White Papers

Basic Building Block Performance
27 Dell EMC Ready Solution for HPC Digital Manufacturing— Altair Performance
Here, the two node results for the Riser benchmark are artificially high, due to the effect of being able to fit
this model into cache on two BBB’s, but not into the reference 14G EBB server. This effect can be seen as
well for the parallel speedup of this model on a system of EBB in Figure 4. Like Radioss, the BBB shows
excellent overall performance both with Windows and with Linux. Again, the Linux based couplet offers a
noticeable performance benefit over its Windows counterpart.
At the writing of this paper, we have yet to complete OptiStruct benchmarks with the BBBs. Overall, we
believe that OptiStruct workloads will be suitable for BBBs, but not as optimal as they would be as the IBBs,
which have a better balance of CPU, memory, and I/O for a typical OptiStruct workload than the BBB.
However, customers wishing to use the BBB for a diverse CAE workload could use BBBs for such a purpose.
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
Riser Windmill
Performance Relative to EBB
Figure 17: BBB performance for AcuSolve
Linux-1Node Linux-2Node Windows-1Node Windows-2Node