8x Blackwell Ultra GPUs: EC2 P6-B300 Arrives in Seoul!
Hey everyone, it's Shii-chan! I've got some big GPU instance news for you today!
AWS What's New
aws.amazon.com
What was announced?
From AWS's What's New feed: Amazon EC2 P6-B300 instances are now available in the AWS Asia Pacific (Seoul) Region. P6-B300 instances pack 8 NVIDIA Blackwell Ultra GPUs and are built for training and deploying trillion-parameter foundation models and large language models (LLMs).
The story so far
P6-B300 was already available in US West (Oregon), US East (N. Virginia), and AWS GovCloud (US-East), but Seoul wasn't an option until now. That meant teams in the Asia Pacific area who wanted this class of massive GPU instance for training had to reach all the way to a US region.
What changes
Now you can spin up P6-B300 directly in Seoul. For teams that need to keep data in Korea or the Asia Pacific area, or that want lower latency, this is great news — you can now build a large-model training setup in a nearby region.
Dive Deep
It's available in the p6-b300.48xlarge size, with specs like these:
- 8x NVIDIA Blackwell Ultra GPUs
- 2.1 TB of high-bandwidth GPU memory
- 4 TB of system memory
- Networking: 6.4 Tbps via Elastic Fabric Adapter (EFA), plus 300 Gbps of dedicated ENA throughput
Compared to the previous-generation P6-B200 instances, P6-B300 delivers 2x networking bandwidth, 1.5x GPU memory size, and 1.5x GPU TFLOPS (at FP4, without sparsity).
It's now available in four regions: US West (Oregon), AWS GovCloud (US-East), US East (N. Virginia), and Asia Pacific (Seoul). Want more detail? Check out the Amazon EC2 P6 instances page.
Wrap-up
- Amazon EC2 P6-B300 instances are now available in the Seoul Region
- Monster specs: 8x NVIDIA Blackwell Ultra GPUs, 2.1 TB of GPU memory, 4 TB of system memory
- Compared to P6-B200: 2x networking bandwidth, 1.5x GPU memory, 1.5x GPU TFLOPS
- Great news for teams training trillion-parameter foundation models or LLMs in the Asia Pacific area
