White Papers

13 Dell EMC Ready Bundle for HPC Digital Manufacturing—Siemens’ Simcenter STAR-CCM+™ Performance | January 31, 2018
All models tests display good parallel speedup up to eight nodes, with the larger models tending to show
better parallel efficiencies than the smaller models, as expected. Two of the models display super-linear
parallel speedup (civil_trim_20m and VtmBenchmark_178M), which can occur when most of the dataset can
be contained within the processor cache when the problem is carried out on larger number of nodes.
Figure 4 displays the parallel speedup for 4 different benchmark inputs for the LeMans racecar.
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
32 (1) 64 (2) 128 (4) 192 (6) 256 (8)
Performance Relative to 32 Cores
Number of Cores (Number of Nodes)
Figure 3: STAR-CCM+ Parallel Scaling—EBB
civil_trim_20m.amg
EglinStoreSeparation
EmpHydroCyclone_30M
EmpHydroCyclone13m
HlMach10Sou
KcsWithPhysics
reactor_9_million.amg
SlidingMorphingNopostHelicopter
TurboCharger-noreports-nostress-combined-amgopt
vtmBenchmark_178M
VtmUhoodFanHeatx68m