Ooops. The nshap=6 run now produces NaNs where it used to work. All I did was re-run but with the "latest" pre38 and 1 year time-averaging. Any ideas? I don't see anything significant thats changed in the CVS logs. Is this the sun optimization bug? A.