Diffusion models have recently exhibited impressive performance in time series analysis. In this paper, We focus on accelerating diffusion models for the task of time series forecasting and put forward a multivariate time series probabilistic forecasting model with implicit diffusion and simplified state-space layer (IDSS). Specifically, IDSS leverages a deep neural network known as WaveNet and integrates denoising diffusion implicit models (DDIMs) to enhance the inference speed while maintaining predictive accuracy. Furthermore, it employs a simplified state-space layer (S5) to efficiently capture the long-range dependencies in time series, contributing to faster and more accurate prediction of time series. Experimental results demonstrate that our method can rapidly produce high-quality prediction results with high computational efficiency.