For the over-estimation problem existing in the performance assessment of linear control loops resulted from nonlinearites introduction, a performance assessment method is presented. The bicoherence is used to quantify the nonlinear property presented in the output signals before selected assessment methods. Then the minimum variance performance assessment is converted into model identification with the help of volterra series. Finally, a simulation example indicates that the proposed algorithm gives better minimum variance performance bound compared with existing methods for linear system performance assessment.