Simultaneous Multithreading has been presented as a throughput enhancement paradigm for processor design. The key to the superior performance of SMT processors is the ability to exploit both thread-level and instruction-level parallelism, thereby minimizing the wastage of processor cycles. An important factor that affects the performance is the choice of threads to select instructions for dispatch. There have been numerous heuristics aimed at maximizing the overall throughput of the system. However in a multi-programming workload, fairness is also important as it affects the per-thread observed latency and throughput. In this project we present three new classes of thread fetch policies and we compare the fairness and throughput of the new schemes with other proposed thread fetch policies. Our results indicate that our Sliding Window Schemes achieve a good tradeoff between the fairness and throughput for most workloads, and it does so without degrading the performance significantly.