PC Magik Lab
Local models, real GPUs.
How we measure
The series is built for comparability. Every test follows the same protocol so results can be trusted and reproduced.
Same prompt
Identical prompts are used across runs to eliminate input variance.
Same rig
All measurements are performed on the same physical system.
One variable
Only one parameter is changed per test for clear cause and effect.
Model reloaded
Every run starts with a fresh model load to avoid state carryover.
Published files
Every prompt and result file is published for transparency.
Hardware
One rig, controlled environment. No swapping hardware between tests.
Test Rig
All series benchmarks run on this single configuration to keep comparisons fair.
Why this matters
Using a consistent rig isolates software and model variables. LM Studio provides a stable local inference stack, and the RTX 3090 24 GB offers a real consumer-class GPU with ample VRAM for testing.
- Controlled thermal and power environment
- Reproducible timings
- Transparent setup
About
PC Magik Lab is a YouTube series about running AI models on real, local GPUs. The focus is on practical benchmarking, clear methodology, and reproducible results.
The channel publishes tests with the same prompt, same rig, one variable changed per test, model reloaded before every run, and every prompt and result file published.
PC Magik
Mateusz Piekut
Computer service and homelab operator from Poland.