Dear gurus,
Recently we run into several problems. We have 20 nodes clusters. CPU: 74F3 two sockets. MEM: 3200 dual rank 64G. Motherboards: h12dsi. 200G connectx6 infiniband. We follow the setting written in "Ansys® Fluent® Performance with AMD EPYC™ 7003 Series Processors". But it does not scale well. The speed up of 10 nodes is 8, 12 nodes is 9, 16 nodes is 11. It is not linear at all. On the other hand, CPU with 7402 scales very well up to 20 nodes.
I am wondering if someone can give us a hint.
Thanks.
K
