Introduction
Chip designs are increasing in complexity and size, which has resulted in additional transistors driving a need for greater processing power and memory. Greater silicon complexity makes physical verification with electronic design automation (EDA) applications more essential as any delays close to tape-out significantly impact production timelines.
Silicon designers now require increased CPU and memory resources to verify their advanced processor designs. Design rule checking (DRC)and layout versus schematic (LVS) jobs for sophisticated designs can now span several days for a full-chip design and require hundreds or thousands of CPU cores to complete in a reasonable time. Enabling designers with a fast and efficient way to verify their designs helps reduce verification time and costs (including hardware, software, and engineering hours).
On-premise systems have traditionally hosted EDA workloads because they required specific performance, memory, and software parameters. But now, with the scale and flexibility of the AWS Cloud for silicon design implementation and verification, customers can leverage the AWS cloud for their EDA workloads and quickly scale their physical verification jobs and reduce their time to results.
AWS introduced X2iezn instances to address multifaceted EDA workload requirements. AWS recognized that customers performing physical verification on advanced process node designs require higher clock speeds, as well as larger CPU and memory resources. Powered by 2nd generation Intel Xeon scalable processors, X2iezn instances use a frequency of up to 4.5 GHz, the highest in the cloud. They feature up to 1.5TB of memory and deliver up to twice the performance per vCPU than X2e instances. X2iezn instances offer 32GiB memory per vCPU, and enable up to 48 vCPUs and 1536 GiB RAM. X2iezn instances are built on the AWS Nitro System, delivering up to 100 Gbps of networking bandwidth and 19 Gbps of dedicated EBS bandwidth. X2iezn instances are ideal for workloads that require high performance per thread and a high memory. AWS worked with Synopsys, a leader in EDA, to scale Synopsys IC Validator™ physical verification solution on Amazon EC2 X2iezn instances.
This blog post describes the testing that Synopsys performed on X2iezn using IC Validator. Synopsys achieved a 15% performance improvement when executing DRC on X2iezn versus Amazon EC2 R5d instances. Not only did Synopsys see a performance improvement, using IC Validator’s unique Elastic CPU management technology the company was able to obtain a 32% resource savings, enabling designers to save time and optimize resource usage.
Synopsys IC Validator on AWS
IC Validator is a comprehensive and high-performance signoff solution that improves productivity for customers at all process nodes, from mature to advanced. It is a physical verification tool architected for massive and efficient distributed processing. IC Validator is scalable to 1000s of CPU cores, delivering fast performance for full-chip physical signoff and faster signoff convergence. Additionally, Elastic CPU Management technology in IC Validator delivers significant value in the design flow in terms of resource/cost optimization and in accelerating design closure to meet tape-out schedules.
While the run is in progress, IC Validator automatically analyzes serial dependencies in the job, and the job command queue. Using this information, it identifies when more compute resources (i.e. more instances) will make the job finish faster and will subsequently release resources no longer needed. IC Validator Elastic CPU management dynamically optimizes compute resources for physical verification jobs.
Synopsys and AWS collaborated to evaluate the performance of IC Validator workloads on AWS and configure infrastructure using EC2 X2iezn instances. During testing on AWS, Synopsys enabled persistent storage on shared file systems using Amazon FSx for Lustre to maintain data distribution.
Read the full blog to learn more about getting the best OpenFOAM Performance on AWS.
Reminder: You can learn a lot from AWS HPC engineers by subscribing to the HPC Tech Short YouTube channel, and following the AWS HPC Blog channel.