I think the problem is how the SDD basis set is defined in Gaussian vs. the Basis Set Library. Using "Stuttgart RSC 1997 ECP" in Gaussian and NWChem, I did a single point energy calculation of a given geometry. These are the results:
NWChem: Total DFT energy = -661.868398129143
Gaussian: SCF Done: E(RB3LYP) = -661.868399501 A.U. after 13 cycles