We recently caught up with Ross McLendon, SIMULIA R&D Product Manager for Abaqus Unified FEA, Dassault Systèmes, to learn more about industry challenges and accelerating structural simulation with the newest computing processors from AMD.
Q: What are some of the industry challenges requiring simulation and how does SIMULIA’s Abaqus/Explicit help address these?
Companies in every field are facing challenges on multiple fronts. Economic factors are driving efforts to reduce costs and improve efficiencies while continually shifting consumption patterns, government regulations, and technological advancements create pressure to rapidly bring new kinds of products to market.
Whether it’s designing electric vehicles with driver assistance and autonomous capabilities, better performing and longer lasting tires, or smaller smart mobile devices with unique foldable displays, companies large and small are using SIMULIA’s extensive portfolio of simulation applications for structures, fluids, acoustics, multibody motion, electromagnetics, and design exploration and optimization. As a result, companies are able to quickly evaluate product performance and reduce development time and costs.
The Abaqus/Explicit finite element solver is one of the products driving these advances. It excels at solving highly nonlinear problems – both short-duration transient dynamic events like automotive crash or drop tests for consumer electronics, and quasi-static events like metal forming and material damage that are challenging for implicit solvers but extremely useful for rapidly developing manufacturing processes or assessing the safety of a design. Abaqus/Explicit also includes specialized solution techniques like coupled Eulerian-Lagrangian (CEL) analysis to model problems like tire hydroplaning or ballistic impact; smoothed particle hydrodynamics (SPH) to model problems like sloshing or airbag inflation; and the adaptive remeshing of arbitrary Lagrangian-Eulerian (ALE) analysis to model forming processes with extreme deformation.
Q: Why are compute processors from AMD important for Explicit FEA users?
When I think about how we want to drive users’ experiences with Abaqus/Explicit, three areas come to mind – bigger, better, and faster. I want tomorrow’s user to be able to run larger models than they can today, simulate more complicated problems with greater realism than they can today, and get answers in less time than they can today. The better part of this triad depends on our work in SIMULIA R&D. The bigger and faster parts depend on advances in compute power and AMD is delivering on that front with their newest EPYC processors with 3D V-Cache.
Bigger means that with the high core counts and larger pool of addressable memory available from AMDs latest EPYC CPUs, you can run problems on a single system today that required a large compute cluster or supercomputing system in years past. If you have access to thousands of those same AMD EPYC CPU cores in a large cluster or cloud compute solution, you can get performance scaling on new classes of problems with well over a hundred million elements. And, arguably the most important piece of the triad – faster – means that by leveraging more cores and improved processor architectures you can run your simulation in less time. This lets you leverage SIMULIA process automation and design exploration applications to accelerate your design schedule or simulate more design iterations to achieve a more fully optimized solution.
Q: Specifically, how does Abaqus/Explicit take advantage of new AMD processors?
SIMULIA’s Abaqus R&D team collaborates with AMD teams to verify compatibility of Abaqus/Explicit with AMD’s newest processors, evaluate performance gains on our Abaqus benchmark models, and ensure our code makes the best possible use of new processor developments like increased cache sizes, core interconnects and layouts, and memory bandwidth. AMD’s recent performance brief highlighting composite performance gains across our SIMULIA e13 (car crash) and e14 (drop test) Abaqus/Explicit benchmarks (models available on the DS Knowledge Base) using their latest EPYC processors with 3D V-Cache Technology demonstrates the benefit of these collaborations and processor advances.
In particular, Abaqus/Explicit performance tends to strongly depend on memory bandwidth, which means it substantially benefits from the greater memory throughput offered by the latest AMD processors as well as the large cache sizes enabled by 3D V-Cache, both of which help to keep the processor cores fed with data to reduce runtimes.
Q: How do you ensure Abaqus continues to meet the industry’s latest challenges?
We continually work to improve Abaqus/Explicit to address new classes of problems to be solved. Some recent examples are improving beam technology for modeling hyperelastic lattices or rebar-reinforced concrete columns in buildings subjected to earthquakes, enabling greater parallelism in our mechanics and preprocessing routines, providing contact mass scaling, and improving our mapping and import technology to enable transient multiphysics simulation. We also collaborate with partners like AMD to find new ways to optimize Abaqus/Explicit performance and improve scaling across multiple processor cores. Our goal is to ensure that Abaqus/Explicit excels at helping customers rapidly solve both routine and complex structural engineering problems with ever-larger and faster-running models to get high-quality, innovative products to market faster.
Q: What final words of advice do you have for Abaqus users?
I definitely encourage users to take advantage of the latest versions of Abaqus and improvements in High-Performance Computing (HPC) processors to help accelerate your simulation workflows and overall product development process.
I also suggest investing some time experimenting with different run configurations (different combinations for the CPUS and THREADS parameters) for your models and hardware. As processors advance and processor core layouts, cache organization, and memory connections become more and more complex, overall performance – even when running on a single compute node – can be sensitive to choices like whether to run multiple processes, multiple threads, or a hybrid with multiple multithreaded processes. What worked best on the previous generation of processors may not give the best performance on the latest generation. A little time invested in benchmarking can yield dividends in improved performance.
Finally, get connected. If you’re not yet a member, I invite you join the SIMULIA Community. Register for our structural simulation eNews, and consider attending your local Regional User Meeting where you can get information about the latest developments and best practices and interact with other users and our folks from R&D (and sometimes even the hardware vendors themselves) to help you get the most out of your simulation tools and compute hardware.
Click here to access AMD EPYC Technical Whitepapers and Briefs related to SIMULIA Abaqus.
Interested in the latest in simulation? Looking for advice and best practices? Want to discuss simulation with fellow users and Dassault Systèmes experts? The SIMULIA Community is the place to find the latest resources for SIMULIA software and to collaborate with other users. The key that unlocks the door of innovative thinking and knowledge building, the SIMULIA Community provides you with the tools you need to expand your knowledge, whenever and wherever.