Quantization Explained: How to Run 70B Models on Consumer GPUs