I've run two MOM6 experiments that are the same in all regards except their MOM6 version. One was run using the MOM6-examples tag dea81c82 from November 24, 2020 and the other using the MOM6-examples tag cec94e4 from May 20, 2026. Both simulations were run using 90 processors for the ocean, in a fully coupled model configuration. When I compare 5-yr annual means between ocean_month_rho2 output, I find differences in only these variables:
- FrictWork
- Kd_layer_xyave
- T_adx
- T_ady
- S_adx
- S_ady
The spatial pattern of differences suggest an issue with processor alignment (see screenshot).
I also ran an experiment using the same new MOM6 version but with 98 processors instead of 90 processors. The answers are identical whether I use 90 or 98 processors for the new MOM6 version.
I've run two MOM6 experiments that are the same in all regards except their MOM6 version. One was run using the MOM6-examples tag
dea81c82from November 24, 2020 and the other using the MOM6-examples tagcec94e4from May 20, 2026. Both simulations were run using 90 processors for the ocean, in a fully coupled model configuration. When I compare 5-yr annual means between ocean_month_rho2 output, I find differences in only these variables:The spatial pattern of differences suggest an issue with processor alignment (see screenshot).
I also ran an experiment using the same new MOM6 version but with 98 processors instead of 90 processors. The answers are identical whether I use 90 or 98 processors for the new MOM6 version.