Cloud giant launches P6-B200 Elastic Compute Cloud instances.
Amazon Web Services (AWS) has launched Elastic Cloud Compute (EC2) instances featuring Nvidia B200 GPUs.
The P6-B200 instances became generally available on May 15, and are designed for AI, machine learning, and high-performance computing (HPC) workloads - particularly large-scale distributed AI training and inferencing for foundation models.
Each instance features eight Nvidia B200s and has 1,440GB of high bandwidth GPU memory, the 5th Generation Intel Xeon Scalable processors, 2TiB of system memory, and 30TB of local NVMe storage.
According to AWS, the instances offer up to 125 percent improvement in GPU teraflops, 27 percent increase in GPU memory size, and a 60 percent increase in GPU memory bandwidth compared to the P5en instances.
In addition, the instances are powered by the AWS Nitro System, which will make it simple for users to scale AI workload within Amazon EC2 UltraClusters to tens of thousands of GPUs.
The P6-B200s are now available in the US West (Oregon) AWS region.
Last week, GPU cloud provider Cirrascale Cloud Services added the Nvidia HGX B200 GPUs to its cloud offering. Other clouds offering the chips include Vultr and Google.
