Some of the most relevant simulation and optimization workloads rely on stochastic processes, which require an ever-increasing source of high-quality, high-speed random numbers.
Current means to generate random numbers may introduce artifacts in highly parallel simulations, while also consuming valuable computing resources. In contrast to these pseudo-random generators, physical sources of randomness may also be used. However, today’s physical RNG devices are typically slow and unavailable in HPC environments.
The Quside Randomness Acceleration cards deliver high performance and efficient randomness generation for a broad class of randomized algorithms.