Hybrid MPI and CUDA Parallelization for CFD Applications on Multi-GPU HPC Clusters

Table 4

The runtime for Tesla V100 multi-GPU clusters.

No.	Two GPUs (ms)	Three GPUs (ms)	Four GPUs (ms)
Mesh 1	3.85	3.41	3.01
Mesh 2	7.16	6.58	5.86
Mesh 3	13.78	11.82	9.31
Mesh 4	23.99	20.43	15.8
Mesh 5	41.59	32.22	25.71
Mesh 6	78.69	58.81	45.16
Mesh 7	151.78	108.85	83.92
Mesh 8	295.13	208.72	161.88