Maximize ROI and Performance on the SIMULIA® Portfolio with Intel® Processors
Author: Rod Mach, CEO, TotalCAE
When it comes to running computer-aided engineering (CAE) applications on high-performance computing (HPC) infrastructure, engineers can never get enough performance. Every new generation of processor technology can deliver greater performance and dramatic financial returns in hyper-competitive industries such as manufacturing. However, the road to consistently getting the best performance for CAE applications is fraught with many obstacles.
Challenges operating CAE solutions
Managing HPC clusters can be very difficult. Manufacturing clients don't really want to be experts at HPC. They want to build the highest quality products at the lowest costs and improve productivity and time-to-market.
Often our client’s IT department isn't completely proficient in all the nuances of HPC technology, CAE simulation tools, and all the software and middleware that's needed to make these complex solutions work seamlessly every day for their production needs.
Manufacturers also need excellent support if they do run into issues with their CAE environment. It is hard for them to determine if there is there something wrong with the application or cluster or the Linux operating system or the middleware.
The solution
To help manufacturers overcome these challenges, TotalCAE partners with leading organizations such as Intel®, Lenovo®, and Dassault Systèmes® to provide best-in-class high-performance computing (HPC) solutions for realistic simulation with the SIMULIA® portfolio of CAE applications.
Deep HPC and SIMULIA expertise
TotalCAE has over 17 years of experience in HPC and speeding up the SIMULIA portfolio – particularly SIMULIA Abaqus® – one of the world’s preeminent Finite Element Analysis (FEA) applications for structural analysis.
We make it easy for engineers to use HPC – on-premises clusters or on the cloud. We:
- Install, manage, support, and maintain all SIMULIA solutions on your HPC infrastructure
- Give you an easy-to-use interface so that you can run your SIMULIA Abaqus, XFlow, Isight, and other types of workloads at a push of a button
- Provide comprehensive support to determine the root cause of any issues and then help you get that problem solved so you can focus on getting your work done and meeting your deadlines
- Constantly innovate with our partners to test and optimize the performance of the most current versions of SIMULIA applications on the latest HPC systems.
Relentless focus on speeding up SIMULIA on the latest HPC infrastructure
Speeding up CAE applications requires four things: optimizing CPU, network, memory, and disk. Here we focus on the CPU. There are several techniques to accelerate SIMULIA applications including Abaqus:
- Various flags that you can use on high-core count processors
- Several job stacking performance measurements if you want to put multiple jobs on one server
- Co-simulation with Abaqus, XFlow, and other CAE applications is possible on modern processors and Lenovo HPC clusters.
There are also several processor choices on Lenovo systems to speed up your simulations.
3rd generation Intel Xeon® Scalable Processors on Lenovo for outstanding performance and scale
These processors provide a significant leap in performance for the same license cost compared to the previous generation. Our data shows a markable increase in performance on standard benchmarks and customer benchmarks. For instance, the classic Abaqus 2022 s4B benchmark saw a 34% increase in performance on the same number of cores. Similarly, on an explicit model, e1 has a 36% improvement for the same license cost.
In addition to benchmarks, real customer models show a 25% improvement for the same license cost. Over the last year, we deployed third-generation Intel Xeon Scalable Processors to many happy SIMULIA customers with significant improvement in performance and throughput.
These Intel processors also enable great scalability for Abaqus 2022 across numerous CPUs on HPC clusters for a single job. So, it is possible to accelerate larger models like e13 and s4e and get great scalability to significantly lower time-to-results. The new 4th generation Intel Xeon® Scalable Processors that Intel launched this week will continue to offer increased performance for Abaqus.
This outstanding performance significantly enhances value for engineers and manufacturers.
Better financial returns with higher performance with SIMULIA 2022 enhancements
In addition to lowering the cost of the application licensing, performance enhancements can help your engineers get more work done, improving the productivity of your human resources and enterprise.
However, it is not just about the processors and the hardware. If you use the latest hardware without the correct setup configuration options, you may not get the performance and reliability needed to run your Abaqus jobs. The whole system needs to be optimized to get the best performance, which is what TotalCAE helps you do.
It also helps if you are on the latest SIMULIA release. There are always big performance gains from release to release, and a range of new functions and features that you can leverage.
For instance, when you have many CPUs in a node and spread a workload across many nodes, you get significant communication overhead that hurts performance and scalability. The 2021 Abaqus version has a new hybrid message parallel functions that use a hybrid threading model to pass data more quickly between the cores and then use a message-passing interface (MPI) between nodes.
This optimization reduces communication overhead, increases scalability, and can make a big difference in the performance of your job. On the same hardware, the same version of Abaqus, you can get anywhere from 20% to 40% speedup just by running Abaqus with the right options. So, when you go from a workstation to a high core count cluster, there are additional, special options to boost performance.
Further, with so many cores now in a single node, you may be interested in stacking jobs to get even more efficiency from a single node. For example, if you typically only run 16 core jobs because it is cost-effective from a licensing-per-job perspective, you can run several of those on a node. With the increased memory bandwidth on these new Intel 3rd and 4th generation processors, there is a very small penalty from job stacking, so you can stack multiple jobs on one single server and get greater throughput from your cluster. You can choose job stacking on the TotalCAE platform with a simple check box depending on your specific model and requirements.
There are several solvers in the SIMULIA portfolio that many of our customers are coupling to solve larger, more complex, advanced multi-physics problems. For example, some clients are coupling XFlow, a computational fluid mechanics (CFD) solver, together with Abaqus in a single run. XFlow can use one set of cores or CPUs on your hardware while Abaqus can use another group. The TotalCAE platform provides a simple way for this. You just pick how many cores you want for CFD and how many for Abaqus. Then you push a button, and the system takes care of laying this all out and automates the execution.
Lenovo HPC clusters with 3rd Intel Xeon Scalable Processors lower limits on the number, size, and complexity of simulation models while delivering faster time-to-results, and more innovation, productivity, revenues, and profits. What’s more, they are affordable contrary to the myth that HPC is expensive.
TotalCAE makes operating a Lenovo CAE cluster simple and economical
There’s a prevalent belief that HPC clusters are only for organizations with large budgets. The reality is that even a mid-size or small company can afford HPC clusters. They have become quite ubiquitous with pricing starting as little as \\\$1000 a month for a capable HPC cluster.
Lenovo provides the workload-optimized building blocks for the cluster including compute, visualization nodes, storage, and networks that TotalCAE can help you assemble, maintain, and manage. TotalCAE also includes TotalCAE platform which includes monitoring, management, scheduler, submission portal, and analytics software needed to ensure the seamless operation of the HPC cluster
TotalCAE also installs the SIMULIA software suite, all the updates, hotfixes, and whatever is required. Lastly, we work with our partners and provide comprehensive support to determine the root cause of any issues with hardware, software, or applications and then help you get that problem solved with 1-hour response times. So, this turnkey CAE cluster solution is very affordable and simple to operate.
We manage numerous award-winning HPC solutions from the Fortune 10 all the way to small engineering firms. For more information, visit www.totalcae.com.
Checkout the webinar from December if you missed it below:
