I ran the same local LLM on an RTX 5070 laptop and one with an iGPU