Amazon EC2 P6-B300 instances are now available in the Asia Pacific (Seoul) Region
AWS has launched Amazon EC2 P6-B300 instances in the Asia Pacific (Seoul) Region, expanding high-performance GPU computing options for AI workloads.
Amazon Web Services (AWS) announced the availability of Amazon EC2 P6-B300 instances in the Asia Pacific (Seoul) Region, marking the fourth global region to offer these instances. The instances feature 8 NVIDIA Blackwell Ultra GPUs with 2.1 TB of high-bandwidth GPU memory, 6.4 Tbps Elastic Fabric Adapter (EFA) networking, and 300 Gbps dedicated Elastic Network Adapter (ENA) throughput. AWS states the configuration is designed to support large-scale AI model training and deployment. The instances are priced on a pay-as-you-go basis, with no upfront commitment required.
P6-B300 instances deliver twice the networking bandwidth, 1.5 times the GPU memory, and 1.5 times the GPU TFLOPS (at FP4 precision) compared to P6-B200 instances, according to AWS. The improved specifications aim to reduce training times and increase token throughput for AI workloads, including trillion-parameter foundation models and large language models. AWS highlights the instances' suitability for sophisticated AI techniques requiring high memory and bandwidth.
The instances are available in the p6-b300.48xlarge size, which AWS describes as optimized for memory-intensive and compute-heavy workloads. The p6-b300.48xlarge configuration includes 4 TB of system memory, supporting large-scale AI model training. AWS notes that the instances are now available in the US West (Oregon), AWS GovCloud (US-East), US East (N. Virginia), and Asia Pacific (Seoul) Regions.
Customers can access P6-B300 instances through the AWS Management Console, AWS CLI, or AWS SDKs. AWS provides documentation and pricing details on the Amazon EC2 P6 instances page. The instances are designed for customers running AI training and inference workloads that require high-performance GPU computing and large memory capacity.