What 57 LLM Benchmark Runs on an AMD Instinct MI300X Taught Me About VRAM, Throughput, and Quantization

I built AMD LLM Lab to understand LLM inference performance before running the model.