Running AI models on real, local hardware. No cloud. No compromise.
How We Measure
Every episode follows the same rigorous protocol: the model is reloaded before every run, ensuring that the only difference between results is the single variable being tested. All prompts and result files are published alongside each video on YouTube and in the repository.
Every test uses the exact same prompt to ensure consistency across all episodes.
The test bench stays identical throughout the series - only the model variable changes.
Per-test, only a single model parameter or configuration variable is altered.
The model is fully reloaded before each test run to eliminate state carryover.
All input prompts and output result files are made publicly available alongside each episode.
Hardware
The flagship GPU driving every test in the series.
Local model serving platform for running inference on the RTX 3090.
Enables running large local models interactively.
About
Welcome to PC Magik Lab, the YouTube series about running AI models on real, local GPUs. We believe in transparency, reproducibility, and the power of local computing.
Every episode is crafted with care, using the same protocol to ensure your results are comparable and trustworthy.