The truly optimal semiactive control establishes the best possible performance of a dynamically excited and a semiactively controlled isolated structure, subjected to the investigated damper constraints. While it is acausal, and cannot be implemented in practice, it provides a performance goal for implementable control schemes. This paper proposes an implementable predictive control to suppress the earthquake response using a nonlinear semiactive damper. The proposed method uses the Kalman Filtering Technique for the prediction of the near-future ground accelerations. The performance of a 2-degree-of-freedom base-isolated structure is investigated for the proposed predictive control and compared with those of the passive damping, passive on-off cases, two causal semiactive controls, the truly optimal control and the uncontrolled case. The results obtained illustrate that the proposed predictive control shows a good agreement with the ideal best passive case and the causal continuous semiactive control while it follows the truly optimal trajectories closely, except for the peak responses. The proposed control is effective in reducing the base drift without increasing the superstructure response.