在雲端加速創新

Leverage the NVIDIA RAPIDS™ Accelerator for Apache Spark within Amazon EMR to accelerate Apache Spark 3.x data science pipelines without any code changes on NVIDIA GPU-accelerated AWS instances. This integration enables data scientists run their extract, transform, and load (ETL), data processing, and machine learning pipelines at massive scale and lower cloud costs by getting more done in less time and with fewer cloud-based instances.

NVIDIA AI software and GPU-accelerated instances can accelerate each step of AI and machine learning workflows within Amazon Sagemaker, including data preparation, model training, and inference serving. To deploy AI models into production faster and lower inference costs, Amazon SageMaker has integrated NVIDIA Triton™ Inference Server, enabling features like multi-framework support, dynamic batching, and concurrent model execution that maximize performance on both CPU and GPU instances on AWS.

A team of experienced scientists and developers at AWS creating Amazon Titan foundation models for Amazon Bedrock, a generative AI service, uses NVIDIA NeMo™, an end-to-end, cloud-native framework for building, customizing, and deploying generative AI models anywhere. 

And the Elastic Fabric Adapter (EFA) from AWS provides customers with an UltraCluster Networking infrastructure that can directly connect more than 10,000 GPUs and bypass the operating system and CPU using NVIDIA GPUDirect®.