[ares]

Hello,

This might not be the right forum (perhaps I should have posted under hardware), but I’m running into a strange issue with Fluent. We are using an EPYC 9654 system with 192 cores on Windows. We experimented with different numbers of NUMA nodes, but regardless of the configuration, Fluent Meshing seems to be limited to only 8 cores during parallelization. At one point we tried 24 NUMA nodes, but now we’ve gone back to 4.

We also tested on Azure Cloud Services, which provides 4 NUMA nodes (4×44 cores). There, we encountered a similar issue: Fluent Meshing is capped at 44 cores. Even if we launch a second meshing job in parallel, it still uses the same 44 cores.

Has anyone else experienced this problem?

I should also mention that we tried running a Linux VM (limited to 40 cores) on our server, and in that case Fluent was able to use all of the cores. So it doesn’t appear to be hardware-related.