Skip to content

Set up GPU hardware transcoding for Jellyfin, Plex and Emby

How-toDeployrr v6.0.42+CurrentMediaHost systemDocker

A GPU lets Jellyfin, Plex and Emby convert video faster and with less CPU load. This page shows how to scan for your GPU in Deployrr, give it to the app, turn on hardware transcoding inside the app, and check that it works.

  • You have a Basic, Plus or Pro license. If you don’t, System > GPU shows a Requirements Not Met (N) box. It says “License - Basic, Plus, or Pro required. Current: …” and “(Prerequisites->Verify)”.

  • You finished the Prerequisites steps. If you didn’t, the same box also says “Prerequisites - All steps must be completed”. See Quick start.

  • Your server has an Intel or AMD GPU. See the caution above for NVIDIA.

  • For an Intel or AMD GPU, the folder /dev/dri exists on the server. This is a general Linux check, not a Deployrr feature:

    Terminal window
    ls /dev/dri

    You should see a name that starts with renderD, such as renderD128. If you don’t, see Prepare your server’s GPU so Deployrr can use it.

  • Plex only: you have a Plex Pass. Plex hardware transcoding needs it. This is general Plex knowledge, not a Deployrr rule.

  1. Open System > GPU. The menu line reads “Graphics Card (for HW Transcoding)”.
  2. Read the Graphics Card box, then select OK.
  3. Wait while Deployrr shows “1/2” (“Scanning for graphics card devices and toolkits…”) and “2/2”.
  4. Read the result. It is one of these three:
What you see What it means
“Graphics card detection completed - AI & Hardware transcoding supported” At least one GPU is ready. Go to part 2.
“GPU hardware detected but required toolkits are missing.” Deployrr found a GPU but it is not ready. Read the lines above that message. They name what is missing. Fix it, then run the scan again.
“No graphics cards detected via lspci.” or “Graphics card detection completed - No GPU hardware found” Deployrr found no GPU.

To see GPU detection in Deployrr 6, watch Deployrr 6 overview: Intelligent GPU Detection at 5:37.

The scan prints a line for each GPU it finds:

Line What to do if it fails
“Intel GPU: Found” then “Intel Iris: Ready” Nothing. It is ready.
“Intel Iris: /dev/dri not available” Install the Intel GPU drivers, or pass the GPU into your LXC or VM, then check that /dev/dri exists. See GPU host prep.
“AMD GPU: Found” then “AMD ROCm: Ready” Nothing. It is ready.
“AMD ROCm: /dev/dri not available” Install the AMD GPU drivers, or pass the GPU into your LXC or VM, then check that /dev/dri exists. See GPU host prep.
“NVIDIA GPU: Found” then “NVIDIA Container Toolkit: Ready” The NVIDIA driver is installed and Docker is set up for NVIDIA.
“NVIDIA Container Toolkit: Not installed or not configured” Install the NVIDIA driver, then set up Docker for NVIDIA with the NVIDIA Container Toolkit. Deployrr does not install them. It prints the link to the NVIDIA install guide below this line. For the steps, see GPU host prep.

If you see “lspci command not found. Install pciutils to enable GPU detection.”, install the pciutils package with your Linux package manager, then scan again.

When the scan works, the GPU line in the System menu shows the GPU types in green, such as “nvidia, intel”. If it shows “Not Set/Found” in yellow, the scan did not find a ready GPU.

Deployrr asks you about the GPU when you install the app. Install Jellyfin, Plex or Emby after the scan.

  1. Install the app from the Apps menu, the way you install any app.
  2. If you did not scan before, you see “GPU detection has not been run yet.” Choose Yes at “Would you like to run GPU detection now?”.
  3. Find the box named “GPU Selection for” and the app name, such as “GPU Selection for Plex”. It shows “Hardware transcoding” for Jellyfin, Plex and Emby.
  4. At “Select GPU acceleration:”, pick your GPU. The first entry is highlighted. There is no default, so check before you press Enter. The options are:
    • NVIDIA GPU
    • AMD GPU
    • Intel GPU
    • CPU Only (no GPU acceleration)
  5. Wait for “GPU acceleration configured”. The text in brackets after it is intel_amd for Intel or AMD, nvidia for NVIDIA, or amd for Ollama with AMD.

The box lists only the GPUs that are ready and that the app supports. For example, Ollama offers only NVIDIA and AMD, and Kasm offers only Intel and AMD. If you pick CPU Only (no GPU acceleration), you see “Using CPU mode (no GPU acceleration).”

