This study develops the global data envelopment analysis (DEA) model and considers non-performing loans (NPLs) as the undesirable output variable to discuss the overall efficiency of 16 A-share listed commercial banks a total of 128 observations in China from 2007 to 2014. Then, this study extended the concept of total-factor energy efficiency (TFEE) from Hu and Wang (2006) to discuss the disaggregate efficiency of each bank to analyze the causes of banks' inefficiency. Finally, the study investigates the relationship between the disaggregate efficiencies of NPLs and interest income. The empirical results find that the Bank of China and the city commercial banks have better overall efficiency performance. The disaggregate efficiency of NPLs is the lowest among the three output disaggregate efficiencies for most banks, so high NPLs are the major cause of poor overall efficiency. The regression analysis results show a positive relationship between the disaggregate efficiencies of interest income and NPLs. It means the higher NPLs will accompany the lower interest income. This could result in the lower profitability and higher risks of the commercial banks. Therefore, to effectively improve the efficiency of banks, China should start from the bottom to improve business operation profitability and prevent NPLs from being too high.