White Papers

Abaqus Performance
14 Dell EMC Ready Solution for HPC Digital Manufacturing—Dassault Systѐmes’ Simulia Abaqus Performance
4 Abaqus Performance
Abaqus is a multi-physics Finite Element Analysis (FEA) software commonly used in multiple engineering
disciplines. Depending on the specific problem types, FEA codes may or may not scale well across multiple
processor cores and servers. Implicit FEA problems often place large demands on the memory and disk I/O
sub-systems. Abaqus contains several solver options, both implicit and explicit. As such it is difficult to
summarize the overall performance potential of Abaqus with a few benchmarks. Each Abaqus release
distribution does contain some standard benchmarks, both for the implicit solver (Sxx or standard) and the
explicit solver (Ex for explicit). These benchmarks are useful to get an indication of the relative performance
potential for different systems, so should be viewed more qualitatively than quantitatively. Figure 4 shows a
single server performance comparison for four standard Abaqus benchmarks when using all processors cores
on the server, where the value for each benchmark is the solver wall time based on the output at the bottom
of the .msg file.
For comparison, the figure also includes performance data for prior generations of the Ready Solution for
HPC Digital Manufacturing the 13
th
generation system using 16-core Intel E5-2697Av4 processors, and the
14
th
generation system using Intel Xeon Gold 16-core 6142 processors. With the exception of the “s3d”
model, all other models were run with “mp_mode=MPI” and “mp_host_split=8”, which we have found is
typically optimal for systems with more than 16 cores. The “s3d” benchmark is a modal analysis which
operates only in “mp_mode=threads” mode. These results demonstrate that the latest Intel Xeon Scalable
(code-name Cascade Lake) processors 62XX deliver noticeably improved performance other its
predecessors. Typically, the processors with more cores, outperformed processors with fewer cores, while
there were a few cases where the 20-core 6248 outperformed the 24-core 6252. However, these standard
benchmarks are much smaller than typical production size cases, where the additional processors likely will
improve overall performance.
Figure 5 shows the singer server performance for the Abaqus Explicit benchmark models E1-E6, on the same
servers noted above. The benchmark timings were based on the wall clock timing listed at the bottom of the
.sta file.
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
S2A S3D S4B S4D S6
Solver Elapsed Time (sec)
Figure 4: Abaqus Standard Performance
E5-2697Av4 6142 6242 6248 6252