If no GPU is ready, you skip the box and see “No compatible GPU detected. Using CPU mode.” If you cancel the box, you see “GPU selection cancelled. Using CPU mode.” In all these cases the app installs and runs on the CPU.

If Deployrr cannot add the GPU lines, it shows an error and then “Installing APP without GPU acceleration (CPU only)”. The app still installs. This is what happens when you pick NVIDIA for Jellyfin, Plex or Emby. See the caution at the top.

These apps ask the GPU question: Jellyfin, Plex, Emby, Dispatcharr, Ollama, Kasm, Chromium, DigiKam, Double Commander, FileZilla, Lollypop, Remmina and XPipe Webtop. Other apps do not.

Deployrr adds the GPU lines when it installs the app. Scanning later does not change an app that already runs. To add the GPU to an installed app, install it again over the existing one:

  1. Open Apps > Browse Apps and pick the app.
  2. Choose Yes at “Do you want to install this app?”.
  3. At “Already Installed”, read what reinstalling does: it stops the existing container, recreates it with the latest image and preserves the data volumes. Choose Yes.
  4. Answer the GPU box as in part 2.

You can also use Apps > Manage Apps > Reinstall. This option shows only when the app is stopped.

4. Turn on hardware transcoding in the app

Section titled “4. Turn on hardware transcoding in the app”

These steps are on the app’s own screens, not Deployrr’s. They are general knowledge, and the names can change between app versions.

Jellyfin

  1. Open Jellyfin and go to Dashboard > Playback > Transcoding.
  2. Set Hardware acceleration to VAAPI. This is for Intel and AMD.
  3. For the device, use /dev/dri/renderD128. If the folder has a different renderD name, use that one.
  4. Under Enable hardware decoding for, tick the formats your GPU supports.
  5. Save.

To see this done by hand, watch Jellyfin Docker Compose: turn on and test hardware transcoding at 11:40 (manual Docker Compose setup). The video picks Intel Quick Sync for an Intel GPU and says VAAPI suits AMD.

Plex (needs Plex Pass)

  1. Open Plex and go to Settings > Transcoder.
  2. Turn on Use hardware acceleration when available.
  3. Turn on Use hardware-accelerated video encoding if you see it.
  4. Save.

To see the Plex transcoder settings and a test, watch Plex Docker Compose: configure and test hardware transcoding at 17:53 (manual Docker Compose setup).

Emby (needs Emby Premiere)

  1. Open Emby and go to Settings > Transcoding.
  2. Pick the hardware acceleration type for your GPU.
  3. Save.

These checks are general Docker and Linux checks, not Deployrr features.

  1. Check that the app container sees the GPU. Replace jellyfin with plex or emby if you use those:

    Terminal window
    sudo docker exec jellyfin ls /dev/dri

    You see the same card and renderD names as on the server.

  2. Play a video that needs transcoding. One way is to set the quality in the player lower than the file’s quality.

  3. Open the app’s activity screen. In Jellyfin it is Dashboard. In Plex it is Settings > Status > Dashboard. In Plex, a hardware transcode shows “(hw)” next to the video line. In Jellyfin, the playback info names the hardware method.

If the check fails, open the compose file for the app. It is at DOCKER_FOLDER/compose/HOSTNAME/jellyfin.yml. Replace DOCKER_FOLDER with your Docker folder path, HOSTNAME with your server’s hostname and jellyfin with your app. For Intel and AMD you see these lines:

devices:
- /dev/dri:/dev/dri

If the lines are missing, the app was installed on the CPU. Repeat part 3.

Ollama and Dispatcharr get a different block for NVIDIA:

deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]

The v6.0.42 release notes say this block is untested on NVIDIA hardware. If it works for you, nvidia-smi shows your GPU inside the container. Run it with sudo docker exec APP nvidia-smi, and replace APP with the container name.

Ollama with AMD gets these lines:

devices:
- /dev/kfd:/dev/kfd
- /dev/dri:/dev/dri

If you edit the file, recreate the app so it picks up the change. Use Tools > Stack Manager > Recreate. See Stack Manager. Or run this Deployrr command if you installed the bash aliases. Replace APP with the app’s name:

Terminal window
dcrec APP

Without the aliases, run this. Replace DOCKER_FOLDER with your Docker folder path, HOSTNAME with your server’s hostname and APP with the app’s name:

Terminal window
sudo docker compose --profile all -f DOCKER_FOLDER/docker-compose-HOSTNAME.yml up -d --force-recreate APP