YouTube series · Local AI benchmarking

PC Magik Lab

Local models, real GPUs.

A series that runs AI models on real, local hardware — no cloud, no abstractions. Every run is measured the same way so the results actually mean something.

The method

How the series measures things

Fair, repeatable comparisons. One thing changes at a time and everything is published.

01

Same prompt, same rig

Every model runs on identical hardware and is asked the exact same question.

02

One variable per test

A single thing changes between runs, so you can see exactly what moved the numbers.

03

Model reloaded every run

The model is reloaded before each test to keep results clean and consistent.

04

Everything published

Every prompt and every result file is published so the work can be checked.

The rig

One setup. Real hardware.

All runs happen on a single local machine so comparisons stay apples to apples.

GPU

NVIDIA RTX 3090 · 24 GB

The single graphics card that powers every local model run in the series.

Model server

LM Studio

The local model server that loads and serves every model tested on the rig.

Setup

One local rig

A single machine, no cloud. Every result reflects what this one GPU can actually do.

About

Running AI where it matters — on your own hardware.

PC Magik Lab is a YouTube series about running AI models on real, local GPUs. The focus is on practical, repeatable testing instead of hype — the same prompt, the same rig, one variable changed at a time.

Every prompt and every result file is published so viewers can verify the numbers themselves. No cloud, no black boxes — just models running on real hardware you could own.

Subscribe on YouTube