AWS PCS now supports custom GRES, hardware topology, and MIG
AWS Parallel Computing Service (PCS) now supports custom Generic Resource (GRES) settings, hardware topology, and NVIDIA Multi-Instance GPU (MIG). These enhancements, coupled with Slurm version 26.05, provide finer control over how Slurm schedules cluster hardware. This improves task placement for high-performance computing (HPC) and machine learning workloads, allowing for optimized rank placement and GPU resource partitioning. These capabilities are generally available in all AWS Regions where PCS operates.
Features (1) ›
- Enhanced Slurm Scheduling with Custom GRES, Hardware Topology, and MIG
AWS Parallel Computing Service now supports custom Generic Resource (GRES) settings, hardware topology, and NVIDIA Multi-Instance GPU (MIG). These enhancements, delivered with Slurm version 26.05, enable finer control over cluster hardware scheduling, optimizing task placement for HPC and machine learning workloads. Users can partition GPUs, configure core affinity, and override default settings for flexible resource management.
https://aws.amazon.com/about-aws/whats-new/2026/09/aws-pcs-gres-hardware-topology-mig/
Related releases
- AWS STS Simplifies Session Token Limits and Adds Monitoring AWS What's New ·
- AWS Elemental MediaTailor adds ad response lifecycle hooks for monetization functions AWS What's New ·
- Discover and Govern Snowflake Data with SageMaker Unified Studio AWS Big Data Blog ·
- AWS Step Functions automatically integrates new AWS services AWS What's New ·
- AWS Direct Connect introduces flat-rate pricing for dedicated 10G and 100G connections AWS What's New ·
- AWS Glue zero-ETL now captures archived Salesforce records AWS What's New ·