Local models, real GPUs.
PC Magik Lab
A YouTube series about running AI models on real, local GPUs. Testing performance, comparing architectures, and publishing every prompt and result file transparently.
A YouTube series about running AI models on real, local GPUs. Testing performance, comparing architectures, and publishing every prompt and result file transparently.
Every test uses the exact same prompt to ensure comparability across models and configurations.
All tests run on the same hardware configuration to isolate model performance variables.
Only one parameter is modified per run to clearly identify what affects results.
The model is fully reloaded before each test to ensure clean state and consistent results.
All prompts and result files are published openly for transparency and reproducibility.
The flagship consumer GPU with 24GB of VRAM, capable of running large language models locally.
Lightweight model serving software used to load and serve AI models locally during testing.
PC Magik Lab is a YouTube series created by Mateusz Piekut, a computer service technician and homelab operator based in Poland. The series explores the practical side of running AI models on consumer-grade hardware.
Each episode focuses on testing one variable at a time, using the same prompt and the same rig to ensure fair comparisons. All prompts used and result files generated are published alongside every video for full transparency.