Local AI,
Real GPUs

Testing AI models on real hardware, transparently and honestly.

A YouTube series dedicated to running AI models on local GPUs. Same prompts, same rig, one variable changed per test. Every result published.

Scroll

How We Measure

Transparent methodology for consistent, comparable results.

Same Prompt

Every test uses identical prompts to ensure fair comparison between models.

Same Rig

Hardware remains constant throughout the series for reliable performance metrics.

One Variable

Only one factor changes per test: the AI model being evaluated.

Model Reload

Each model is freshly loaded before every run to ensure consistent performance.

Full Transparency

Every prompt and result file is published for complete reproducibility.

Our Hardware

Powered by NVIDIA RTX 3090 and LM Studio.

GPU

NVIDIA RTX 3090

  • VRAM24 GB
  • ArchitectureAmpere
  • Memory TypeGDDR6X
  • Power350W

Server

LM Studio

  • PlatformDesktop App
  • FunctionModel Server
  • InterfaceLocal API
  • SetupLocal Only

Location

Poland

  • CountryPoland
  • OperatorPC Magik
  • FocusHomelab
  • ServiceComputer Service

About PC Magik Lab

A YouTube series dedicated to running AI models on local GPUs. We believe in transparency, reproducibility, and real-world testing.

Every test follows strict methodology: same prompts, same hardware, one variable changed per run. Results are fully published for anyone to reproduce.

MP

Mateusz Piekut

PC Magik