This paper develops an efficient simulation method to calculate credit portfolio risks when the risk factors have heavy-tailed distributions． In modeling heavy tails，the features of return on the underlying assets are captured by multivariate t-copula． Moreover，a three-step importance sampling ( IS) technique is devel-oped in the t-copula credit portfolio risk measure model for further variance reduction． This broadens and enri-ches credit portfolio risk measure models． Simultaneously，the Levenberg-Marquardt algorithm associated with nonlinear optimal technique is applied to estimate the mean-shift vector of the systematic risk factors after the probability measure changes． Numerical results show that IS technique based on t-copula is more efficient and accurate than plain Monte Carlo simulation in calculating the tail probability of distribution of portfolio loss ( or VaＲ of credit portfolio risk under a given confidence level) and that the IS technique can decrease the vari-ance of estimation on the tail probability to a great degree．