Recommended System Requirements
This article complements FlexSim's System Requirements. We'll go over each of the major components, discuss their importance to a FlexSim model, and make recommendations to help you decide on the right hardware for your situation.
Single Model Run
A single run of a FlexSim model (ie not an experiment or optimization run) is single threaded, which means it runs on only 1 CPU core. Buying a processor with more cores won't speed up a single simulation run. To complete single model runs more quickly, use instead a processor with a faster single core clock speed (gigahertz, or cycles per second) or better efficiency (instructions per cycle).
In short, for the fastest individual model runs, you want the fastest processor possible, regardless of the number of cores.
Experimenter or Optimizer - Multiple Simultaneous Replications
When you're doing multiple simultaneous model runs using the Experimenter or Optimizer, FlexSim takes advantage of multiple CPU cores to run simulation replications in parallel.
By default, FlexSim spawns as many replications as you have cores; for example, a Quad core processor enables four single-threaded simulation runs to process simultaneously, one on each core, getting you results four times faster than by using a single core processor at the same speed/efficiency.
If necessary, you can manually limit the number of concurrent replications by setting the Max Cores value from FlexSim's Main Menu under Statistics > Experimenter > Advanced tab > Max Cores input field. Each replication consumes some amount of system RAM, so if you run out of RAM before you run out of cores, it may be necessary to limit the number of cores used to run replications.
If you’re primarily doing single simulation runs, choose a faster processor speed regardless of the number of cores. If you’re primarily doing experiments, choose a processor with a high number of processing cores (and make sure your system has enough RAM to support them - see Memory below). If you’ll be doing both, choose a processor that strikes a good balance between core speed and core count.
Random Access Memory, or RAM, is a PC's fast "working" memory, where it stores the data and programs it is currently running (don't confuse this with your much slower hard drive, where data is stored "at rest"). Most consumer Windows computers top out at 32 or 64 GB of RAM, though Windows is capable of supporting much more when coupled with high-end hardware (see limits per OS here).
Your computer is always doing many other things besides just running FlexSim. Your operating system itself has memory needs for all the processes it is constantly running. You probably have an antivirus solution, maybe a backup software, and often other utilities running invisibly in the background. Then there are all the applications you currently have open - your email, a web browser, music streaming, etc. All of these are consuming RAM.
As you plan your FlexSim hardware, be aware that you should have enough RAM to run all these other applications, plus your FlexSim model, plus have a good amount of extra RAM as a buffer.
Many FlexSim models will use less than 4 GB of RAM in a single model run, but it all depends on your simulation and data. Some large and complex models can and do go above 4 GB of RAM utilization (sometimes well above). If FlexSim's RAM utilization starts approaching a high percentage of your total available system RAM, your computer will turn sluggish or totally unresponsive as it runs out of memory and begins paging out to disk for its memory needs (which is orders of magnitude slower than RAM).
A FlexSim system should have 8 GB of RAM as a minimum. For the best experience we recommend 32 GB of RAM or more. Many modern high-end consumer PCs meet this recommendation. See FlexSim's latest minimum and recommended system requirements here.
Experiments and Optimizations
Experiments and optimizations run several replications of your model simultaneously, so make sure you have enough total memory in your system so that each replication can hit its peak memory needs, while still leaving enough memory for the system and other running applications (OS, antivirus, other background processes, etc.).
In addition, the experimenter saves statistics data for each replication by default. Each completed replication will send its statistics collectors' data to the main FlexSim thread, where it will ultimately be stored in the results database file on the hard drive. However, several stages of the transfer process require two copies of the data. In addition, all child processes can transfer data simultaneously. So, if you have a model that generates 100MB of statistics data, and you run an experiment with 8 CPUs, and all 8 CPUs finish a replication at roughly the same time, you'll need at least 1600MB (100 MB x 2 x 8) to transfer the data successfully, on top of all the other RAM utilization.
I've developed a simulation model and now I'd like to run multiple replications using the Experimenter. Watching my simulation run normally, while monitoring FlexSim's memory usage with Windows Task Manager, I see that over the entire course of its run my simulation peaks at about 2.1 GB of RAM at its highest utilization.
