Mine is running as a container on an old Mac Pro 2013 (running Proxmox). My GPU is not even as good as yours.

It runs great.

The “beta timeline” required hashing that took forever but this was on my wife’s iPhone and nothing to do with the back-end (she has 50,000 photos).

The machine-learning stuff uses the GPU. Facial recognition for example. It takes longer on low-end hardware but I am not sure why that is a problem.

The image categorization can use GPU acceleration, many seem to be noting it runs well enough on CPU mode even on modest hardware, but importing say hundreds of gigs may take longer for classification (ingress) steps.

It runs great in a container on my synology NAS which has no GPU at all.