chris_fletcher@utoronto_ca
New Member
I'm running CESM1-CAM4 at T31 resolution and I've noticed that the model runs ~10-20% slower while performing a transient integration (B20TRCN) compared to a control integration (B1850CN). Looking at the timing summary it appears that the ATM and OCN components are both running more slowly in the transient simulation.
This might be an obvious point, but could someone explain why this is? My understanding was that the differences between B1850 and B20TR were in the prescription of time-varying boundary conditions (CO2, ozone, land use, etc.), but that all model components were active in both cases. So I'm unclear why the transient case would cause such a significant performance hit to both ATM and OCN?
As a footnote, it appears from the CESM timing table that an even larger performance hit is observed for the 1-degree model in transient mode:
1-deg B1850CN on 512PEs = 15.02 ypd
1-deg B20TRCN on 512PEs = 9.54 ypd
http://www.cesm.ucar.edu/models/cesm1.0/timing/
Thanks in advance.
This might be an obvious point, but could someone explain why this is? My understanding was that the differences between B1850 and B20TR were in the prescription of time-varying boundary conditions (CO2, ozone, land use, etc.), but that all model components were active in both cases. So I'm unclear why the transient case would cause such a significant performance hit to both ATM and OCN?
As a footnote, it appears from the CESM timing table that an even larger performance hit is observed for the 1-degree model in transient mode:
1-deg B1850CN on 512PEs = 15.02 ypd
1-deg B20TRCN on 512PEs = 9.54 ypd
http://www.cesm.ucar.edu/models/cesm1.0/timing/
Thanks in advance.