Hardware Recommendations for SOLIDWORKS Simulation and Flow Simulation

Article by Scott Durksen, CSWE updated March 29, 2021

Article

This article provides recommendations related to hardware for SOLIDWORKS Simulation and SOLIDWORKS Flow Simulation software.

You can learn more about general SOLIDWORKS hardware requirements for SOLIDWORKS 2021 on our blog.

Be aware that more powerful and expensive computers may not solve simulation studies in a practical timeframe if best practices are not observed.  To learn more about tools and techniques to improve your Simulation studies, we recommend attending a SOLIDWORKS Simulation training course on the topic.  This will provide much greater performance improvements compared to an overly expensive workstation.

The following gives details on specific hardware components.  However the entire system should be designed and assembled for optimal performance.  Custom machines can be a cheaper option, however selecting components individually may not be compatible or have bottlenecks.  Cooling and thermal management is also an important aspect.  Purchasing a machine from a computer manufacturer ensures all components will work together correctly.

NEW Dell Hardware offer for US Customers: Get a discount on a Dell Certified SOLIDWORKS Workstation through our partner TriMech. View the SOLIDWORKS Workstations »

SOLIDWORKS Simulation Hardware considerations

There is a general order of importance of hardware when running Simulation studies:

  1. CPU Clock Speed (GHz)
  2. Number of CPU cores
  3. SSD Hard Drive
  4. RAM
SOLIDWORKS Simulation Hardware

SOLIDWORKS Simulation Hardware

Computer Processor (CPU)

All hardware of a computer is important to consider, but the CPU is still the most important component in terms of the speed for running Simulation studies.  The CPU Clock Speed should be the first item to consider.  A number of Simulation processes can only utilize a single CPU core while solving.  Therefore the CPU Clock Speed would be the bottleneck as the other cores will not be used during this time.  Look for a CPU speed closing in on 4GHz.

Be aware that “Boost” or “Turbo” speeds represent the max speed that can be hit, but not sustained.  It may not reach these speeds in all scenarios or for an extended period of time.  It must remain within specification limits for workload, temperature and power.  Given Simulation studies can be long to solve, the CPU will not be able to sustain peak speed.

SOLIDWORKS Simulation and Flow Simulation do take advantage of multiple CPUs and cores.  However don’t expect doubling the number of cores will cut your solution time in half.  Certain operations are inherently linear and therefore cannot take advantage of multiple cores.  Power users typically favour between 4 and 12 cores, though recent tests have found that optimal performance scalability is 8 cores.  It has been suggested that a single CPU with the same amount of cores has slightly better performance compared to the same amount of cores split over multiple CPUs.

Hyperthreading can be misleading.  A system may show 4 cores with 8 logical processors, which indicates this is a quad core machine with hyperthreading enabled.  Each core is divided into two “virtual” cores though each virtual core only has half the bandwidth and power.  While performance is not degraded with hyperthreading enabled, we can’t expect the additional virtual cores to improve performance.

  • SOLIDWORKS Simulation 2021 has made improvements in multi-core processing.
  • SOLIDWORKS Flow Simulation multithreading scalability is optimal on up to 20 cores.

We cannot expect the CPU to have 100% usage all the time.  Larger studies may see better utilization of cores, however Microsoft Windows ultimately manages the CPU usage for each program based on requirements.

Storage

SOLIDWORKS Simulation and Flow Simulation creates a large amount of data.  All of this data is been written to storage drives while the study is running.  Having a high-performance CPU does not define how fast the calculated information is being saved.  Therefore storage drives are a key performance aspect of a system.

SSDs (solid state drives) are significantly faster at read/write commands compared to standard HDDs (hard disk drives).  It’s common to see two drives in a computer, an SSD for the operating system and programs, and a larger capacity HDD for files.  This was typically done for cost savings as high-capacity SSDs can be expensive.  However they have become more affordable over the years and are worth the investment.  The SOLIDWORKS files and Simulation results should also be located on the SSD drive while solving.

NOTE: Never solve a study with the results folder referencing a network location as this will cause significant performance issues.

Use an SSD drive with a read/write of at least 500MB/s if possible.  The temporary result files can be much larger in size while solving compared to the final result file.  Have at least 100GB+ of free space as the temporary results can balloon to many GBs in size for large/complex studies.

Consider investing in an NVMe (Non-Volatile Memory Express) SSD drive for additional performance.  The NVMe interface was designed for SSD storage.  This type of drive looks more like a memory chip.  Standard SSDs have a matching form factor to HDDs so it could connect with the CPU using the legacy SATA (Serial AT Attachment) interface.  NVMe drives are connected to the CPU with a high-speed PCIe socket.

Consider the various possible speeds:

  • 125 MBps – Network connection with Gigabit switch, assuming all other cables/switches/ethernet cards match and absolutely no other network traffic
  • 220 MBps – Local SATA HDD
  • 550 MBps – Local SATA SSD
  • 3000 MBps – Local NVMe SSD

Random Access Memory (RAM)

Adding additional memory won’t necessarily speed up your studies.  However without enough RAM you may start running out of physical memory during the solution process, at which point virtual memory is required.  Virtual memory temporarily stores information on local hard drives rather than the RAM chip.  Even solid state drives (SSDs) are significantly slower at saving data compared to RAM chips.

Typically 32GB of RAM is sufficient for most basic studies.  If you are a heavy Simulation user, 64GB of RAM or more is recommended.

Depending on the size/complexity of the study and which solver is being used, more memory may be required.  For example the Direct Sparse solver in SOLIDWORKS Simulation uses roughly 5GB of RAM per 1 million DOFs (degrees of freedom).  Best practices should be taken to reduce the DOFs as much as possible for faster solve times and to minimize the amount of memory required (i.e. mesh controls, symmetry, etc.).

Adding RAM to all memory slots in a computer can actually improve performance, though this does reduce the upgradability.  Plan for current and future models.  If you expect to get into larger projects at some point, consider installing additional memory.

Error correcting-code memory (ECC) can also be beneficial for stability and may be required for motherboards that support Intel Xeon processors.

Graphics Cards (GPU)

While the graphics card does not improve the overall calculation time to solve the study, it does affect the time it takes to view the results.  Having an old graphics card installed with an out-of-date date driver will cause slow performance and issues when trying to view the plots, especially animations.  Be sure to have a SOLIDWORKS certified OpenGL workstation graphics card and install the certified driver.  Read through our graphics card drivers article for more details.

Purchasing the highest end graphics card may sound like it will give you amazing performance, but the cost will be high for marginal improvement.  Consider a mid-range graphics card to have the best balance.  Read through our SOLIDWORKS System Requirement posts for more detail.

Related Links

Certified SOLIDWORKS Services available from Javelin

Javelin can help you to:

Scott Durksen, CSWE

Scott is a SOLIDWORKS Elite Applications Engineer and is based in our Dartmouth, Nova Scotia office.