Guides
Local-AI guides
Running models at home, explained the friendly way. What to buy, what fits, and how to get going without the jargon.
LM Studio, start to finish
Install it, pick a model that fits your card, and set the two options that decide fast or unusable.
Ollama vs LM Studio
Same engine, same speed. The choice is how you drive it and whether you are building on top.
vLLM, set up properly
Docker, compose, model swapping and the flags we measured. The fastest way to serve a model to several people at once.
llama.cpp, run directly
The engine under both of them. The flags that matter, and how to measure your own machine.
Run an LLM on your own hardware
Why bother, which GPU you need, and getting a model answering you tonight.
The best AI PC for local LLMs
DGX Spark, Ryzen AI Max or Mac Studio? What the memory numbers mean and which to buy.
The best prebuilt AI-ready PC
Prebuilt towers with an RTX 5090 or 4090 - what matters for AI and which one I would buy.
Strix Halo for local AI, explained
The Ryzen AI Max chip that puts 128GB in a mini-PC - what it runs, and which box to buy.
How to undervolt a GPU (and why I power-test every card)
The 108-watt screenshot, cap vs undervolt, and the one-evening test almost nobody runs on 24/7 hardware.
nvidia-smi, and power-capping your GPU for AI
Read your GPU from the command line, and cap its power so it runs cooler and quieter with barely any speed lost.