PC Specifications for Simscape Power Electronics

I have been given the chance to specify a PC for Model Based Design of power electronics. At the moment, mostly running Simscape models of power electronic converters using the "Specialized Power Systems" library. Typically running the Simulink model and powergui solver at 10ns timestep, runtimes in excess of 100ms. I am also interested in cosimulation of VHDL modules using Modelsim.

For this specific use case, what are the key factors affecting execution time? I am looking to specifiy a processor, RAM and SSD arrangement. My current PC meets the "recommended specs" but ideally looking to go well beyond that and build something optimised for full-time professional use.

Answer 1

[Sabin]

You can use MATLAB R2022b System Requirements for Windows/Mac/Linux as a starting point for upgrade. That would be good enough to run any MathWorks toolbox today. From there it really depends how much do you plan to invest; sky is the limit. A 6-core CPU and 64GB RAM should be a good option for engineering.

[Euan]

Thanks for you reply. I am familiar with the receommendations in the link. However, I am looking to specify something well beyond the minimum spec since I spend much of my week running simulations. I wondered if there were any specific recommendations applicable to Simscape Power Electronics models. For example, whether core count was more important than clock speed, whether they were particularly RAM hungry, whether they need a fast SSD etc.

At the moment, I'm basing the optimum specification off the top listing on the benchmark tool.

[Joel Van Sickel]

Hello Euan,

You will see the biggest benefit from a high end processor at a fast clock rate. Number of cores is not as important except for running multiple simulations at the same time.

I'd recommend extra RAM, for convenience. I don't have an official number, but I'd say for most cases, 64 is overkill while 4 to 8 is going to slow you down assuming you are doing other things at the same time with your computer.

Any SSD should be fast enough, but having lots of space is convenient. If you are using VHDL tools, installs between MathWorks tools and the FPGA toolchain can easily be over 30 GB. Also, if you are doing any synthesis of the FPGA bitstream, you can expect an extra 2 GB per bitstream. I'm trying to over estimate for these to provide conservative estimates.

Answer 2

[Euan]

Thanks for your replies. We're leaning towards an i9-12900K with 64GB RAM and 1TB SSD which ties in well with your recommendations and keeps the other engineering disciplines using different tools happy as well!