BioNeMo beta benchmarks show up to ~2.9x structure prediction throughput on 8x H100
AllThingsApx · x · 2026-09-11
A supplementary post on NVIDIA's BioNeMo Inference Runtime public beta: an open PyTorch-native library that accelerates protein and biomolecular structure prediction inference on NVIDIA GPUs without model conversion.
- Supports Boltz-2, OpenFold2, OpenFold3, and Protenix v2
- With specialized kernels plus CUDA Graphs and Ray-based replicas, throughput scales up to 2.9x in 8x H100 benchmarks
- Aimed at proteome-scale prediction and drug discovery
Useful as the benchmark datapoint behind the main release.
More from Infra
- EdgeRock-4B holds KV cache at 72 MiB across 256K context, 99.1% smaller than Qwen3.5-4B — alifcoder · 2026-10-27
- RTX 5080 price jumps from $1658 to $2318 in ten days on Amazon — Frizzy-MacDrizzle · 2026-10-27
- ComfyUI on 12GB VRAM: measured peaks show offloading turns 24s into 11 minutes — Impossible-Mail-4619 · 2026-10-27
- Neon rebuilds its docs search entirely on its own stack: Lakebase Search, Functions, AI Gateway — matei_zaharia · 2026-10-27
- PyTorch details 2.12x faster on-device streaming ASR on Arm SME2 via ExecuTorch — PyTorch · 2026-10-27
- NVIDIA shows speculative decoding speeding up LLM inference on Jetson with Qwen3.5 9B — NVIDIA Developer · 2026-10-27