Abstract: Scientific computing is a process wherein the laws and properties of the objective world are represented, predicted and discovered by the computer. With the development of high performance computers, scientific computing has forged rapidly ahead to meet the needs of new applications, and has matured as the third approach to propel scientific discovery and technological innovation alongside traditional theory and experiment. I present my understanding of scientific computing that its essence is to seek the truth, and the degree of confidence is its kernel. It is emphasized that the requirements of the application problem, collaboration among physicists, mathematicians and computer scientists, and the fusion of physics, mathematics and computer disciplines, are needed to accomplish good scientific computing. I also discuss how to heighten the confidence, with some comments on how to improve scientific computing in general.