Some drawbacks of the latest reaching laws are proposed as follows: the second term of the fal function reaching law is slower and unnecessary compared with the first term; the multi-power terms reaching law is chattering and slower than the sole power term reaching law under bounded input restriction, no matter how parameters are selected, nor how the initial conditions are given. Then, a variable power reaching law is proposed aiming at three goals: chattering free, reaching at sliding mode in finite time, and speeding up convergence. The analytical expression of convergence time and its upper bound are derived, and the sufficient conditions are given. Simulation results show the superiority of the proposed reaching law.