The three real multiplier implementation takes advantage of the pre-adder in the DSP Slice, saving general fabric resources. In general, the three multiplier solution uses more slice resources (LUTs/flipflops) and have a lower maximum achievable clock frequency than the four multiplier solution.