In the post-quantum Kyber algorithm, the Hash computation directly affects the processing speed of key generation, noise polynomial, and key exchange processes. To address the issues of multiple types of Hash computations, complex variables, and high resource consumption, this paper proposes an efficient Hash computation unit. The proposed scheme first establishes independent data interaction units through the sponge construction, achieving higher parallelism by combining the logically merged Keccak-1600 permutation function. Subsequently, the central binomial distribution sampling and optimized rejection sampling are integrated into the Hash unit, resulting in a compact and highly flexible Hash computation unit. Finally, the scheme is tested and verified under the Taiwan Semiconductor Manufacturing Company (TSMC) 65nm process. The proposed method consumes only 36.4K equivalent logic gates, and experimental results show that the Hash computation cycle is reduced by up to 30%, rejection sampling cycle by 16%, making it more efficient compared to related implementations.