Same prompt
Every model is tested with the same prompt, so responses are compared apples to apples.
A YouTube series
PC Magik Lab
Local models, real GPUs.
A hands-on series about running AI models on real, local hardware — no clouds, no marketing numbers. Every run happens on the same physical rig, under the same conditions, so what you see is what the metal actually does.
Method
How the series measures things
The only way to compare models fairly is to keep everything constant. The series follows a strict testing discipline:
Same rig
Every run happens on the same physical machine, so hardware differences never skew a result.
One variable per test
Only one variable is changed at a time, so any difference in output can be attributed to it.
Model reloaded before every run
The model is reloaded before every run, so no warm cache or leftover state carries over.
Everything published
Every prompt and result file is published to the repository, so anyone can check the work.
Verifiable by anyone
With the same prompt, the same rig and the same rules, results can be reproduced and checked.
The lab
One rig. No surprises.
The whole series runs on a single, fixed setup. Know exactly what's doing the work.
-
GPU
NVIDIA RTX 3090 · 24 GB
The graphics card doing the actual work in every single test.
-
Server
LM Studio
Runs locally as the model server that serves every run.
One rig for the whole series — the same machine behind every test you see.
About
Who's behind the lab
PC Magik Lab is a YouTube series about running AI models on real, local GPUs. The idea is simple: skip the cloud hype and show what models actually do on hardware you can touch.
Every test follows the same rules — same prompt, same rig, one variable changed per test, the model reloaded before every run — and every prompt and result file is published so the work can be checked and reproduced.
The series is run by PC Magik (Mateusz Piekut), a computer service and homelab operator from Poland.
Local models, real GPUs.
Follow the series on YouTube, and check every prompt and result file in the repository.