I am running the UTCI comfort map calculation on a large 3D model with over 60k radiance sensor grid points.
I am using accelerad which has definitely helped decrease runtime on tests I did with the model at lower sensor grid counts, but I find the run time takes a big toll once it reaches the “ProcessUtciMatrix” step.
Would anyone know what would be a good solution to reducing the time taken for this step, would this be a simple case of getting a CPU with higher single core performance?
If you’re using your GPU for the Radiance calculations, you might actually be able to make better use of your CPU resources by increasing the number of _workers_ used in the calculation to something that’s a bit more than the number of CPUs on your machine (maybe twice this number). You can do this using the HB Recipe Settings component.
The number of default workers is set up so that you don’t max out your CPU resources doing the Radiance calculations but, given that these calculations are being offloaded from your CPUs with Accelerad, you might benefit from setting a higher number of workers.
Granted, at a certain point, you may need to accept that you’ll need a better machine to run things faster. But the strategy above could probably speed up the calculation time by up to a half for your particular case.