Simulation performance is heavily dependent on the number of unknowns that the solver needs to calculate. No matter how well chosen the solver settings are to the problems at hand, there will always be situations where a straightforward simulation will need very high computational resources, a very long solver time, or even fail altogether.
These include full-wave simulation of complex or electrically-large systems, which need very large numbers of mesh-cells for accurate modeling, along with particle-in-cell simulations of devices containing very dense clouds of charged particles, and parameter sweeps and optimizations of systems over many variables.
To make these sorts of complex simulations more practical, CST STUDIO SUITE® supports multiple High-Performance Computing (HPC) methods for improving simulation performance:
The use of HPC in CST software is governed by a flexible acceleration token concept. Acceleration tokens can be purchased as part of the license, and depending on the situation, the tokens can be used to access and combine different HPC options. For example, a token can be used in one run to use multiple computers in a cluster working closely together to solve a simulation problem too large to be handled by a single machine (MPI), and in the next run, to distribute multiple simulations of a parameter sweep across multiple machines and run them in parallel (DC).
Selecting the hardware and acceleration technique best suited to your needs can sometimes be challenging. CST hardware experts are glad to help you with your decision, performing benchmarks with models you provide on different hardware configurations to give you confidence when deciding for a specific hardware configuration. For large investments (like a complete computer cluster) benchmarking can be performed in cooperation with the test centers of your preferred hardware vendor.