Amazon EC2 P4d instances for pushing machine learning boundaries now available

Amazon Web Services announces general availability.

Amazon Web Services (AWS) has announced the general availability of Amazon Elastic Compute Cloud (Amazon EC2) P4d instances with EC2 UltraClusters capability.

The cloud giant has touted the offering as the next generation of GPU-powered instances for providing the best performance in machine learning training and high-performance computing (HPC) in the cloud.

Applications include natural language processing, object detection and classification, seismic analysis, and genomics research.

These instances are powered by the latest Intel Cascade Lake processors and feature eight of the latest Nvidia A100 Tensor Core GPUs, each connected by NVLink and with support for Nvidia GPUDirect.

AWS said they boast 2.5 petaflops of mixed-precision performance and 320GB of high bandwidth GPU memory in one EC2 instance.

It claims the increased performance of P4d instances will speed up the time to train machine learning models by around 2.5x and the additional GPU memory will help customers train larger, more complex models.

"Using P4d instances with AWS's Elastic Fabric Adapter (EFA) and Nvidia GPUDirect RDMA (remote direct memory access), customers are able to create P4d instances with EC2 UltraClusters capability," AWS said.

"With EC2 UltraClusters, customers can scale P4d instances to over 4,000 A100 GPUs … by making use of AWS-designed non-blocking petabit-scale networking infrastructure integrated with Amazon FSx for Lustre high performance storage, offering on-demand access to supercomputing-class performance to accelerate machine learning training, and HPC."

P4d instances are built on the AWS Nitro System.

AWS said customers can run containerised applications on P4d instances with AWS Deep Learning Containers with libraries for Amazon Elastic Kubernetes Service (Amazon EKS) or Amazon Elastic Container Service (Amazon ECS). Customers can fully manage use of P4d instances via Amazon SageMaker.

P4 instances are available in one size -- p4d.24xlarge -- and customers can launch them in the US East (North Virginia) and US West (Oregon) Regions today. Nvidia A100 drivers and the most recent ENA driver are required.

LATEST FROM AWS