Partitioned parallel radix sort is a parallel radix sort that shortens the execution time by modifying the load balanced radix sort which is known one of the fastest internal sorts with parallel processing. Parallel sorts usually consist of a few phases of local sort and data movement across processors. In load balanced radix sort, it requires data redistribution in each round for perfect load balancing, whereas in partitioned parallel radix sort, it is needed only once in the rst round. The remaining work is only computation and data movement within each processor, requiring no further interprocessor communication. The proposedmethod has been implemented on IBM SP2, PC Cluster, and CRAY T3E. The experimental results show that partitioned parallel radix sort outperforms the load balanced radix sort in all three machines with various key distributions, by 13 up to 30 in SP2, and 20 to 100 in T3E, 2.5 fold or more in PC Cluster, in the execution time.