11:15 - 11:30
Matching Workloads to Blackwell: Power-Resolved Benchmarking of NVIDIA B200 and B300 across HPC and AI Workloads
Mohsen Zand, Trung Dac Nguyen, H. Birali Runesha
University of Chicago, IL, United States
We evaluate NVIDIA B200 and B300 (Blackwell Ultra) GPUs on two independent eight-GPU clusters, spanning computer-vision inference and training, LLM inference with Llama-3-8B and 70B, and six production HPC codes. The B300 is not a uniform successor to the B200: it trails by up to 5.6× on convolutional inference, 8× on 3D vision transformer training, 24–27% on 70B LLM throughput, and 16× on double-precision Quantum ESPRESSO. We resolve the convolutional deficit using per-GPU board power rather than throughput alone: on the affected workloads the B300 draws only 36–53% of its 1100 W envelope, whereas on ViT-Large, which it wins, both parts saturate at 95–97% of TDP. Because our harness keeps inputs resident in device memory, a host-side data pipeline is excluded by construction, localizing the cause to per-kernel cost. Two B300 nodes of identical silicon differ by 2.1× in time-to-first-token solely from a development vLLM build. We give per-workload procurement guidance.
