Performance

Simulation speed is a red hot topic at the moment. It is worthwhile discussing the meaning of the terms speed and performance in simulation software.

What is Performance?

Just reducing it to computational speed is not sufficient. The simulation accuracy is of paramount importance. An inaccurate result will be useless, no matter how long it takes to reach it. Therefore we define performance as the combination of accuracy and speed. It is inversely proportional to the time it takes you to reach your predefined goals. And of course, also the usability contributes to the total performance of a software package.

CST simulators have proven their outstanding accuracy and speed in numerous published benchmarks. All kinds of HF devices have been analysed: a wave-guide power splitter, microstrip phase shifter, waveguide fed dual mode cavity filter and a Vivaldi antennas as illustrated here.

Numerical Advantages

Leaving aside the usabitlity, the reasons for this success are manifold. First of all there are the numerical advantages offere by the method used in most of the solvers, the Finite Integration Technique (FIT). CST MWS transient solver based on FIT is certainly the foundation of CST's success as an EM simulation software vendor. The efficiency is continously improved by developments like SmartGrid™. CST's Complete Technology provides various numerical approaches, seamlessly integrated into one user interface. The user can select the most appropriate one for his problem class. In many cases there is the possibilty to verify results by applying another solver to the same problem.

CST Simulation Acceleration

The CST simulation acceleration scheme summarizes the hardware related mechansism implemented into the CST STUDIo SUITE to minimize simulation time.

Code Optimization

There is continous research ongoing on the hardware level in order to make use of the latest technological developments. As an result of CST's cooperation with Intel®, the performance on Multi-Core Intel® Xeon® processors was increased up to an factor of four (Read the Intel®/CST solution profile).

GPU Computing

GPU-computing is an increasingly popular alternative to using traditional, expensive high performance computing equipment. CST employs NVIDIA® CUDA™ technology to dramatically shorten simulation time.

Cluster Computing

CST’s parallelization based on the Message Passing Interface (MPI) enables the usage of computers in a network. It increases performance in 2 ways:

1. Reduction of simulation time through parallel use of CPUs on several computer systems
2. Ability to simulate models previously too large to solve through domain decomposition and distribution over the memory of several computers

Distributed Computing

An additional effort to speed up complex simulations is undertaken by employing available network resources. CST's distributed computing scheme efficiently uses available computing power to minimize simulation time for optimization, parameter studies, broadband or multiport devices.

It’s the accuracy that makes your ideas run faster.

Numerical solutions to Maxwell’s equations vary greatly, and the performance of a code can only be judged in the time taken to reach a certain level of simulation accuracy. This involves serious benchmarking, and performance values may vary depending on the application example. At CST we are sure that our simulation technology will give you the competitive edge.