AWS Graviton2 processor gives ARM higher throughput and reduces local computing

Amazon Web Services (AWS) announced that Arm, a global leader in semiconductor design and intellectual property development and licensing, will use AWS cloud services for a majority of its electronic design automation (EDA) workloads. Arm will lead the semiconductor industry’s transformation journey by migrating EDA workloads to AWS using instances based on AWS Graviton2 processors (powered by Arm Neoverse cores). Traditionally, the semiconductor industry has used on-premises data centers for compute-intensive tasks such as semiconductor design verification. To perform verification more efficiently, Arm uses cloud computing to simulate real-world computing scenarios and leverages AWS’s virtually unlimited storage and high-performance computing infrastructure to scale the number of simulations it can run in parallel. Since beginning its migration to the AWS cloud, Arm has increased the responsiveness of EDA workflows on AWS by 6x. In addition, by running telemetry (collecting and integrating data from remote sources) and performing analytics on AWS, Arm generates more powerful engineering, business, and operational insights that help improve workflow efficiency and optimize costs and resources across the company. After completing the migration to AWS, Arm ultimately plans to reduce global data center footprint by at least 45% and reduce on-premises computing workloads by 80%.

Highly specialized semiconductor devices are increasingly powering everything we do at work and in our lives, from smartphones to data center infrastructure, from medical devices to self-driving cars. Each chip can contain billions of transistors, which are designed down to a few nanometers (about 100,000 times thinner than a human hair) to achieve optimal performance in the smallest space. EDA is one of the key technologies that makes this extreme engineering feasible. The EDA workflow is very complex, including front-end design, simulation and verification, and an increasingly large back-end workload (timing and power analysis, design rule checking, and other applications that prepare chips for production). Traditionally, these highly iterative workflows take months or even years to produce a new device (such as a chip system), requiring a lot of computing power. Semiconductor companies that run these workloads locally must constantly balance cost, schedule, and data center resources to advance multiple projects at the same time, so they may face insufficient computing power, slow progress, or bear the cost of maintaining idle computing power.

By migrating EDA workloads to AWS, Arm has overcome the constraints of traditional hosted EDA workflows and gained elasticity through massively scalable computing power, enabling it to run simulations in parallel, simplify telemetry and analysis, reduce iteration time for semiconductor design, and increase test cycles without affecting delivery schedules. Arm optimizes EDA workflows and reduces costs and time using a variety of dedicated Amazon EC2 instance types. For example, the company uses AWS Graviton2-based instances to achieve high performance and scalability, enabling more cost-effective operations than running thousands of local servers. Arm uses the AWS Compute Optimizer service, which uses machine learning to recommend the best Amazon EC2 instance type for specific workloads, simplifying workflows.

In addition to the cost advantage, Arm also uses the high performance of AWS Graviton2 instances to improve the throughput of engineering workloads, and the throughput per dollar is consistently increased by more than 40% compared to the previous generation of x86 processor-based M5 instances. In addition, Arm uses the services of AWS partner Databricks to develop and run machine learning applications in the cloud. Through the Databricks platform running on Amazon EC2, Arm can process data from various steps in the engineering workflow, generate actionable insights for the company's hardware and software teams, and achieve considerable improvements in engineering efficiency.

“By working with AWS, we’ve focused on improving efficiency and maximizing throughput, freeing up valuable time for engineers so they can focus on innovation,” said Rene Haas, president of Arm IPG. “Now, we can run Amazon EC2 instances based on AWS Graviton2 processors, powered by Arm Neoverse, to optimize engineering workflows, reduce costs, accelerate project schedules, and deliver powerful results to customers faster and more cost-effectively than ever before.”

“AWS provides the truly elastic high-performance compute, exceptional networking, and scalable storage that next-generation EDA workloads require,” said Peter DeSantis, senior vice president of global infrastructure and customer support at AWS. “We are excited to collaborate with Arm to power the most performance-demanding EDA workloads with our high-performance Arm-based Graviton2 processors, which can deliver up to 40% price/performance advantages over current x86-based instances.”