Definitely go with K3s instead of K8s if you want to go the Kubernetes route. K8s is a massive pain in the ass to setup. Unless you want to learn about it for work I would avoid it for homelab usage.
I currently run Docker Swarm nodes on top of LXCs in Proxmox. Pretty happy with the setup except that I can’t get IPv6 to work in Docker overlay networks and the overlay network performance leaves things to be desired.
I previously used Rancher to run Kubernetes but I didn’t like the complexity it adds for pretty much no benefit. I’m currently looking into switching to K3s to finally get my IPv6 stack working. I’m so used to docker-compose files that it’s hard to get used to the way Kubernetes does things though.
I use an 6900 XT and run llama.cpp and ComfyUI inside of Docker containers. I don’t think the RX590 is officially supported by ROCm, there’s an environment variable you can set to enable support for unsupported GPUs but I’m not sure how well it works.
AMD provides the handy rocm/dev-ubuntu-22.04:5.7-complete image which is absolutely massive in size but comes with everything needed to run ROCm without dependency hell on the host. I just build a llama.cpp and ComfyUI container on top of that and run it.
I’m curious if cars would be bricked if they couldn’t call home, or if you could selectively allow certain messages through.
I can’t speak for every car but at least Teslas do not mind being offline. You cannot control which messages they send because they connect via a VPN to the mothership. So it’s an all or nothing kinda deal.
You can also pretty easily remove the SIM card on older models with just a few screws. Newer ones use eSIMs, never looked into how to get rid of that one but I assume it is more complicated.
Your comment makes me wonder if one could get around AT by installing faraday cages around where the chips are.
The antennas are usually external, mounted somewhere else in the car and can be unplugged. Never checked if it can still get a signal without the antenna though.
edit: Also, the PCB itself is mounted inside a faraday cage because the entire thing sits inside of RF shielding.
My comment was mostly meant as a joke. I’m aware most of them use their networking capabilities for IPC and being able to use them remotely is just a cool feature resulting from that (except X11).