My computer system has 16 GB of RAM. When FlexSim is not running at all, the system uses about 20% of its resources for the operating system and other background processes. This leaves about 12.8 GB of RAM free. How many concurrent replications of FlexSim can my PC support?
12.8 GB / 2.1 GB per replication = 6.1 concurrent replications
So, I might get by with allowing 6 concurrent replications of FlexSim, but I feel like 5 is a safer bet to give the system a bit of headroom, and because running the experimenter and collecting results data also requires some additional memory overhead.
Since my computer has a quad-core processor and uses simultaneous multithreading (SMT) (which allows each core to run two FlexSim replications at once), FlexSim's default would be to run 8 concurrent replications.
My calculations showed that 8 concurrent replications would be too many, and will probably bog down my PC, slowing or halting the entire process. I will limit the number of concurrent replications the experimenter will use by setting the Max Cores value to 5. This option is set from FlexSim's Main Menu under Statistics > Experimenter > Advanced tab > Max Cores input field.
Your settings will differ from this example. Check your computer's amount of RAM, the peak memory needs of your simulation models, and your baseline RAM utilization when you're not running any simulations. Factor in a reasonable % as a buffer. Use these values to determine how many concurrent FlexSim replications your computer may support.
When allowing FlexSim to use its default of running the max number of replications your CPU supports, please keep in mind that many multi-core CPUs actually run multiple threads simultaneously per core (simultaneous multithreading - SMT). To fully take advantage of all the simultaneous replications that your PC could run at once, be sure to get enough RAM for all the replications your CPU is capable of. For example, a 4-core CPU with support for SMT could run 8 simulation replications at once. Make sure you have the memory to do so! If not, be sure to use the Max Cores option to limit the number of concurrent replications.
A system meeting FlexSim's minimum RAM requirement will have enough memory for a single model run of most simulation models. If your simulations are large or complex, or if you intend on running experiments or using the optimizer, you should meet the higher recommended spec of 32 GB of RAM or more.
The FlexSim installer may be up to 1.2GB in size, depending on the FlexSim version.
After installation, FlexSim's program files use up to 1.2GB of disk space. Your model files, CAD layouts, images, custom 3D shapes, import data, exported reports, and other simulation related assets that you provide will take disk space in addition to the software's installation footprint.
Most FlexSim models will run from memory, so hard drive speed isn't really an issue. However, if you're reading/writing with files/databases while the model is running, disk speed could make a huge difference; a Solid-State Drive (SSD) is a must in that case. Alternatively, try to do data reading and writing before and after the simulation run, if possible.
FlexSim recommends SSDs in general because they make every interaction with your computer feel faster and more responsive. You won't regret an upgrade from a regular, spinning hard drive to an SSD.
ConclusionConsider upgrading your hard drive to an SSD if you will be reading or writing to files or databases during a model run, or to generally improve system responsiveness.
The better/newer/bigger/faster the graphics processor, the better the performance. Integrated graphics, where a smaller GPU is packaged into the CPU, have gotten much better in recent years, but a discrete GPU remains the most powerful graphics solution.
FlexSim works best on Nvidia GeForce RTX or GTX GPUs, but Nvidia Quadro and AMD Radeon are also good solutions. Most recent integrated graphics solutions meet FlexSim's minimum graphics requirements, but in general will not perform as well as discrete graphics.
Graphics Driver Updates
It is also important to note that you should keep your graphics drivers updated to the latest version. If your graphics hardware meets FlexSim's minimum requirements but you are experiencing graphical issues in FlexSim, a graphics driver update may solve the problem. Please check out this article for some common symptoms and solutions to graphics issues within FlexSim.
FlexSim supports Oculus Rift/Meta Quest 2, HTC Vive, and Windows mixed reality virtual reality platforms (see demo models). Be sure to meet FlexSim's higher recommended specifications if you plan to use VR.
While recent integrated graphics solutions meet FlexSim's minimum requirements, a discrete graphics solution is a better choice. For maximum performance, FlexSim recommends a recent NVIDIA GeForce RTX GPU.