PC Magik Lab
Local models, real GPUs.
A series that runs AI models on real, local hardware — measuring them the same way every time so the results actually mean something.
How the series measures things
Every test follows the same rules, so comparisons across models are fair and repeatable.
Same prompt
Every run of a given test is asked the exact same prompt — no variations that could skew the numbers.
Same rig
Every test runs on the one and only machine, so hardware is never a hidden variable.
One variable at a time
Each test changes exactly one thing, so you always know what moved the result.
Model reloaded every run
The model is reloaded before every single run, so nothing leaks between measurements.
Everything published
Every prompt and every result file is published, so anyone can check the work end to end.
One machine. Nothing virtual.
All benchmarks run on a single local setup with a real NVIDIA GPU — no cloud, no emulators.
GPU
A full-size consumer GPU with plenty of VRAM to run large local models in hardware.
Model server
LM Studio serves the models, so every run uses the same engine and settings.
Real hardware. Real local models.
PC Magik Lab is a YouTube series about running AI models on real, local GPUs. The focus is on honest, repeatable measurement — the same prompt, the same rig, one variable changed per test.
Everything tested comes from a single local machine, and all of the prompts and result files are published so the process can be checked by anyone. No cloud, no black boxes — just models running on hardware you can point at.
Watch and dig in
Head to the YouTube channel for the videos, or open the repository for every prompt and result file.