NVIDIA open-sources cuPotential, speeding molecular dynamics up to ~4x on GPUs
AllThingsApx · x · 2026-11-17
NVIDIA released cuPotential, a CUDA-X library for efficiently computing molecular interactions from standard force fields in molecular dynamics simulations, usable via an OpenMM Python plugin or an experimental C API.
Official benchmarks (integrated with OpenMM 8.4.0, NPT mixed precision, vs stock OpenMM throughput) show roughly 3–4x speedups across systems from 24k to 409k atoms (DHFR, TYK2, BACE, water, ApoA1, cellulose): RTX PRO 6000 Blackwell up to 3.95x, L40S 3.99x, H100 PCIe 3.34x, B200 3.24x, T4 2.64x. The code is open-sourced on GitHub with docs, samples, and benchmark details.
More from Infra
- PiPNN removes the search step from graph index builds, cutting billion-vector build times from days to minutes — qdrant_engine · 2026-11-17
- tinygrad.js runs the real tinygrad in your browser on your GPU via WebGPU — CatAstro_Piyush · 2026-11-17
- Data Center Backlash Grows Loud, With Historical Echoes of 'A Nuisance Has No Rights' — justin_hart · 2026-11-17
- Constrained Decoding for Function Calling Cuts Endpoint Throughput by 17%-90% — remilouf · 2026-11-17
- Terray taps NVIDIA CUDA kernels to speed drug-design model TerraBind up to 7x — AllThingsApx · 2026-11-17
- Big Tech locks up 60% of DRAM supply it can't deploy yet — Reddit proposal for use-it-or-lose-it auctions — politefella0 · 2026-11-17