KCODE only normalized to power (with 1/keff) to obtain tally absolute value. I'm not sure if I use SDEF instead of KSRC in KCODE problem, do I normalize tally with source strength and the 1/keff factor? Or only the source strength?
Be very cautious here: KCODE problems are not fixed source (i.e., normalized per source neutron) even if the initial source distribution is given by an SDEF code rather than KSRC. In KCODE, SDEF only defines the initial source distribution. After the first cycle, the fission source points generated in the previous cycle are used.
As for normalization of tallies, the tallies in KCODE are normalized PER FISSION NEUTRON. So, you must normalize to the value related to the reactor power, nu-bar, keff, and energy released per fission.