Nvidia releases cuDSS to move simulation calculations onto GPUs

Nvidia made cuDSS generally available, releasing a library that moves the solution of large sparse systems of equations onto GPUs. Those calculations recur in engineering and scientific simulations and can limit how many designs teams can evaluate. Nvidia says Applied Materials achieved up to 10x faster sparse solves than CPU-only approaches. A ZEISS SMT plasma simulation benchmark using COMSOL recorded a 4x speedup and finished in six days and 14 hours. Developers can integrate the library through C, C++, Python and Fortran interfaces.