YouTube series • Local AI testing
PC Magik Lab
Local models, real GPUs.
A series about running AI models on real, local GPUs. Honest measurements, same rig, transparent results.
Series focus
Same prompt
Same rig
One variable
Reloaded every run
Published data
Clean comparisons for local LLM performance. No marketing, just measurements.
How we measure
Consistency first. Every test follows the same protocol so results are comparable.
1
Same prompt
Identical input across all runs to eliminate prompt variance.
Identical input across all runs to eliminate prompt variance.
2
Same rig
One hardware setup for the entire series.
One hardware setup for the entire series.
3
One variable changed per test
Isolate the change to see its real impact.
Isolate the change to see its real impact.
4
Model reloaded before every run
Cold start each time for fair timing.
Cold start each time for fair timing.
5
Every prompt and result file published
Full transparency. Check the repository for raw data.
Full transparency. Check the repository for raw data.
Hardware
One rig, documented and constant.
Rig
- GPU: NVIDIA RTX 3090 24 GB
- Model server: LM Studio
- Setup: Local, no cloud
Why this matters
Using a single, real GPU keeps variables low. LM Studio provides a consistent local server for loading and running models under the same conditions.
About
PC Magik Lab is a YouTube series about running AI models on real, local GPUs. The goal is clear, repeatable comparisons with published prompts and results.
Transparency
All prompts and result files are published in the repository. No hidden setups, no cherry-picked numbers.