Homelab · self-hosting · local AI

Build the box.
Measure the watts.

We run a small test bench of local-AI and always-on boxes and write up what they actually draw, how hot they get and what they are worth. Every part was bought with our own money, and every number was read off a meter, not a spec sheet.

See the builds →

HOW EVERY ARTICLE IS MADE

Bought. Run. Measured.

01Buy the part and put it in the rack

02Run it for weeks on real workloads

03Publish watts, temps and tokens per second →

Idle and load power from a wall meter. No vendor numbers.

Build logs Local AI on a budget Quiet, low-watt boxes GPU head to heads Parts we run

How we test

Wall meter, driver telemetry, named models.

Idle and load watts at the plug, GPU temperatures and fan speeds under a repeatable load, tokens per second with the model and quantisation named. The method sits beside every number so you can repeat it.

Read the method →

First articles

The first build logs are being measured now.

The best used GPU for local LLMs under $300, what an always-on box really costs per month, and a two-GPU Ollama build.

Own hardware, own meter, own money. Affiliate links are disclosed on every page and never decide a verdict. Read our editorial policy.

Build the box.Measure the watts.

Bought. Run. Measured.

Wall meter, driver telemetry, named models.

The first build logs are being measured now.

Build the box.
Measure the watts.