Amazon ECS now supports fractional GPU scheduling with Amazon EC2 G6f instances
Amazon ECS now supports fractional GPU scheduling on EC2 G6f instances, allowing tasks to request as little as one-eighth of an NVIDIA L4 GPU to reduce costs for smaller workloads.
The useful question is what changes for users, developers or buyers, and whether the announcement stays industry context or becomes something people can actually use.
Amazon Elastic Container Service (Amazon ECS) has introduced fractional GPU scheduling for Amazon EC2 G6f instances, enabling users to allocate GPU resources in increments as small as one-eighth of an NVIDIA L4 Tensor Core GPU with 3 GB of memory. This feature allows workloads such as small-model AI inference, model experimentation, and graphics rendering to run without requiring a full GPU, potentially lowering infrastructure expenses. Tasks can specify fractional GPU requirements in the container definition using values like GPU=0.125, GPU=0.25, or GPU=0.5, and ECS will place them on compatible G6f instances.
The fractional GPU configuration is supported on both Amazon ECS Managed Instances and Amazon ECS on EC2, providing flexibility in deployment models. ECS Managed Instances offer a fully managed experience, handling instance provisioning, scaling, patching, and lifecycle management automatically. This allows users to focus on workload execution rather than infrastructure operations, while also benefiting from features like GPU metrics via Amazon CloudWatch Container Insights and automatic health monitoring for GPU hardware failures.
With automatic health monitoring, ECS can detect GPU hardware issues and replace unhealthy instances to minimize disruptions to workloads. This capability is currently available in all AWS Regions where Amazon EC2 G6f instances are offered. Users can configure capacity providers with G6f instances and specify fractional GPU values through tools such as the AWS Management Console, AWS CLI, AWS SDKs, AWS CloudFormation, or other infrastructure-as-code solutions.
To implement fractional GPU scheduling, users must define a capacity provider with G6f instances and include the fractional GPU value in the container definition of their ECS task definition. Additional details are available in the Amazon ECS fractional GPU documentation and the Amazon EC2 G6 instance page. This feature aims to optimize resource utilization and cost efficiency for GPU-accelerated workloads.