In this paper, The ARMA-GARCH、ARMA-EGARCH and ARMA-TARCH  model under the assumption of that the random error is Normality , Student t or GED distribution is used to analyze the asymmetry volatility of China stock market. The samples are the close prices of shanghai and Shenzhen A share index from December 1995 to March 2008.The result suggests that: 1) both Shanghai and Shenzhen A share stock market have the asymmetry of information on volatility of the stock return rate; 2) in the Maximum Likelihood norms and principles of AIC, ARMA-EGARCH is the optimal model to describe the China A-shares index volatility.