Amazon ECS supports fractional GPU scheduling with EC2 G6f instances
Amazon Elastic Container Service (ECS) now enables fractional GPU scheduling for NVIDIA L4 Tensor Core GPUs on EC2 G6f instances, allowing workloads to use GPU partitions as small as one-eighth. This feature offers cost savings for smaller GPU-intensive tasks like AI inference and experimentation by enabling right-sizing. Support is available across AWS regions via the AWS Management Console, CLI, SDKs, and CloudFormation.
- →Amazon ECS enables fractional GPU scheduling on EC2 G6f instances
- →Fractional GPU requests via container definitions
- →Managed Instances offer enhanced GPU workload support
- →Availability and getting started with fractional GPUs
Features (1) ›
- Amazon ECS enables fractional GPU scheduling on EC2 G6f instances
Amazon ECS now supports scheduling workloads on fractional GPUs, down to one-eighth of an NVIDIA L4 Tensor Core GPU with 3 GB memory, using EC2 G6f instances. This allows for more granular resource allocation for GPU-intensive tasks, potentially reducing infrastructure costs.
Enhancements (2) ›
- Fractional GPU requests via container definitions
Users can request fractional GPUs (0.125, 0.25, or 0.5) directly within the container definition of their Amazon ECS task definitions. ECS will then place the task on a compatible G6f instance that meets the specified GPU partition requirement.
- Managed Instances offer enhanced GPU workload support
ECS Managed Instances now provide improved support for accelerated workloads, including GPU metrics via Amazon CloudWatch Container Insights and automatic health monitoring for GPU hardware failures. This offers a more robust and automated experience for GPU-focused operations.
Notes (1) ›
- Availability and getting started with fractional GPUs
Fractional GPU scheduling is available in all AWS Regions where Amazon EC2 G6f instances are offered. Configuration can be done using the AWS Management Console, AWS CLI, SDKs, or infrastructure-as-code tools.
https://aws.amazon.com/about-aws/whats-new/2026/08/amazon-ecs-fractional-gpu/
Related releases
- Amazon EC2 R8a Instances Now Available in Canada (Central) Region AWS What's New ·
- AWS EC2 Image Builder streamlines Microsoft Office integration Modernizing with AWS Blog ·
- AWS EC2 High Memory U7in Instances Now Available in São Paulo AWS What's New ·
- Amazon EC2 introduces application status checks for app-level monitoring AWS What's New ·
- Amazon EC2 R8i and R8i-Flex instances available in Europe (Milan) AWS What's New ·
- Amazon GameLift Servers adds 21 EC2 instance types AWS What's New ·