Power consumed per unit switching activity in a CMOS based computational unit is greatly dependent on the power consumption of the sum and carry generation units. This paper is focussed on reducing area and PDP of the computational units through gate level and transistor level optimisation. At transistor level 6T Mux based CMOS-CPL XOR gates with stable rise and fall times are used. The concept of gate level Boolean equivalent substitution is used in optimization of logics used in carry generation in a computational unit. Optimised carry block exhibits 50% lesser delay and 45% lesser power consumption compared to a ANDOR based carry generation system. An optimized computational unit at layout level is realised with proposed logical substitutions and with Mux based XOR gates. The resulting computational unit exhibits 60% reduced power consumption compared to a standard realisation. Synthesis of layout and simulations are done by using 45nm technology.