Sign in to my dashboard Create an account
Menu

Accelerate EDA builds by an order of magnitude? Yes please!

fidgeting tool in hand
Table Of Contents

Share this page

Krish Vitaldevara
Krish Vitaldevara
440 views

Chip design is evolving rapidly, accelerated by powerful electronic design automation (EDA) tools. A critical piece of the EDA workflow is data access speed during the design, build, and iteration process. We shared our performance results landing us the top spot in the SPECstorage® Solution 2020 EDA Blended benchmark this past May. Today I am pleased to share that we’ve done it again. NetApp, the leader in high-performance data storage for EDA, has once again proven its superiority by claiming the #1 spot in the SPECstorage Solution 2020_SWBUILD benchmark among results published at www.spec.org as of November 16, 2023.

Not only did we come out on top, we did so by a huge margin. Our tests show the number of builds to be more than an order of magnitude (12 times) as many as the next closest competitor for an 8-node NetApp® AFF A900 cluster and 3.8 times as many for a 2-node NetApp AFF A900 cluster.

Designer time accounts for nearly 70% of chip design costs. Empowering the designers and engineers to be more productive by enabling more build iterations and faster development means that customers improve productivity and can turn designs into products faster. By improving design performance by as much as a factor of 12 compared to the next closest competitor, we’re reducing the build iteration time to less than one-tenth that of the next closest competitor and removing this layer of friction from the process.

What the benchmark measures

What is the SPECstorage Solution 2020_SWBUILD benchmark, and what does it test? SPEC® stands for the Standard Performance Evaluation Corporation, a nonprofit organization that maintains and endorses a standard set of benchmarks for high-performance computers and infrastructure. Various workload definitions are provided for benchmarking; software build (SWBUILD) is one of them. This workload represents metadata-intensive workloads, which are common in EDA. During the test process, metadata for thousands of files is checked; when meeting a criterion, files are read and compiled, and then data is written out to storage. The ultimate metric for the benchmark is the “BUILD,” which represents the number of “MAKE” commands run during a given period specified in the benchmark. Additional information is tracked, such as overall response time (ORT), IOPS, latency, and throughput.

For this benchmark, we tested two configurations: a NetApp AFF A900 8-node cluster and a NetApp AFF A900 2-node cluster. Both clusters were running NetApp ONTAP® 9.14.1, the latest ONTAP release. For the top-line metric, BUILDS, our 8-node cluster achieved 6120 BUILDS with an overall response time of 1.58 ms. Our 2-node cluster achieved 1900 BUILDS with an overall response time of 1.16 ms. In contrast, the next closest competitor achieved only 504 BUILDS with an overall response time of 1.02 ms.

The results: High performance, better scaling

These performance improvements result from a concerted effort by NetApp, in partnership with our customers, to design the highest-performing and most reliable storage system for EDA available.

  • Our benchmark results are achieved on AFF A-Series all-flash systems running ONTAP 9.14.1, the industry’s leading data management software. The ONTAP FlexGroup technology enables large scale-out NAS containers—up to 20PB and 400 billion files—to deliver high performance, with automatic load distribution and scalability.
  • Predictable, consistent high-performance results in more concurrent job sets than any other vendor.
  • The results show better scaling, with continued low latency as demands increase from more SWBUILD workloads.

That high performance and low latency empower designers to run more simulations—faster. With this increased productivity, designers have more iteration cycles to improve the design so that they can deliver chips with better power, performance, and area (PPA) and greater differentiation.

For IT teams responsible for supporting engineer and designer teams, consider what it would mean to achieve an order of magnitude improvement in build times for your teams. To see the results for yourself, you can access a full report of this benchmark. Join us at NetApp.com to learn more about the powerful AFF systems and ONTAP storage operating systems that made these results possible.

specstorage diagram

SPEC® and the benchmark name SPECstorage® are registered trademarks of the Standard Performance Evaluation Corporation. Learn more information about SPECstorage.

Krish Vitaldevara

Krish is a passionate product leader and engineer with demonstrated success at building compelling multiyear strategies for large platform and product teams and delivering significant impact for large enterprises and billions of consumers.

Krish is the senior vice president for Core Platforms at NetApp. The Core Platforms team is responsible for the unified storage platform, manageability platform, Customer Experience Office (CXO), and Chief Design Office (CDO), and it enables the delivery of various NetApp offerings across on-premises, hybrid cloud, and data services. Prior to joining NetApp, Krish spent more than 6 years at Google in multiple roles, including leading product for Android and Plays Trust and Safety and Product for Google Maps. Krish joined Google after a long stint at Microsoft, where he led product teams for O365 Foundations, Outlook.com, and Microsoft Consumer Trust and Safety teams. Early on in his career, Krish worked at a couple of startups—LoudCloud, started by Ben Horowitz and Marc Andreesen, and Brience, which enabled enterprises to customize customer experiences on mobile. Krish holds an MBA degree from Santa Clara University and a master’s degree in Information Systems Engineering from Arizona State University. Krish is also a proven innovator and hacker with more than 30 patents primarily in distributed systems, spam detection models, using graphs and networks for anomaly detection.

View all Posts by Krish Vitaldevara

Next Steps

Drift chat loading