We compare in an analytical way two leader-based and decentralized algorithms (that is, algorithms that do not use a leader) for Byzantine consensus with strong validity. We show that for the algorithms we analyzed, in most cases, the decentralized variant of the algorithm shows a better worst-case execution time. Moreover, for the practically relevant case t <= 2 (t is the maximum number of Byzantine processes), this worst-case execution time is even at least as good as the execution time of the leader-based algorithms in fault-free runs.