The multiple power supplies are for redundancy. It will work with one plugged in but you are “supposed” to plug in both.

It’s fundamentally not that different from a consumer desktop. Plug in a monitor and a keyboard and a USB with your preferred flavor of headless Linux installer on it. Configure ssh as the first thing you do because you won’t want to plug the monitor and keyboard into it every time you need to tweak something.

You can probably find VGA to HDMI adapters if you can’t find a monitor with VGA support.

TBF 0C and 0F are kinda arbitrary values.

Wire guard is pretty easy to set up. Also Docker is great.

I only let things I trust are secure (e.g. ssh) have access from the internet, other services I hide behind a VPN (e.g. Tailscale).

Most routers have a feature to assign static IPs to a specific MAC address. You can also tell most devices to try to take a specific IP instead of using DHCP.

There are multiple ways to set it up, but it’s very possible to set a specific device to always have the same local IP, which is usually the first step to many self-hosting scenarios.

Neat!

Any chance you could share some resources on how you did it? I’d kinda like to give it a try…

I have a similar project called PiKVM. I can remotely turn on my computer from a full shutdown, navigate the BIOS to select an OS, and log in, after which I typically switch to a software-based Remote Desktop which is more performant. But you can’t power on a computer and navigate a BIOS with a software solution.

Could I run larger LLMs with multiple GPUs? E.g. would 2x3090 be able to run the 48GB models? Would I need NVLink to make it work?