Collaboration between OEMs, ISVs, and Intel boost CAE workload performance

[AdvEXperts]

Computer-aided engineering drives the design of innovations that make up our modern lives—from cars to cell phones. Learn how HPE, Intel, and independent software vendors are collaborating to improve CAE performance.

Computer-aided engineering (CAE) drives the design of innovations that make up our modern lives—automobiles, aircraft, ocean-going vessels, trains, even cell phones. The faster manufacturers can identify design flaws, the faster they can bring new products to market.

HPE, Intel, and independent software vendors (ISVs) are working together to improve CAE performance and speed time to insight.

Key takeaways

• HPE and Intel are collaborating to help manufacturers reduce time to market, provide more performance in a smaller footprint, minimize system downtime, and identify CAE performance bottlenecks.
• Substantially boost workload performance by combining 3rd Generation Intel® Xeon® Scalable processors, the HPE Apollo 2000 Gen10 Plus scale-out server, and optimized CAE applications, compared to older processors.
• See improvements across all the critical CAE vectors: compute, memory, and I/O with 3rd Gen Intel Xeon Scalable processors compared to 2nd Gen Intel Xeon Scalable processors.
• HPE Apollo 2000 Gen10 Plus server packs substantial performance and workload flexibility into a small data center space

Manufacturers need more efficient CAE

Today’s manufacturers are increasingly using high-performance computing (HPC) clusters to run CAE workloads. One goal is to shrink development time and cost. But it is equally important for business success to improve product quality, which can lead to greater customer satisfaction.

But manufacturers face challenges in meeting these goals. Modern CAE applications can outstrip the computing capacity of older CPUs. Legacy servers can take up precious data center space and consume more than their fair share of power and cooling resources. Even when the hardware is upgraded, older versions of the most popular CAE applications may not be able to take advantage of newer hardware capabilities.

HPE, Intel, and various CAE vendors are working together to solve these challenges and help manufacturers realize their full potential.

Choose the right hardware for the job

Not all CAE workloads have the same memory, frequency, core count, and I/O characteristics. Therefore, Computational Fluid Dynamics (CFD), Structural Modeling/Finite Element Analysis (FEA), and Mechanical Engineering/Noise, Vibration, and Harshness (NVH) workloads may benefit from different processors. Intel can provide guidance about which 3rd Gen Intel Xeon Scalable processor SKU is the best choice for a particular application.

The server in which the CPUs reside is also an important part of the CAE efficiency puzzle. The HPE Apollo 2000 Gen10 Plus scale-out server offers many benefits, including keeping server footprint to a minimum, lowering TCO, and reducing the cost to the environment (TCE) by shrinking energy consumption. Specifically, this platform is characterized by the following:

• Plug-and-play
• Shared infrastructure
• Highly secure
• Flexible configuration options
• Storage and I/O flexibility
• Comprehensive manageability

Deploying the right compute resources—in the most efficient manner possible—is key to improving CAE workload performance.

HPE tested a variety of CAE benchmarks, comparing the performance of the 16-core, 3.4 GHz 2nd Gen Intel Xeon Scalable processor (35.75 MB L3 cache) and the 18-core, 3.0 GHz 3rd Gen Intel Xeon Scalable processor (39 MB L3 cache). Below are some of the results of those benchmarks.[1]

ANSYS Fluent (Computational Fluid Dynamics)

In more than a dozen industry-standard benchmarks, using the latest Intel processor delivered a 13–18% improvement in jobs per day, with consistently less walltime. Examples of benchmarks included a gasoline direct injection model, a landing gear analysis, and external airflows over an aircraft wing and various types of cars.

ANSYS CFX (Computational Fluid Dynamics)

We ran benchmarks for a LeMans race car, a pump, and three air foils. All benchmarks ran better on the 3rd Gen Intel Xeon Scalable processor, with 19–29% more jobs-per-day and walltimes of up to 24% faster.

ANSYS LS-DYNA (Finite Element Analysis)

We ran three crash simulations, all of which illustrated a substantial increase in jobs per day on the 3rd Gen Intel Xeon Scalable processor—up to 41%—with consistently lower walltimes.

