Performance and scaling

Datasets used in EDA workloads are somewhat unique. A typical design engineer’s work area consists of large number of binary files, which are large, often in GB size. The dataset used to test the AWS Cloud infrastructure is a typical to design teams that create high performance SOCs, and includes analog and digital design libraries.

For performance benchmarks, the reference architecture in the following figure was set up in the AWS Cloud with a goal to compare it to a typical, on-premises infrastructure. ClioSoft used a number of standard benchmark tests to compare the two environments, which are similar, but not exactly the same because the options available on the two platforms are different. The results of the comparison of the on-premises environment to the cloud environment showed the typical operations gaining 2–3.5x in performance.

Note that this is not an equal comparison between the two infrastructures since comparable options are not available on the two platforms. The following chart shows the performance speedup of some of key operations observed on EC2 instances for a typical EDA workload.

Performance speedup of EC2 instances for a typical EDA workload