Designing a microprocessor is extremely time-consuming. Computer architects heavily rely on architectural simulators, e.g., to drive high-level design decisions during early stage design space exploration. The benefit of architectural simulators is that they yield relatively accurate performance results, are highly parameterizable and are very flexible to use. The downside, however, is that they are at least three or four orders of magnitude slower than real hardware execution. The current trend towards multicore processors exacerbates the problem; as the number of cores on a multicore processor increases, simulation speed has become a major concern in computer architecture research and development. In this dissertation, we propose and evaluate two simulation techniques that reduce the simulation time significantly: statistical simulation and interval simulation. Statistical simulation speeds up the simulation by reducing the number of dynamically executed instructions. First, we collect a number of program execution characteristics into a statistical profile. From this profile we can generate a synthetic trace that exhibits the same execution behavior but which has a much shorter trace length as compared to the original trace. Simulating this synthetic trace then yields a performance estimate. Interval simulation raises the level of abstraction in architectural simulation; it replaces the core-level cycle-accurate simulation model by a mechanistic analytical model. The analytical model builds on insights from interval analysis: miss events divide the smooth streaming of instructions into so called intervals. The model drives the timing by analyzing the type of the miss events and their latencies, instead of tracking the individual instructions as they propagate through the pipeline stages.