --enable-libwavpack --enable-libwebp --enable-libx265 --enable-libxml2 --enable-libxvid --enable-libzmq --enable-libzvbi --enable-lv2 --enable-omx --enable-openal --enable-opengl --enable-sdl2 --enable-libdc1394 --enable-libdrm --enable-libiec61883 --enable-chromaprint --enable-frei0r --enable-libx264 --enable-shared
  libavutil      56. 22.100 / 56. 22.100
  libavcodec     58. 35.100 / 58. 35.100
  libavformat    58. 20.100 / 58. 20.100
  libavdevice    58.  5.100 / 58.  5.100
  libavfilter     7. 40.101 /  7. 40.101
  libavresample   4.  0.  0 /  4.  0.  0
  libswscale      5.  3.100 /  5.  3.100
  libswresample   3.  3.100 /  3.  3.100
  libpostproc    55.  3.100 / 55.  3.100
Input #0, matroska,webm, from 'input_media.mkv':
  Metadata:
    CREATION_TIME   : 2017-03-09T20:53:19Z
    ENCODER         : Lavf57.7.2
  Duration: 00:29:19.79, start: 0.000000, bitrate: 3233 kb/s
    Stream #0:0: Video: hevc (Main), yuv420p(tv, bt709), 1920x1080 [SAR 1:1 DAR 16:9], 23.98 fps, 23.98 tbr, 1k tbn, 23.98 tbc (default)
    Stream #0:1: Audio: aac (LC), 48000 Hz, 5.1, fltp (default)
    Metadata:
      title           : Surround
Unknown encoder 'nvenc'

I only have installed nvidia-cuda-toolkit but now I am thinking that my system does not have actual cuda installed (I just cannot make it happen under Debian10):

 sudo apt-get -y install cuda
Reading package lists... Done
Building dependency tree
Reading state information... Done
Some packages could not be installed. This may mean that you have
requested an impossible situation or if you are using the unstable
distribution that some required packages have not yet been created
or been moved out of Incoming.
The following information may help to resolve the situation:

The following packages have unmet dependencies:
 cuda : Depends: cuda-10-2 (>= 10.2.89) but it is not going to be installed
E: Unable to correct problems, you have held broken packages.

^^ Followed instructions from the official cuda drivers I remember I couldnt even install it from the local file as well because of same problem: cuda : Depends: cuda-10-2 (>= 10.2.89) but it is not going to be installed..

Thank you, Michal

[zocker-160]

have you tried blender? Can you use your GPU with it? This is normally a very good indication if CUDA works properly or not

but since nvidia-smi did work for you, it seems to be alright, on the other hand Unknown encoder 'nvenc' is normally an indication, that the driver is not properly installed

For Debian I personally prefer installing the nvidia driver directly using the installer from nvidia https://www.nvidia.de/Download/index.aspx

keep in mind that you need to go into init 3 before installing

EDIT: it might as well just help to purge the current nvidia installation and reinstall it; nvidia drivers can brake, when doing kernel updates or similar

[aronmgv]

For Debian I personally prefer installing the nvidia driver directly using the installer from nvidia https://www.nvidia.de/Download/index.aspx

keep in mind that you need to go into init 3 before installing

I used to have it installed via official drivers but then went for the official debian package..

Seems blender can see it:

So next step purge the nvidia package drivers and install them from official page?

[zocker-160]

So next step purge the nvidia package drivers and install them from official page?

hmm I don't know if that really fixes your problem, since Blender can see it, it seems to be installed correctly

I can only imagine that a newer driver might help. I personally use the latest 440.59, it could be, that the latest docker needs the latest nvidia driver as well, but since the Plex docker did work for you.......

I'm running out of ideas :-/

[aronmgv]

I will try, anyway thanks for all the attention here :)

Per this image from the nvidia-docker2:

I need nvidia-cuda-toolkit for container and cuda on the host - which does not install for me because of "broken dependencies". Do you have cuda successfully installed as well? Thanks

[zocker-160]

Do you have cuda successfully installed as well?

I haven't installed this manually, all I did on my debian install, was installing the nvidia driver via the official installer and then installed nvidia-container-toolkit and nvidia-container-runtime

maybe you are missing the container runtime?

I installed Docker from the official Docker repo, the Docker inside the debian repos is too old which docker version do you have?

[aronmgv]

maybe you are missing the container runtime?

I do have it installed - had to in order for Plex HW transcoding to work.

I installed Docker from the official Docker repo, the Docker inside the debian repos is too old which docker version do you have?

docker -v Docker version 19.03.6, build 369ce74a3c docker-compose -v docker-compose version 1.25.3, build d4d1b42b (via deb [arch=amd64] https://download.docker.com/linux/debian buster stable)

-- Anyway will try using driver from the official nvidia page - but a bit later and will let you know. Maybe it will help.

[aronmgv]

Package drivers sucks :/ Once installed official ones it started working:

Another lesson learnt. Thanks again! Michal

[zocker-160]

@MacGyver27

Package drivers sucks :/ Once installed official ones it started working:

yes on Debian I have to agree, that's why I prefer the nvidia installer, even though it is buggy sometimes as well.

glad it works now for you! :+1: