Our simulation results show that for memory-intensive, multiprogramming workloads, the new policy improves the overall performance by 10.7% on average and up to 17.7% on a four-core processor, when ...