SIMULIA Abaqus/Explicit (Finite Element Analysis)

Three very different benchmarks—an airbag deployment, a thick plate under uniform pressure, and a car crash—demonstrated significant performance improvement on the 3rd Gen Intel Xeon Scalable processor. In particular, the thick plate test ran up to 73% more jobs per day.

SIMULIA Abaqus/Standard (Mechanical Engineering/NVH)

We ran seven benchmarks. 3rd Gen Intel Xeon Scalable processors enabled up to 53% more jobs per day with walltimes reduced by a similar amount.

Are you ready to learn more about how collaboration between HPE, Intel, and CAE ISVs can accelerate your CAE workloads?

Read the white paper to find out more about how HPE and Intel technologies can help drive your business forward.

Advantage EX Experts
Hewlett Packard Enterprise

twitter.com/hpe_hpc
linkedin.com/showcase/hpe-ai/
hpe.com/info/hpc

---

[1] Baseline Configuration: Testing by HPE as of April 2020. 4 nodes, 2x Intel® Xeon® Gold 6246R processor (16 cores, 3.4 GHz/4.1 GHz, 35.75 MB L3, 6 memory channels), 192 GB memory per node (12x 16 GB, 2933 MHz), boot drive: diskless, storage: Lustre Parallel File System, BIOS: version = SAED1229 (01/22/2019) revision 5.14, Thermal Design Power = 205W, Interconnect = InfiniBand HDR, Intel® Hyper-Threading Technology = OFF, Intel® Turbo Boost Technology = ON, Intel® Volume Management Device = disabled, OS = Red Hat Enterprise Linux 7.6, HPC Workload Profile. Ansys Fluent v2020 r1 (Boeing Landing Gear Analysis [landing_gear_15M], External Flow Over Aircraft Wing [aircraft_wing_14M], Flow Through Combustor [lm6000_16M]); Ansys CFX v2020 r1 (Air Foil 100M, Air Foil 10M, Air Foil 50M); Ansys LS-DYNA vR9_3_1 avx2 (3cars, car2car, Neon); Abaqus/Standard v2018 HF14 (S11 Block with Uniform Load , S4b Cylinder Head Bolt-up, S8 Shipping Port); Abaqus/Explicit v2018 HF14 (E10 Plate with Uniform Load, E12 Airbag Deployment, E9 Taurus Crash).

System under Test: System under Test: Testing by HPE as of September 2021. 4 nodes, 2x Intel Xeon Gold 6354 processor (18 cores, 3.0 GHz/3.6 GHz, 39 MB L3, 8 memory channels), 512 GB memory per node (16x 32 GB, 3200 MHz), boot drive: diskless, storage: Lustre Parallel File System, BIOS: version = U47 (08/18/2021) revision 1.52, Thermal Design Power = 205W, Interconnect = InfiniBand HDR, Intel Hyper-Threading Technology = OFF, Intel Turbo Boost Technology = ON, Intel Volume Management Device = disabled, OS = Red Hat Enterprise Linux 7.9, HPC Workload Profile. Ansys Fluent v2021 r2 (Boeing Landing Gear Analysis [landing_gear_15M], External Flow Over Aircraft Wing [aircraft_wing_14M], Flow Through Combustor [lm6000_16M]); Ansys CFX v2021 r2 (Air Foil 100M, Air Foil 10M, Air Foil 50M); Ansys LS-DYNA vR11_2_2_x64_avx13 (3cars, car2car, Neon); Abaqus/Standard v2021 HF6 (S11 Block with Uniform Load , S4b Cylinder Head Bolt-up, S8 Shipping Port); Abaqus/Explicit v2021 HF6 (E10 Plate with Uniform Load, E12 Airbag Deployment, E9 Taurus Crash).

Test setup: The HPE Application Engineer used a script to initialize each test run using Altair PBS Professional as a workload manager. Output files were collected and stored. Number of  MPI tasks: 3rd Generation Intel Xeon Scalable processor: 36, 72, 108, 144; 2nd Generation Intel Xeon Scalable processor: 32, 64, 96, 128.