Ollama Llama3-8b Speed Compairson with different NVIDIA GPU and FP16/q8_0 quantification