Hello, Spare Cycles

The whole homelab is one refurbished Lenovo ThinkStation P330 in my closet, with an EliteDesk G2 on the shelf as a backup. The P330 cost about $300 on eBay in September 2026, and the first thing it needed was an ethernet cable. I took the one from my printer. The printer had opinions about that, and they’re their own post.

This is Spare Cycles, a blog about whatever tech is currently eating my free time. Right now that’s this box and the local LLMs I’m working toward running on it. My day job is integrations and solution architecture, so I spend a lot of time drawing boxes and arrows for other people. This is where I draw some for myself and write down what happened. The name covers the CPU’s idle time and mine.

The box

It’s quiet, it sips power, and it has full-height PCIe slots. Remember the slots.

What runs on it today

Proxmox VE is the hypervisor. On top of it:

• A media container running a media server and a reverse proxy that handles HTTPS.
• An apps container with the usual download-automation stack behind a VPN, plus uptime monitoring.
• A Home Assistant VM with a Zigbee coordinator passed through, running a few smart plugs, door sensors and too many light bulbs.

Admin access stays private over Tailscale. Config lives in git, and backups hold the state. Keeping those two separate has already paid off once.

P330 (Proxmox VE)
├── media container     media server, reverse proxy
├── apps container      downloads behind a VPN, monitoring
└── Home Assistant VM   Zigbee, lights, automations

The number that matters

Tokens per second is the number that decides whether a local model is a tool or a demo. The P330 currently has no GPU and 16 GB of RAM, so for anything big the honest answer is “not many.”

That’s the first experiment. What I want to find out:

• which models are worth running on modest hardware, and at what quantization
• how to serve them so the rest of the homelab can use them
• how to wire them into the integration tools I use at work

What’s next

Unglamorous groundwork first: ad-blocking DNS, proper off-site backups, and a UPS so a power cut doesn’t take out a drive. Then the fun part, picking inference hardware and measuring it honestly.

The slots are waiting.