AWS Parallel Computing Service adds NVIDIA Blackwell UltraServer support
AWS Parallel Computing Service (PCS) now supports the new Amazon EC2 P6e-GB200 and P6e-GB300 UltraServer instances, which utilize NVIDIA Blackwell architecture. This enhancement allows customers to run large-scale GPU workloads within Slurm-managed clusters, simplifying HPC deployment and scaling. The new servers offer significant GPU memory and compute capabilities, with availability in regions supporting both PCS and EC2 Capacity Blocks for UltraServers.
- →Support for NVIDIA Blackwell-based EC2 UltraServers in PCS
- →Simplified GPU workload deployment with UltraServers
- →Enhanced compute and memory capabilities with UltraServers
- →Managed HPC environment with PCS
Features (1) ›
- Support for NVIDIA Blackwell-based EC2 UltraServers in PCS
AWS Parallel Computing Service (PCS) now supports Amazon EC2 P6e-GB200 and P6e-GB300 UltraServer instances. These servers enable running large-scale GPU workloads using NVIDIA Blackwell architecture within Slurm-managed clusters.
Enhancements (2) ›
- Simplified GPU workload deployment with UltraServers
Customers can use EC2 Capacity Blocks for UltraServers and associate them with PCS compute node groups. PCS automatically configures Slurm with the correct topology plugin for seamless integration.
- Managed HPC environment with PCS
AWS PCS simplifies running and scaling HPC workloads on AWS using Slurm. It manages cluster operations, including updates and observability, to build elastic environments integrating compute, storage, networking, and visualization.
Notes (1) ›
- Enhanced compute and memory capabilities with UltraServers
P6e-GB200 UltraServers offer up to 72 NVIDIA Blackwell GPUs with substantial FP8 compute and HBM3e memory. P6e-GB300 UltraServers provide increased GPU memory and FP4 compute performance compared to the GB200 variant.
https://aws.amazon.com/about-aws/whats-new/2026/06/aws-parallel-computing-service/