As semiconductor technologies are aggressively advanced, the problem of parameter variations is emerging. Process variations in transistors affect circuit delay, resulting in serious yield loss. Considering the situations, variation-aware designs for yield enhancement interest researchers. This paper investigates to exploit the statistical features in circuit delay and to cascade dependent instructions for reducing variations. From statistical static timing analysis in circuit level and performance evaluation in processor level, this paper tries to unveil how efficiently instruction cascading improves performance yield of processors. Cascading instructions increases logic depth and decreases the standard deviation of the circuit delay. That might improve performance yield of microprocessors. Unfortunately, however, it is found that variability reduction in the circuit level does not always mean yield enhancement in the microarchitecture level.