 Abstract Reservoir Computing, RC, is a powerful technique for efficiently processing temporal data with low training costs. This paper demonstrates an all-ferroelectric RC system which consists of a volatile ferroelectric diode, FD, and a non-volatile ferroelectric diode, FD, derived from the same PT-by-FE03-SRU03 structure. The volatile FD with an imprinted electric field, IMP, exhibits short-term memory and non-linearity while the non-volatile FD with no IMP shows long-term potentiation, depression. These two FDs are used as the reservoir and readout networks, respectively, allowing the all-ferroelectric RC system to handle various temporal tasks. The system is able to achieve an ultra-low normalised root mean square error of 0.017 in the Hennan MAP time series prediction, and both the volatile and non-volatile FDs demonstrate long-term stability, high endurance, and low power consumption, making them suitable for use in neuromorphic hardware. This article was offered by Xiwei Chen, Wenjie Li, Chen Fan, and others. We are article.tv, links in the description below.