MLPerf Inference v6.1 summary.csv CPU-only Xeon 6 rows

Top-level summary.csv of the MLPerf Inference v6.1 results repo, filtered to rows whose accelerator_model_name is N/A (CPU-only). Host processors are Xeon 6 Granite Rapids: 6980P (128c, 2S), 6972P (96c, 2S), 6787P (86c, 2S), 6788P (86c, 2S and 4S). Models filed: llama3.1-8b INT8 Offline and Server (Tokens/s, aggregate) and one whisper INT8 Offline row (Samples/s). Framework column is PyTorch; Dell's note says vLLM with FP8 KV cache. Not a per-request decode number.