This PC Was In The TRASH - Now It's A SERVER
Вставка
- Опубліковано 6 лют 2025
- I turned an e-waste PC into a capable home server with a few cheap upgrades. I replaced the i3-540 with a Xeon X3470, installed more memory, an SSD and an Nvidia Quadro K4200 GPU.
For the software, I installed Proxmox and virtualised a Minecraft server and Jellyfin. I tried to get hardware acceleration working in Jellyfin, but it didn't quite go to plan.
Cool video, it was really neat to see a glimpse into your experience with setting up this virtualization environment on old/repurposed hardware! I have a growing homelab that makes use of several older generation servers though most of them are old enterprise equipment. If you decide to re-visit the GPU passthrough with this machine it could be a fun follow-up. Given that most management is better done remotely via SSH disabling the host PVE access to the GPU is really only a few flags as an update to GRUB. You basically just blacklist the Nvidia drivers in the kernel and reboot allowing you to properly pass it through to your VM. In regards using Jellyfin in an LXC over a full VM I tend to lean towards being more cautious and using a VM for anything accessible from outside my local network. LXC isolation is great though the cautious side of me is remains skeptic.
I'm definitely going to revisit hardware passthrough and transcoding in the future, but it might not be on this system. I recently got an SFF system that I think would do pretty well as a jellyfin server. That's good to know about the LXC containers, fortunately I'm not opening ports on my router for external access.
Good video, your editing has improved. It would have been nice to have seen some stats of CPU and RAM usage for the Minecraft server. You could also try using a basic GT 210 / 610 or something as the primary video adapter and use the quadro only to pass through to Jellyfin.
A bit late to the party, but using an LXC instead of a VM for Jellyfin would solve the issue with not passing through the only GPU. An LXC container would have native access to the (i)GPU without any configuration.
You make good content, kid. Keep at it!
You should leave it in trash...just kidding, great video! 😉
Good find, if it is really just running as a media or game server, you can let it sleep and wake on own, when you need it 👍
Nice Mac collection!
Big jumps in nvenc encoders came from Kepler to Maxwell, Pascal to Turing, and Ampere to Ada Lovelace. So, realistically speaking, the Kepler Quadro you used is the bare minimum for nvenc, and you would do well to at least use a Maxwell or Pascal based card as the quality difference is noticeable.
Wahoo!
I Love RePurposing Trashed PCs. My lil ladies get a kick outta helping dad out.
Great Video, i like to ask. Have you consider using docker on that pc. As far as i understand docker containers are light compared to VM's and with very limited resources i think docker would be the best bet.
i7-860S from aliexpress? - actually a plain 870 is cheaper, but an 860S would lower the power consumption = or an i5-750S if you could make do with 4 cores
Next time consider using PaperMC as your server software, its even more faster, better, than Spigot/Vanilla. Anyways nice video :)
dammit bobby! just run these apps on bare metal and do a backup with guymager - overall a good video #don't break debian