Amazon SageMaker HyperPod now supports G7e and r5d.16xlarge instances

This article announces that Amazon SageMaker HyperPod now supports G7e and r5d.16xlarge instances for AI/ML workloads.

• G7e instances feature NVIDIA RTX PRO 6000 Blackwell GPUs with 2.3x better inference performance than G6e
• G7e offers up to 768 GB total GPU memory for deploying larger language models
• G7e provides 1.27x more TFLOPs and 4x GPU-to-GPU bandwidth compared to G6e
• G7e suitable for LLMs, agentic AI, multimodal generative AI, and physical AI models
• r5d.16xlarge provides 64 vCPUs, 512 GB memory, and 3 TB NVMe SSD storage
• r5d.16xlarge ideal for distributed training data preprocessing and feature engineering
• G7e available in US East, Ohio, Tokyo, and US West regions

SageMaker HyperPod now offers enhanced infrastructure options for large-scale foundation model development and deployment with improved performance and memory capabilities.