Compare measured LLM inference throughput and efficiency across consumer NVIDIA GPUs for the model you care about.