A Performance Study of a Dual Xeon-Phi Cluster for the Forward Modelling of Gravitational Fields

Figure 17

Behaviour of the speed-up when using ten nodes of the cluster. The serial fraction using the Xeon CPU is around 1% and for the Xeon-Phi, 4%; this means that the Xeon-Phi nodes introduce more overhead although they are faster; the speed-ups are calculated using their corresponding baseline, that is, the computation time used by one node.