In modern data centers, reducing power consumption and power emissions is imperative. According to the International Energy Agency, data centers and the data transmission networks that connect them account for over 2.1% of global electricity demand, and high-performance computing (HPC) is a significant contributor.1 Given the power requirements of modern data centers, it is no wonder manufacturers and data center operators are focused on sustainability. With increased regulation and carbon taxes in many jurisdictions, reducing a data center's carbon footprint is good for the planet and the bottom line. Fortunately, the combination of Altair software and 3rd Gen AMD EPYC™ processors provides manufacturers with easy-to-implement solutions that can put CAE data centers on a path to better energy efficiency. Here's how an organization can do it:
1. Leverage More Energy-Efficient Servers
An obvious way to lower an organization's energy consumption is simply using more energy-efficient servers. Boosting throughput per watt can not only reduce power and cooling requirements by reducing the number of servers required to meet performance goals — it increases productivity and efficiency as well. Today, AMD EPYC processor-powered systems deliver the industry's best throughput per watt, holding the top spots in the industry-standard SPECpower_ssj® 2008 benchmark.2 Moreover, AMD is on track to deliver an ambitious 30x increase in energy efficiency for AMD processors and accelerators powering servers for HPC and AI-training from 2020 to 2025.3
2. Reduce Server Counts with Dense, High-Throughput Systems
Higher throughput not only enables greater productivity, it also means that organizations can often run the same workload with fewer servers. AMD 3D V-Cache™ technology is a big part of this advantage for cache-sensitive applications, helping customers reduce server footprint and associated power and cooling costs. In benchmarks conducted by Altair, standard finite-element analysis (FEA) and computational fluid dynamics (CFD) workloads ran anywhere up to 1.5x to 1.8x faster using Altair® AcuSolve® and Altair® Radioss®, respectively, on the latest AMD EPYC™ processors with AMD V-Cache™ technology.4 If an application runs 1.8x faster, the same result can be achieved with a 44% reduction in server capacity.5
3. Boost Efficiency with Smart Scheduling
One of the best steps CAE data center managers can take to boost efficiency is to implement a workload management solution. Altair® PBS Professional® and Altair® Grid Engine® help customers use resources more efficiently and boost utilization. These advanced workload managers maximize throughput and parallelism, keeping every core busy and placing workloads where they run most efficiently. Green provisioning features in Altair PBS Professional can automatically shut down cluster nodes when not in use, which reduces energy costs.6 Some of the world's largest HPC users have turned to Altair schedulers along with AMD HPC architectures and accelerators.
4. GPU-Aware Scheduling
Increasingly, some CAE workloads benefit from high-performance GPUs that offer better performance and power efficiency than CPUs for numerically-intensive calculations. The AMD® Instinct® MI250X accelerator is the world's fastest, delivering 95.7 TFLOPS peak double-precision floating-point performance.7 Altair simulation tools like Altair® AcuSolve® leverage AMD ROCm™ APIs to offload linear algebra computation to AMD Accelerators. AcuSolve supports three generations of AMD Radeon™ Instinct and AMD Instinct™ accelerators and delivers performance increases ranging from 3x to 5x.8
In addition, both Altair PBS Professional and Altair Grid Engine support GPU-aware scheduling, obtaining key metrics from AMD Instinct™ and third-party GPUs and placing jobs to maximize GPU utilization. Both schedulers maximize throughput with core affinity features that bind OS-level tasks to cores close to appropriate GPU execution units.
5. Cloud Bursting
Rather than provisioning for peak capacity on-premises, customers can use cloud bursting to tap eco-friendly cloud resources during busy periods and reduce data center footprint. While sometimes easier said than done, both Altair® Control® and Altair® Navops® make policy-based cloud bursting seamless, enabling customers to augment on-premises capacity with powerful 3rd Gen AMD EPYC™ CPU-based cloud instances. Customers can also use built-in cloud bursting facilities in PBS Professional. Done right, cloud bursting can be not only energy-efficient, it can be friendly to a CAE design center's bottom line.
6. Addressing Both Sides of the Energy Efficient Coin
The techniques discussed above have focused on reducing a data center's footprint. However, there's a whole other side to the energy efficiency challenge. Customers increasingly demand safer, more innovative, and more energy-efficient products. This is where CAE shines, especially when powered by systems based on AMD EPYC processors owing to their leading energy efficiency.9 Altair’s structural optimization tools use efficient simulation-driven approaches helping engineers design greener, more sustainable products. Using these tools, engineers can maximize product performance while minimizing material usage. Based on the idea that reducing is better than recycling, Altair helps customers design long-lasting products that require fewer resources to manufacture and operate. The Altair Enlighten Award honors advancements in sustainability in the automotive industry, recognizing lightweight designs that boost mileage and result in more eco-friendly designs.
These innovations lead to a virtuous cycle. Altair and AMD technologies make CAE data centers more energy-efficient, while the design activities they support may lead to more sustainable products and processes. The resulting products, in turn, help further improve data center efficiency.
How can Altair and AMD CPUs also reduce carbon emissions and energy usage? View the infographic here.
To learn more about third-gen AMD EPYC processors, visit amd.com/en/processors/epyc-7003-series.
To learn more about Altair HPC solutions, visit altair.com/hpc-cloud-applications.
1. International Energy Agency, November 2021 - Data Centres and Data Transmission Networks – 1% of global electricity use is attributable to data centers and another 1.1-1.4% is attributable to data transmission networks.
2. See details in AMD EPYC™ Family Claim Information – EPYC-028.
3. Includes high-performance CPU and GPU accelerators used for AI training and High-Performance Computing in a 4-Accelerator, CPU hosted configuration. Goal calculations are based on performance scores as measured by standard performance metrics (HPC: Linpack DGEMM kernel FLOPS with 4k matrix size. AI training: lower precision training-focused floating-point math GEMM kernels such as FP16 or BF16 FLOPS operating on 4k matrices) divided by the rated power consumption of a representative accelerated compute node including the CPU host + memory, and 4 GPU accelerators. More details at www.amd.com/en/corporate-responsibility/environmental-stewardship.
4. These results compared the latest AMD EPYC 7003 series processors with AMD 3D V-Cache to earlier AMD 7003 series processors. See the article Breakthrough Computing Performance with Altair and 3rd Gen AMD EPYC™ Processors with AMD 3D V-Cache™ Technology for details.
5. A throughput increase of 1.5x corresponds to a 33% reduction in required infrastructure (1-(1/1.5)) to achieve the same aggregate throughput. Similarly, a 1.8x throughput gain corresponds to roughly a 44% reduction in infrastructure - (1-(1/1.8)) = 44%.
6. Altair Unveils PBS Professional® 10, Adds Green Provisioning Feature, Improves Performance and Administrator Controls.
7. Calculations conducted by AMD Performance Labs as of Sep 15, 2021. See AMD Instinct MI250X product information for additional details.
8. Based on a comparison of an AMD Radeon™ Pro VII graphics card vs a processor with 32 CPU cores. See the article AMD Radeon™ Pro VII speeds up steady-state simulations with Altair AcuSolve.
9. See EPYC-028 – AMD EPYC provides leading results on SPECpower_ssj® 2008 benchmark.