flow5 is available for Windows 10, macOS and Linux platforms running on x86-64 bit processors.
mac-OS on ARM processors
Apple announced in June 2020 that the future Macs will run on ARM processors. Until further notice, there is no guarantee that flow5 will be able to run natively on the new platforms. The main question is whether Intel’s MKL libraries will be available for the new OS, and if so what will be their performance.
Updates will be posted here when more information is available.
Processor – x86
Since the program performs intensive calculations for the analyses and in the pre- and post-processing phases, its speed of execution will scale with the computer’s CPU power.
Also worth mentioning is that multi-threading has been implemented as much as possible, which means that the application’s execution speed will increase with the number of processor cores. A multi-core configuration is therefore recommended for performance.
Restriction: flow5 makes use of the Intel® MKL library, which is specifically optimised for Intel processors. Although preliminary testing has shown that the application runs on other processors with comparable performance, please note that there is no guarantee from Intel® that this will always be the case.
Boundary Element Methods such as the ones implemented in flow5 require the construction of large size influence matrices. Unlike matrices used in finite element analysis, these matrices are dense and not sparse. This means that they make use of large amounts of RAM. Consequently, mesh sizes may be limited by RAM availability.
For instance, a mesh with 10,000 elements will require a 10K x 10K matrix that will make use of approximately 1 GB of memory when stored in double precision. Nonetheless, this is something that PCs with standard 8 GB RAM configurations can handle without difficulty.
Testing has shown that an 8 GB configuration can support matrix sizes up to 15,000. At higher matrix sizes, the OS may become sluggish or even unresponsive.
Lower RAM sizes will limit maximum mesh sizes accordingly.
The RAM memory is requested at the start of the analysis, i.e. when the “Analyze” button is clicked, and is freed at the end of the analysis, i.e. when the analysis window closes.
A standard GPU is sufficient to run flow5 and display the 3d graphics, providing it supports OpenGL 2.1 or a more recent version. This should be the case for all graphic cards less than five years old. Again, any GPU that displays xflr5 3d graphics will support flow5. See the troubleshooting guide for more detailed explanations.
At this time, flow5 does not run any calculations on the GPU. There is a plan to investigate whether transferring some tasks to the GPU can improve performance, but given the improvements already made with the use of Intel’s MKL and with multi-threading, this is now far down in the list of priorities.
A recommended configuration such as the one used to run the demos is
- windows 64
- processor Intel i5 4 cores
- 8 GB RAM
- Screen resolution 1900 x 1200
- OpenGL 3.1+
- flow5 will not run on a configuration which does not support xflr5v6 64 bits. Please test xflr5v6 before purchasing flow5.
- win64, macOS and linux 64 bits; flow5 does not run on 32-bits configurations
- OpenGL 2.1 or more recent