NVIDIA announced that its cuDSS library is now generally available. cuDSS moves the sparse‑system solve step of engineering and scientific simulations onto GPUs. Early adopters say it cuts compute time dramatically – Applied Materials saw up to 10× speedup, COMSOL reported a 4× gain, and Dassault Systèmes noted 8‑10× faster runs. The library works with C, C++, Python and Fortran and can run on anything from a workstation to a large GPU cluster.
Why it matters
Engineers can test many more design ideas faster, reducing development costs and time‑to‑market.
