Introducing Amazon EC2 P4d instances
Dave Brown, VP Amazon Elastic Compute Cloud (EC2)
For a ResNet50 model using the TensorFlow framework trained on the Imagenet2012 dataset, P4d instances are 2.1x faster than the P3 instances.
Speech to text
For a Jasper model using the PyTorch framework trained on the LibriSpeech dataset, P4d instances are 2.3x faster than the P3 instances.
Natural Language Processing
For a BERT-Large model using the TensorFlow framework trained on the Wikipedia-corpus dataset, P4d instances are 3x faster than the P3 instances.
EC2 UltraClusters Overview
EC2 UltraClusters of P4d instances democratize access to supercomputing-class performance for everyday ML and HPC developers via an easy pay-as-you-go usage model. EC2 UltraClusters are comprised of -
- More than 4,000 latest NVIDIA A100 Tensor Core GPUs
- Petabit-scale non-blocking network infrastructure
- High throughput and low latency storage with FSx for Lustre
EC2 UltraClusters - How it works
Request more information on EC2 UltraClusters
Any ML or HPC developer, researcher, or data scientist can spin up P4d instances in EC2 UltraCluster using the AWS console. For help or additional questions regarding EC2 UltraClusters, contact us.Contact Us
Getting started with Amazon EC2 P4d instances
P4d instances are supported by all major AWS services such as Amazon SageMaker, Amazon EKS, Amazon ECS, AWS Batch, AWS ParallelClusters, Deep Learning AMIs (DLAMI), and Deep Learning Containers, and Deep Learning Containers. You can also get started with Amazon EC2 P4d instances optimized to run AI & HPC software from the NVIDIA NGC Catalog on AWS Marketplace.Learn More