In this paper, based on the low-power version of parallel Booth multiplier we will show one novel sub-word parallel (SWP) multiplier and one low-error reduced width multiplier. The new SWP multiplier can finish one n by n multiplication or two n/2 by n/2 multiplications or the multiplication of two n/2 complex numbers in the same architecture. In the low-error reduced-width multiplier, an input number dependent compensation vector is derived that is superior to other proposed results in error performance and operation speed. The implementation results of 16 by 16 to 16 case show that the area is 50.67% and the critical path delay is 66.18% of the original Booth multiplier.