Transcoding with an Nvidia GPU & Unlocking it for Maximum Streams in Plex, Emby & Jellyfin

Goal

Use an NVIDIA GPU for hardware transcoding on Unraid 6.9+ for:

• Plex
• Emby (Jellyfin-compatible setup via NVIDIA/encoding stack)
• Jellyfin

For users without Quadro cards, the video explains how to remove NVIDIA’s consumer “multiple session” limitations—described as “unlocking” the card to enable maximum streams.

---

Key concepts / constraints discussed

1) NVIDIA “multiple sessions” limitation + Quadro recommendations

Common advice: use Quadro GPUs since they can transcode multiple streams.

The presenter argues this is not strictly necessary. Cheaper NVIDIA consumer GPUs can achieve similar behavior by:

• Installing appropriate drivers
• Applying an “unlock” method

---

2) EULA / compliance debate and “VM passthrough” context

The video covers:

• NVIDIA restrictions for GPU use in VMs (mentions Error 43).
• Unraid has patches that reduce detection/driver disabling during GPU passthrough.
• The presenter claims these VM patches don’t modify NVIDIA driver code, but warns that unlocking for multi-transcode may violate NVIDIA’s driver EULA.

Viewer discretion advised.

---

Installation & setup guide (Unraid 6.9+)

Step A — Update Unraid to 6.9+ (if needed)

1. Go to Tools → Update OS
2. Install the Stable branch (video notes stable released Feb 22, 2021)
3. Reboot after updating

---

Step B — Install NVIDIA driver support on Unraid

1. Open the Apps tab
2. Search for “nvidia”
3. Install the NVIDIA plugin
4. After installation:
   • Stop and restart Docker to ensure drivers are ready

---

Step C — Install GPU monitoring

Install “GPU Statistics” plugin and set:

• Vendor = NVIDIA
• Select the specific GPU (if multiple GPUs exist)

---

Docker container setups for transcoding

Common container requirements

Use NVIDIA-enabled containers (examples shown):

• LinuxServer Plex
• LinuxServer Emby
• LinuxServer Jellyfin

For each container’s template:

• Map media paths (movies/tv/music)
• Set NVIDIA_VISIBLE_DEVICES to the GPU ID

• Add container parameter: --runtime=nvidia

• Switch to Advanced view to edit Extra parameters

---

Plex (hardware transcoding verification)

• Install LinuxServer Plex container
• Configure:
  • NVIDIA GPU ID
  • --runtime=nvidia
• In Plex settings, ensure hardware accelerated video encoding is enabled by checking:
  • “Use hardware acceleration when available”

Requirement highlighted:

• Hardware transcoding requires a Plex subscription / Plex Pass

Testing / verification:

• Force a transcode (e.g., “Convert to 720p”)
• Check Plex Dashboard for HW (hardware encoding)
• Confirm on Unraid using GPU Statistics (video later indicates multiple concurrent transcodes, e.g., 3)

---

Emby (Emby Premier requirement + NVENC check)

• Install LinuxServer Emby container (default branch used in the video)
• Enable hardware acceleration when available in Emby settings

Requirement highlighted:

• Emby Premiere account (not just the free tier) for hardware transcoding

Testing / verification:

• Force a transcode to 720p
• In Emby UI, verify details show:
  • Transcoding steps like H.265 → H.264
  • Use of NVENC (video shows “nvidia encode decode” / H.265 to H.264 using NVENC)

Monitoring:

• Unraid GPU Statistics shows activity
• The presenter notes minor UI confusion about whether Emby appears correctly as an “active app,” though GPU usage is still indicated

---

Jellyfin (free, config + transcoding type)

• Install LinuxServer Jellyfin container
• Configure:
  • NVIDIA GPU ID
  • --runtime=nvidia

Jellyfin-specific settings:

• Admin Dashboard → Transcoding
  • Set Hardware acceleration to something other than “none”
  • Select transcoding backend NVENC
  • (Video mentions choosing from multiple options under that section, described as selecting “top five”)

Testing / verification:

• Force a transcode (e.g., “Conan the Barbarian” to 720p)
• Verification relies heavily on Unraid GPU Statistics, since the Jellyfin UI text is not clearly readable in the video

Key point:

• Unlike Plex/Emby, Jellyfin setup is presented as not requiring a paid key

---

Multi-container / multi-stream validation

• The video claims the NVIDIA GPU can handle multiple containers simultaneously.

Demonstration:

• Run Jellyfin + Plex + Emby concurrently with multiple transcodes

Confirmation:

• GPU Statistics shows three processes

Note:

• Emby may have a minor display/UX issue where activity doesn’t appear as expected in the “active app” view, but hardware usage still appears confirmed via GPU stats.

---

“Unlocking” NVIDIA for more concurrent transcodes (spike-cushion analogy)

After transcoding demos, the presenter demonstrates an automated patch workflow:

1. Stop all GPU-using containers
2. Install User Scripts plugin
3. Create a script named “unlock nvidia”
4. Paste a patch script fetched from a link in the description
5. Schedule it to run at array start
6. Run it immediately to patch/unlock the driver

Analogy: “spike cushions” are used to explain the concept of bypassing NVIDIA limitations by effectively changing what the system allows.

This is also tied back to the earlier discussion about EULA.

---

Main speakers / sources

Primary speaker

• The video author/presenter (“hi there guys…”), who installs/configures everything and discusses Plex/Emby/Jellyfin setup and the unlocking approach.

Primary software sources referenced

• Unraid OS 6.9+
• NVIDIA Unraid NVIDIA driver plugin
• GPU Statistics plugin
• LinuxServer.io containers:
  • Plex
  • Emby
  • Jellyfin
• Plex (Plex Pass requirement)
• Emby (Emby Premiere requirement)
• Jellyfin (NVENC transcoding)