Same Initial Conditions, Different Results.

[cassain]

Hi,

I am setting CAM3.1 for seasonal weather forecast.

I decided to make two equal experiment runs, both using the same initial condition and namelist.

I ran the model for 5 months, and analyzed the accumulated precipitation for the last 3 months.



I used 16 cores to perform the runs.

Can anyone give me a hint as to why these results are so different?

 

[eaton]

All CAM versions produce bit for bit reproducible results when keeping the task/thread count constant. And generally the results are bit for bit identical even when changing task/thread counts. If that's not the case then I'm afraid there is a system problem.