High Performance Model Configs on A3 GPU Expected performance results for Llama2-7B model running on A3 GPU: Llama2-7B Hardware TFLOP/sec/chip 1x A3 (h100-80gb-8) 492 2x A3 (h100-80gb-8) 422 4x A3 (h100-80gb-8) 407 8x A3 (h100-80gb-8) 409 16x A3 (h100-80gb-8) 375