EM simulations can be classified as high performance computing tasks. This means that computers used for CST Studio Suite® must meet high requirements in terms of CPU, RAM, and graphical specifications in order to achieve optimal performance. Sufficient power supply and cooling must also be ensured for the workstation or server. We strongly recommend buying a complete package from a brand-name manufacturer, e.g. DELL, HP, or IBM, and that the chosen hardware meets the following requirements.
|Minimum requirement||High end server/ workstation recommendation||Notes for recommended hardware|
|Processor||Intel x86-64processor||Dual Intel Xeon Scalable (also known as “Skylake” ) processor||We recommend high processor base clock frequency (>3 GHz) and 8-10 cores per CPU for general-purpose simulation workstation|
( RAM )
|16GB||32– 64 GB per CPU||We recommend fastest RAM memory modules available, currently DDR4-2666|
|Graphics card||100% OpenGL compatible graphics card||NVIDIA Quadro series card dedicated to CAD/CAE applications|
|Storage||30GB of free disk space||At least 500GB hard disk drive||SSDs may be used, but are not necessary for good simulation performance|
( optional )
|Supported GPU card||High-end NVIDIA Quadro or Tesla card||Please refer to the GPU computing guide|
( optional )
|Dedicated compute cluster hardware||Fast network interconnects with low latency, e.g. Infiniband or Intel OmniPath, are strongly recommended.
Please refer to the MPI computing guide
( optional )
|For optimal simulation performance, the solver servers should run on separate computers from the frontend and the main controller. A fast network connection between the solvers servers, the main controller, and the frontend is recommended as simulations may generate a lot of data that needs to be transferred.|
We recommend the latest Intel Xeon processors in a dual socket configuration for a high-end workstation or server configuration; currently these are the Intel Xeon Scalable Processors, also known as "Skylake". For fast performance, a base frequency greater than 3 GHz is recommended. Please note that the processor’s turbo frequency cannot usually be used for long periods due to electrical and thermal limits, so for general performance and for long simulation times the base frequency gives a more realistic performance expectation.
We also recommend 8-10 cores per CPU for a general-purpose simulation workstation. It is in general advisable to have a high processor base frequency rather than a large amount of cores. The performance scaling as a function of number of cores depends on the used solver technology, the simulation model, and other factors.
For some applications and solver technologies, a high amount of processor cores and more than two processors may be a good option to obtain better performance. If you are planning a large hardware investment in high-end HPC hardware like a cluster system or a system with more than two CPU sockets, we recommend that you contact our technical support team directly so that we can help you during the configuration process.
A single processor system with a high base clock frequency may also deliver sufficient simulation performance for many applications. We recommend the Intel Xeon W processors for such lower end workstations. The above advice is also valid for such configurations.
Simulation memory requirement is highly application and solver technology dependent. For a high-end workstation or server system, we recommend at least 32-64 GB RAM per CPU depending on the complexity of your application and the used solver technology. We recommend the fastest RAM memory modules available, currently DDR4-2666.
For Intel Xeon Scalable processors the memory modules should be arranged in multiples of six (6 modules, 12 modules, etc.) per processor socket to fully utilize the available memory channel bandwidth. A high memory channel bandwidth is essential to obtain the best possible performance for many of the CST solvers. The maximum memory channel bandwidth depends on the number of RAM modules as well as the type of the modules. Please ask your hardware vendor to provide you with a configuration that achieves the best possible memory channel bandwidth.
For the best performance of the 3D modeling and post-processing interface, we recommend a fast 3D graphics card. The NVIDIA Quadro-series graphics cards are well tested with CST STUDIO SUITE and we recommend using a card of this series that is dedicated to CAD/CAE applications.
The base installation of CST STUDIO SUITE requires approximately 7GB of disk space while additional space is required for the installation of the service packs and other CST programs and tools.
Simulations may generate a lot of data, so sufficient storage space should be ensured. We recommend at least a 500 GB hard disk drive for a typical simulation workstation. You may use SSDs for storage, but they are not necessary for good simulation performance.
Some CST STUDIO SUITE solvers support GPU computing with NVIDIA Tesla and Quadro GPU devices using a minimum of CUDA version 8.0. The high memory bandwidth and parallel processing abilities of GPU cards provide a significant simulation speed-up compared to CPU computing alone. Options are available for server-class and workstation configurations. Please see the GPU Computing Guide for a list of supported GPU devices as well as information about the requirements that the host system must fulfill.
Some CST STUDIO SUITE solvers support MPI computing. It typically requires dedicated compute cluster hardware (e.g. InfiniBand or Intel OmniPath network interconnects). Please refer to the MPI Computing Guide for more information about the requirements and general setup information.
If you are planning a large hardware investment in high-end HPC hardware like a cluster system, we strongly recommend that you contact us directly so that we can help you during the configuration process.
Distributed computing divides the simulation workload across three different components: frontend, main controller, and one or more solver servers. For optimal simulation performance, the solver servers should run on separate computers from the frontend and the main controller. A fast network connection between the solvers servers, the main controller, and the frontend is recommended as simulations may generate a lot of data that needs to be transferred.
For solver server computers the above advice for hardware configuration is valid, as they run the most resource intensive part of the simulation. The frontend is used mainly for post-processing and graphical analysis of the results, so it does not require powerful hardware. The main controller maintains a simple job queue and transfers simulation data from the solver servers to the frontend, so it also does not require powerful hardware.
We support the latest 64-bit Microsoft Windows operating systems as well the latest 64-bit Red Hat Enterprise Linux and Suse Linux Enterprise versions. Please note that there are some limitations when running CST STUDIO SUITE or other CST tools on Linux, i.e. not all solvers and modules are supported. For more information, please refer to the supported operating systems document.
The following system configurations provide examples of low-end and high-end workstations that are suited to EM simulations using CST STUDIO SUITE.
Low-End Configuration (Laptop):
Note that this configuration is not expected to deliver ideal performance for most practical use-cases. This configuration is appropriate for simple simulations, post-processing calculations and graphical rendering purposes.
High-End Configuration (Workstation):
If a system is used only for results post-processing and analysis, e.g. for distributed computing frontend, then it does not need powerful hardware. A powerful graphics card and enough disk space for storing the results are usually sufficient.
If you have further questions, our technical support engineers are available to review your desired hardware configuration and provide you with feedback and suggestions to improve it.
NOTE: CST assumes no liability for any problems caused by this hardware recommendation.