luoyangcheng96@pku_edu_cn
New Member
Hello everyone, I'm running CESM1.2.2.1/CAM5 in F_2000_CAM5 compset on Cheyenne. In order to optimize the speed and to lessen the loss in the connection between nodes, I carried out a node number test for the 1-degree resolution and the 0.5-degree one. Theoretically the loss in connection should be less in a finer resolution than in a coarser one. However, the results of my tests turned out to be opposite. Attached is the normalized production rate of the both resolutions. We can see when using 6 nodes, the loss for the 1-degree run is ~12%, but for the 0.5-degree run the loss is ~24%. What might be the possible reason for this issue? Any suggestions may help. Thank you Yangcheng