Number of Cores to use in 2D Modeling

[Scott Miller]

The User’s Manual shows speed results of testing a few data sets with different numbers of cores. What happens in the gap between 8 cores and 16 cores for data sets above 100,000 cells?

What would be the number of cores to use as a function of model size?

[Chris G.]

All projects are different. A lot of it depends on the shape and orientation of your geometric elements. Long and narrow 2D areas will not do as well with lots of cores as a wide 2D area. Also, more cores do better with much larger numbers of cells. There’s a lot behind this and I don’t understand it all. That’s why I suggest people just try their data sets with 1 core, all available cores, and a few numbers of cores in between. Then you’ll know for sure which is best for your project.

[Scott Miller]

Thank you, Chris.

There are definitely some long and narrow parts in my 2D area. I’ll choose a discreet, intense storm and do some speed vs. core testing. (It looks like I should start with comparing 5.0.3 and 5.0.4!)

Please consider touching on how to configure a smart 2D area in your blog. Thanks again.

[Author]

Posts