Figure 4. The average performance improvement (runtime reduction) when estimating the marginal likelihood on a simple phylogenetic model without pre-burnin phase, recorded over 10 repeated runs.

The analyses were performed on the San Diego supercomputer cluster Gordon using 1, 2, 4, 8, 16, 32 and 64 CPUs. The runtimes were measured in seconds. The graph shows the actual and the expected performance increase compared with a single CPU, where performance is nearly linear.