
AWS announces general availability of Amazon EC2 P5 instances powered by NVIDIA H100 GPUs, delivering up to 6x faster performance and 40% lower training costs for generative AI and HPC workloads.


<div><p>This article announces the general availability of Amazon EC2 P5 instances, powered by NVIDIA H100 Tensor Core GPUs, designed for generative AI and HPC workloads.</p><ul><li>Delivers up to 6x faster time to solution and 40% lower ML training costs</li><li>Supports training and deploying large language models and diffusion models</li><li>Enables generative AI applications: question answering, code generation, image/video generation</li><li>Suitable for HPC applications: pharmaceutical discovery, seismic analysis, weather forecasting</li><li>Features 2x higher CPU performance, memory, and 4x higher local storage than previous generation</li><li>Provides up to 3,200 Gbps networking with second-generation Elastic Fabric Adapter</li><li>Deployed in EC2 UltraClusters supporting up to 20,000 H100 GPUs</li><li>Available in US East (N. Virginia) and US West (Oregon) regions</li></ul><p>P5 instances provide significant performance and cost improvements for demanding AI and HPC applications at scale.</p></div>


Related articles