Servers: The Right Compute
cancel
Showing results for 
Search instead for 
Did you mean: 

How Arm-powered high-performance systems change the economics of chip design

PankajGoyal

The advent of high-performance computing with Arm chips offers a path to fast, economical EDA that can satisfy demand in today's extremely demanding market environment.

HPE HPC-Arm-blog.jpgFor everyone in the business world, competition is heating up. Customers seem to want everything faster, with more features and, of course, at a lower cost.

Modern electronics manufacturing and chip designers provide a particularly dramatic example of what it's like to be under these pressures. New market trends like the Internet of Things (IoT) create opportunities, but the onus is on companies to capitalize on those opportunities quickly. Semiconductor testing has always been difficult and time-consuming, but now it's even more urgent, as chip designers must accelerate the electronic design automation (EDA) verification process to test highly complex system-on-a-chip (SoC) designs and bring them to market more quickly.

Here is the good news: The advent of high-performance computing (HPC) powered by Arm processors offers a path to fast, economical EDA workflows that can meet the demands of today's market environment. Here's how new approaches to HPC are affecting the chip manufacturing business and beyond.

Added pressure on chipmakers

The semiconductor business has always been competitive, but today it is even more so. As the product development cycle accelerates, consumer and commercial electronics need a broader array of processors that can do more. Automobile, appliance, and industrial product manufacturers are innovating feverishly and outfitting their products with new digital capabilities. They need never-before-seen numbers of new chip designs. In the IoT sector alone, industry data forecasts that there could be as many as 75 billion connected devices online by 2025. Consumer electronics companies are similarly broadening their portfolios by offering products such as games, digital watches, and phones.

Likewise, chipmakers are under pressure to produce more than new designs. Customers are demanding more battery life, more safety, more security, and more durability. Chips are getting bigger with more complexity. They have more transistors than ever. Some use cases call for a SoC design, which could contain tens of millions of gates and billions of transistors.

Understanding the increasing demands of EDA

Electronics manufacturers and chip designers use specialized electronic design automation (EDA) tools to design integrated circuits, printed circuit boards, and SoCs relatively quickly. However, the EDA-produced digitized chip design represents just one phase of the product development process. To manufacture the chip, the chipmaker must tape out a photomask for the fabricator. This process can cost millions of dollars.

The final design must be error-free before a company can produce the tape-out. As a result, every chip design must undergo rigorous verification testing before it goes into production. This verification workflow involves a highly iterative engineering process that can take months, and in some cases years, to produce a single integrated circuit. The process grows harder as the chip becomes larger and more sophisticated. When designers add registers and memory elements on an SoC, regression testing requirements increase exponentially.

How Arm and HPE Apollo 70 change the economics of EDA

The market's time and innovation demands, coupled with the increased testing load created by bigger and more complex chips, have generated an unprecedented level of pressure on makers of EDA verification software and server hardware providers. In response, EDA vendors and their hardware partners have devised a transformative solution: the use of high-performance systems for EDA powered by Arm processors.

Building a high-performance compute environment that could tackle EDA workloads took a focused partner approach. For example, Cadence Design Systems, which makes the EDA software solution the Cadence Verification Suite, partnered with Marvell Semiconductor and HPE to offer a cost-effective and performant EDA solution. This solution combines the Cadence Xcelium Parallel Logic Simulation with an HPE Apollo 70 System outfitted with Marvell ThunderX2 processors. These Arm chips use the Armv8-A architecture, which provides high multicore throughput.

The HPE Apollo 70 system provides up to 64 cores and 256 threads of server compute resources. The Apollo 70 architecture includes high memory bandwidth that delivers superior throughput by allowing more jobs to run simultaneously without impacting the performance of each individual job.

Collectively, these distinctive elements of the HPE and Cadence solution are changing the economics of semiconductor device verification. For instance, engineering productivity and time to market improve as extra cores enable multicore waveform dumping. This results in faster diagnostics and problem resolution with little slowdown in total regression time, so companies can go to market quicker with more complex products.

Growing applications for high-performance systems

EDA is but one of several emerging use cases for Arm-powered high-performance systems.

Sandia National Laboratories' "Astra" system, for example, is the world's first Arm-based supercomputer to exceed 1 petaFLOPS, as well as the first Arm-based system to enter the TOP500 list of supercomputers. Featuring a remarkable 2,592 HPE Apollo 70 compute servers in 648 racks, Astra runs more than 138,000 cores with 332TB of memory. It performs 1,758 teraflops per second. Its architecture enables its owner, the U.S. National Nuclear Security Administration, to run extremely complex data-driven engineering and science simulations.

In the UK, the Catalyst Program uses HPE Apollo 70 machines in HPC clusters running SUSE Linux on Marvell's ThunderX2 Arm-based processors. These supercomputers facilitate scientific and economic research at three leading universities.

HPC has historically been used in scientific or research capacities, but the EDA use case demonstrates that supercomputers have a real place in the commercial sector. As consumers and commercial customers continue demanding more high-tech products and expecting to get them faster, high-performance solutions will only become more commonplace.

Interested in learning more about the future of high-performance solutions? Discover what changes Arm-based solutions might be bringing to energy, manufacturing, health sciences, and even climate science.


Pankaj Goyal
Vice President, HPE AI Business
Hewlett Packard Enterprise

twitter.com/pango
linkedin.com/in/goyalpankaj
hpe.com/ai

0 Kudos
About the Author

PankajGoyal

Pankaj is building HPE’s Artificial Intelligence business. He is excited by the potential of AI to improve our lives, and believes HPE has a huge role to play. In his past life, he has been a computer science engineer, an entrepreneur, and a strategy consultant. Reach out to him to discuss everything AI @HPE.