Hey, I currently run a general 8-core 16-thread processor, which takes 1/2 day for a simulation to complete. It is very slow when I have to do 50/60 simulation for a project which takes several months to complete. I use ansys for simulation. How would it perform if dual epyc 64 core processor is used.
thanks in advance.
The question isn’t that straightforward; like the simulations you’re probably running, there are a lot of complex interactions with different hardware configurations, and software is designed to take advantage of some of them. Other configurations may actually perform worse, because the rest of the hardware and software won’t be able to use the available cores.
First stop for something like this would be Ansys: they have a page with some technical briefs and initial recommendations for AMD hardware. If you haven’t already, I’d reach out to Support directly, as they’ll be able to recommend hardware to get the most out of your particular use case.
You can always follow up here with details, the community may have some suggestions, but the ability to scale and the way in which it does so is dependent on the software and its usage.
How many degrees of freedom are the simulations and which physics interfaces are they using?
Also have you done a mesh convergence study to make sure you’re not over meshing the problem? That many permutations of a simulation deserves a mesh convergence study.
If the number of degrees of freedom is large, tens of millions, you will likely benefit from more memory channels especially and from more cores somewhat.
(post deleted by author)
Rent a VM from GCP/AWS/Azure for a day and check?
GCP c2d-standard-112 is a dual socket EPYC Milan machine with some cores reserved for t he hypervisor.
This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.