Hello everyone,

 

this is a simple guide to use your Intel iGPU or dGPU for transcoding in Immich.

 

Install Intel-GPU-Top Plugin from the Community App Store, no configuration needed:

 

(optional) install the GPU Statistics plugin:

 

Navigate to the Docker Tab and edit the Immich Docker.

 

Enable Advanced View in the top right corner.

 

Add the Extra Parameter: --device=/dev/dri

Save the change to rebuild the docker.

 

Open the Immich Web Interface, select your User Profile in the top right. Choose Administration.

 

Navigate to Settings:

 

Scroll down to the bottom to Video Transcoding Settings.

 

Transfer the Settings from the Screenshot and save:

/dev/dri/renderD128

 

Start a transcode and check using intel_gpu_top on the cli will show a process ID and load.

 

looking at the docker logs will show the intel GPU being used for Transcoding.

 

Thats it, enjoy.

Edited February 6, 20251 yr by frogwell

Thank you for this. I did a google search for a quick way to get this done and your post popped up. It worked like a charm!

Would that work with the Alder Lake N100?

 

I applied the settings as described above but when transcoding there is nothing showing in intel_gpu_top

 

The logs seems to show success though:

 

[Nest] 721  - 03/08/2025, 9:51:13 PM     LOG [Microservices:MediaService] Transcoding video c1fa4d38-d44d-4ce6-9a3d-a7e51e80c671 with QSV-accelerated encoding and decoding
[Nest] 721  - 03/08/2025, 9:51:13 PM     LOG [Microservices:MediaService] Successfully encoded c1fa4d38-d44d-4ce6-9a3d-a7e51e80c671

 

Edited March 8, 20251 yr by googleg

Hi all, just installed Immich and got it running no prob. I did a google search for Immich + iGPU and your post popped up.

Stuck for an hour at Add the Extra Parameter: --device=/dev/dri.

Read the instructions again, and noted that "Enable Advanced View in the top right corner" AFTER clicking "Edit docker", NOT the "Enable Advanced View in the top right corner" ON the Docker page.

I don't think this works for the machine learning OCR right?

That has different steps over here: https://docs.immich.app/features/ml-hardware-acceleration/