NVIDIA Launches AIPerf to Benchmark LLM Inference at Scale

NVIDIA launched AIPerf, a ground-up rewrite replacing GenAI-Perf, using a multiprocessed ZMQ architecture to bypass the Python GIL for accurate LLM inference benchmarking. It supports over 15 endpoint types and datasets like ShareGPT.

• Radeon RX 10800 XT may beat RTX 5090 by 15-25% in 4K gaming and local AI2026-09-20
• Nvidia vs. Micron: AI stock picks for 20272026-09-19
• Prediction: Nvidia-backed AI stock Marvell could 10X by 20332026-09-19
• Jensen Huang calls for developing AI 'as fast as we can'2026-09-19
• Nvidia CEO says '0% chance' world ends in 20302026-09-19