Hi Jim,
As I understood from the forum and our system administrator, the Duke Compute Cluster, which is a shared cluster using virtualized, hyper-threaded CPUs, is not an ideal platform for the CESM MPI code. Specifically, as I tested B1850 cases on the Duke Cluster with various CPU configurations, the maximum speed could only be 2.4 simulated year per day, which is far below our expectation.
Now I'm trying running CESM2.1.5 on the TACC Stampede 3. Is this version fully supported on Stampede 3? If so, does it mean that there have already been a full set of prerequisite softwares for running the model on the Stampede 3?
Thanks,
Xiang
As I understood from the forum and our system administrator, the Duke Compute Cluster, which is a shared cluster using virtualized, hyper-threaded CPUs, is not an ideal platform for the CESM MPI code. Specifically, as I tested B1850 cases on the Duke Cluster with various CPU configurations, the maximum speed could only be 2.4 simulated year per day, which is far below our expectation.
Now I'm trying running CESM2.1.5 on the TACC Stampede 3. Is this version fully supported on Stampede 3? If so, does it mean that there have already been a full set of prerequisite softwares for running the model on the Stampede 3?
Thanks,
Xiang