The invention relates to a full-pipeline high throughput rate accumulator and a data processing method thereof. The accumulator comprises an adder unit which comprises n stages of pipelines, is provided with two data input ports A and B, and is used for carrying out addition operation on the input data of the data input ports A and B; an input delay module which is used for receiving and caching the inputted original data; an output delay module which is used for caching an intermediate result outputted by the adder unit; a first input selection network which is used for selecting a data source of the data input port A, wherein the data source of the data input port A is one of output data of the input delay module, the adder unit and the output delay module; and a second input selection network which is used for selecting the data source of the data input port B, wherein the data source of the data input port B is one of the output data of the input delay module and the output delay module. According to the present invention, the full-pipeline, high throughput and real-time processing of the data processing are achieved, and the problems that an existing accumulator is low in throughput, a production line is blocked, and a large number of caches are needed, are solved.