Probabilistic seismic demand analysis reflects the transfer process of uncertainty from ground motion intensity measures (IMs) to engineering demand parameters (EDPs). Different probabilistic seismic demand analysis methods are compared in this paper, including cloud method, single strip method and multiple strips method. Based on the real-time interaction of MATLAB and OpenSEES, a probabilistic seismic demand analysis framework based on CPU parallel computing is proposed, which significantly improves the analysis efficiency. Combined with CPU parallel computing, the parallel-cloud method, parallel-single strip method and parallel-multiple strips methods are proposed. Through probabilistic seismic demand analysis of RC frame-shear wall structures, the seismic fragility analysis of RC frame-shear structures with different floors is carried out. The research shows that the logarithmic standard deviation of cloud method is generally greater than that of single strip method and multiple strips method. The seismic performance of the structure is overestimated. There are differences between the single strip method and the multiple strips method in determining the median value and the logarithmic standard deviation for the seismic fragility function. The logarithmic standard deviation of the single strip method is generally lower than that of the multiple strips method.