Simulations of complex systems are often hindered by slow relaxations from local energy minima. One method to overcome this difficulty is tempering, in which one attempts to reduce relaxation times by repeatedly heating and cooling the system. Tempering has been especially popular in biomolecular simulations. In this talk I will introduce two algorithms of tempering: serial (aka simulated tempering) and parallel (aka replica exchange or parallel tempering). I will discuss statistical properties of each algorithm and in the end will try to make a case for serial tempering.