The new Amazon EC2 DL1 instances, powered by Gaudi accelerators from Habana Labs, an Intel company, are optimized for low-cost training of deep learning models in the cloud. In this report, we present a comparative study of the cost and performance and ease of deployment of this new instance with the most widely used current generation GPU-based training AWS EC2 instances. The use case chosen for this study is the segmentation of liver in medical images using a deep learning model. This paper provides feedback about the experience a developer might have while migrating a training application to the Habana-based Amazon EC2 DL1 training instance.
Our benchmark showed that the Habana-based Amazon EC2 DL1 training instance
delivered from 51% to 73% better price performance compared to the current GPU-based AWS EC2 instances on our deep learning benchmarking training tasks.
In this white paper, learn how our data scientists and data engineers are leveraging the new Amazon EC2 DL1 instance powered by Gaudi® accelerators from Habana® Labs (an Intel company) to train their neural networks.