Assume (a). Then we have 1=dimkMi=nidimkDi, which implies that ni=1 and dimkDi=1. Therefore, Di=k, and we have kG⪯k×⋯×k (since radkG=0 here). Clearly, this gives (b). Conversely, if (b) holds, then ni=dimkDi=1 for all i, and we have dimkMi=nidimkDi=1, as desired.
Comments