Introduction to Self-Hosted Media Management

In the modern digital landscape, data privacy and sovereign cloud solutions have moved from niche hobbies to critical business imperatives. For organizations and individuals seeking a robust, self-hosted alternative to mainstream photo platforms like Google Photos or Apple iCloud, Immich has emerged as the premier open-source solution. It offers a highly polished user interface, multi-user support, and advanced features such as facial recognition and object detection.

However, running Immich at scale introduces significant computational overhead. When thousands of high-definition videos and images are uploaded, tasks like video transcoding and machine learning inference can quickly saturate standard CPU resources, leading to bottlenecked performance and sluggish response times. To solve this, leveraging a Virtual Private Server (VPS) equipped with a dedicated GPU combined with hardware acceleration is the definitive solution.

The Importance of Hardware Acceleration (HWA) in Immich

By default, Immich utilizes the host CPU to process tasks. While modern server processors are highly capable, they are fundamentally designed for general-purpose computing. In contrast, specialized tasks like encoding videos via FFmpeg or executing AI models thrive on parallel processing architectures.

Key Insight: Offloading video transcoding and machine learning tasks from the CPU to a dedicated GPU (Graphics Processing Unit) dramatically optimizes server efficiency, reduces latency, and lowers overall operational costs by preventing CPU throttling.

When you enable Hardware Acceleration (HWA) on a GPU VPS, you unlock several critical benefits:

Blazing-Fast Video Transcoding: High-efficiency video codecs (such as HEVC/H.265 or AV1) are processed in fractions of the time compared to software encoding.
Efficient Machine Learning: Facial recognition, CLIP object search, and image tagging execution times drop from seconds to milliseconds.
System Stability: Preventing CPU spikes ensures that other containers or services running on the same VPS remain responsive and stable.

Prerequisites for Deployment

Before initiating the installation process, ensure your infrastructure meets the following structural and technical requirements:

GPU-Enabled VPS: A cloud instance from providers like Linode, Vultr, or specialized GPU clouds, running a clean installation of Ubuntu 22.04 LTS or Ubuntu 24.04 LTS. An NVIDIA GPU (e.g., T4, A10G, or RTX series) is highly recommended due to superior driver maturity in Linux environments.
Docker and Docker Compose: The latest stable versions installed on the host system.
NVIDIA Container Toolkit: Necessary to expose the physical GPU resources safely into the isolated Docker container environment.

Step 1: Installing NVIDIA Drivers and the Container Toolkit

To allow Docker containers to communicate directly with the host's GPU hardware, we must install the proprietary NVIDIA drivers followed by the runtime toolkit.

First, update your package repository and install the recommended headless drivers for server environments:

sudo apt update && sudo apt upgrade -y
sudo apt install -y nvidia-driver-535-server nvidia-utils-535-server

After installation, a system reboot is required to load the kernel modules. Once back online, verify the installation by executing the nvidia-smi command. You should see a detailed readout displaying your GPU model, driver version, and current temperature.

Next, configure the production-ready repository for the NVIDIA Container Toolkit and install it:

curl -fsSL [https://nvidia.github.io/libnvidia-container/gpgkey](https://nvidia.github.io/libnvidia-container/gpgkey) | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg

curl -s -L [https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list](https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list) | \
  sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
  sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

sudo apt-get update
sudo apt-get install -y nvidia-container-toolkit

Restart the Docker daemon to apply the new runtime layer configuration: sudo systemctl restart docker.

Step 2: Configuring Docker Compose for Immich HWA

Immich utilizes a microservices architecture distributed across multiple containers. To enable GPU hardware acceleration, we must modify the docker-compose.yml file specifically for the immich-server (or immich-microservices depending on your version) and the immich-machine-learning containers.

Below is an optimized configuration blueprint demonstrating how to inject the NVIDIA runtime access layer:

version: "3.8"

services:
  immich-server:
    container_name: immich_server
    image: ghcr.io/immich-app/immich-server:release
    # ... keep standard environment and volumes variables ...
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    restart: always

  immich-machine-learning:
    container_name: immich_machine_learning
    image: ghcr.io/immich-app/immich-machine-learning:release-cuda
    # Note the '-cuda' tag modifier above to pull the GPU-optimized build
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    restart: always

Crucial Technical Note: Ensure you change the image tag for the machine learning service from the standard release to the -cuda variant. This ensures the container contains the necessary internal libraries (like CUDA and cuDNN) to utilize the hardware processing pipeline effectively.

Step 3: Activating Hardware Acceleration via the Immich UI

Once the containers are deployed using docker compose up -d, navigate to your Immich web administration panel to complete the software-level routing.

Follow these structural steps within the dashboard interface:

Log in as the primary administrator and click on the Administration settings gear in the top corner.
Navigate to the System Settings sidebar tab and locate the Video Transcoding options block.
Change the Hardware Acceleration toggle switch from None to NVENC (NVIDIA Video Encoder).
Review the preferred video resolution options and click Save to commit the parameters to the database.

Step 4: Monitoring and Validating Performance

To guarantee that your configuration is running efficiently and utilizing the underlying hardware instead of falling back to software emulation, you should observe the live metrics.

Execute the following monitoring command directly in your VPS secure shell console:

watch -n 1 nvidia-smi

Trigger a substantial event within Immich, such as importing a multi-gigabyte 4K video file or forcing an all-hands "RE-RUN ALL FACIAL RECOGNITION" command from the jobs panel. If configured correctly, you will observe the volatile GPU utilization percentage rise, alongside a clear process listing showing active tasks inside the container namespaces.

Conclusion and Best Practices

Integrating Immich with hardware acceleration on a GPU-powered cloud VPS creates an uncompromised, institutional-grade media solution. It solves the performance challenges associated with scaling a media catalog, ensuring smooth operations even during high-volume workflows.

To maintain peak operating efficiency, always ensure your underlying host NVIDIA drivers remain updated in lockstep with Docker runtime updates. Regularly check container resource consumption metrics to maintain a secure, sustainable self-hosted ecosystem.

Optimizing Immich: Implementing GPU Hardware Acceleration on a Cloud VPS for Seamless Media Backups