EDA and the silicon heart of electronics
Silicon based chips are the heart and soul of virtually every electronic device that improves human lives today. AMD is a leader in providing silicon chips for a wide variety of devices. AMD solutions are used across sensors to supercomputers, cameras to climate modeling. Systems powered by AMD EPYC™ CPUs are used in innovative scientific breakthroughs – from simulating the inner structure of atoms, to mapping the edges of our universe.
Electronic Design Automation (EDA) is a critical part of how AMD designs these technologies for the world. EDA is not a single application, but a diverse mix of high-performance compute (HPC) tools that require powerful hardware. EDA tools are compute, memory, and data heavy. There is no “one size fits all” hardware configuration that works across all aspects of EDA.
Why is AMD IT’s story relevant for you?
AMD is one of the world’s most advanced semiconductor companies. While that could very well be a possible author’s bias, this blog is not intended to highlight the technical excellence of our processors – that theme is effectively addressed by our Marketing and Engineering presentations/blogs/articles.
Here, we'd like to share how AMD IT uses our own technology to help advance our design engineers’ products. We refer to ourselves as “Customer Zero” - we experience the same challenges that our customer EDA IT environments face, and our internal IT solutions address the same problems.
Together, we advance with EDA
EDA workloads are our foundation. They provide a diverse set of challenges. AMD is continuously finding innovative ways to address them. The Chiplet architecture pioneered by AMD drives innovation in year on year improvements that feed compute heavy workloads.
EDA workloads require exponentially more compute every year to support the inexorable march of miniaturization and performance efficiency in silicon chips. Year on year, AMD IT is expected to deliver more computation, power efficiency, higher performance and throughput. We do this by introducing newer AMD CPUs into our production environment as soon as we can benchmark them.
Using current generation silicon to build our next generation products
In our experience running datacenters and cloud environments for many years, AMD EPYC processors have turned out to be the best bet to tackle the diversity of EDA demands. The highest frequency AMD EPYC 9004 CPUs can be paired with very large memory to deliver faster execution times for heavy duty, compute-driven workloads. EPYC 9004 series processors with 3D V-Cache technology can deliver significant performance increases for technical computing applications and high throughput verification workloads . AMD processor core count per socket is also the highest in the x86 industry today. These high core-count SKUs can be used to deliver throughput at massive scale.
Thus, the improvements in every current generation of AMD EPYC CPUs are the raw material for designing the next generation CPUs, GPUs, NPUs, and APUs.
Energy efficiency
AMD silicon are the among the most energy efficient compute you can use for the most demanding compute tasks. The high performance of AMD EPYC processors for massive, demanding workloads is one of the key reasons it was selected as the compute engine of the Frontier supercomputer at the U.S. Dept of Energy’s Oak Ridge National Lab—the world’s first Exaflop computer. Just as critically, the same Frontier architecture topped the Green500 list of energy efficient systems upon its debut in June 2022. Even today, 8 of the top 10 systems (Green500 Nov/2023 rankings) are all based on AMD EPYC – a testament to our goals to make the most efficient compute engines.
AMD partnership approach within
AMD Engineering and IT organizations work together to continuously test and improve the AMD EPYC processor-powered datacenter for every major type of workloads, including EDA.
As customer zero, we partner with our peer teams across multiple aspects of the EDA workflow pipelines, and help them deliver to their roadmaps on time, within budget. Here are some examples of how AMD IT and Engineering groups partner to deliver impact to our internal engineering community and our external customers/partners:
- Evaluate, deploy, and provide feedback on the performance of the latest generation AMD EPYC processors to internal teams.
- Work with our CAD teams to optimize our compute GRID that delivers the infrastructure to design our next generation Silicon. We partner with Cloud providers and Enterprise hardware providers to deliver this capacity both on-premises and multi-cloud.
- Collaborate with Engineering leadership team on the right mix of SKUs to deploy in our GRID for optimal use of EDA tools to deliver the next generation silicon.
- Partner with AMD field support teams by using our large fleet for validating and measuring rare reports of suspected field defects.
- Provide Enterprise Compute/Storage ecosystem with our view on how to modernize datacenters to realize efficiencies and introduce newer workloads like AI with minimal investment.
- Collaborate with our cloud partners to provide feedback on how our production EDA workloads perform on their AMD processor-based cloud instances, so that they can bring those same benefits to their own EDA customers.
- Provide creative IO/compute optimization solutions for EDA that can help EDA customers minimize data storage costs without affecting quality or timelines.
- AMD CIO and Executive leadership team enables innovative efforts to go towards an “unlimited compute” experience for our engineers and enterprise teams alike.
In summary, we work with virtually every part of the EDA tool ecosystem to advance the delivery of the most impactful silicon chips in the industry today. Our teams experiment, share, and learn continuously to improve EDA workloads that are crucial for our business.
Sharing the results of our partnerships
Speaking of sharing, for an in-depth discussion of many diverse types of EDA tools and how AMD performance improvements deliver tangible benefits to EDA customers, see the latest blog on EPYC CPUs excelling at EDA workloads from Raghu Nambiar. You are also welcome to listen in on May 8th as our marketing team gets quality time to talk with Cadence Design Systems on how they are enhancing the popular Spectre® simulation tool in a BrightTalk -hosted webinar describing how AMD & Cadence are accelerating EDA workloads .
From on-premises to the cloud, there is innovation among all our partners. You can see great examples in these stories:
As you may expect, these stories will also feature the recent advancements in AI.
An AI future
It is an exciting time to be in technology today. AI is possibly the most significant technology to appear in the last 50 years. Recent innovations in AI promise to accelerate the chip design lifecycle significantly. EDA tools are adopting more pervasive AI to increase the EDA engineer’s productivity and quality of output. Such advancements are a compelling ingredient in our own internal AMD recipes to create more innovative silicon.
Today, AMD is at the leading edge of the AI wave, in providing CPU solutions for ML and AI inference. We provide leadership GPU solutions for AI, ML, and scientific workloads. We provide AI features across all our silicon.
We use EDA to bring efficient, performant, and energy efficient solutions for every one of your use cases. We are enthusiastic champions of pervasive use of AI to advance EDA semiconductor solutions.
Additional Performance Briefs:
ANSYS CFX https://www.amd.com/system/files/documents/amd-epyc-9004x-pb-ansys-cfx.pdf ,
ANSYS LS-DYNA https://www.amd.com/system/files/documents/amd-epyc-9004x-pb-ansys-ls-dyna.pdf ,
ANSYS Fluent https://www.amd.com/system/files/documents/amd-epyc-9004x-pb-ansys-fluent.pdf,
OpenFOAM https://www.amd.com/system/files/documents/amd-epyc-9004x-pb-openfoam.pdf.
