Source: Tony Webster from Minneapolis, Minnesota, United States [CC BY]

AWS announced on Monday, the launch of next-generation GPU equipped EC2 P4 instances for machine learning and high performance computing. These instances will be powered by NVIDIA’s new A100 tensor core GPUs and Intel Cascade Lake processors.

Amazon Web Services (AWS) launched its first-generation Cluster GPU instances back in 2010, and the previous gen was launched in 2019. According to AWS, the upgraded P4 Cluster GPU instances can deliver up to 2.5x the deep learning performance, and up to 60% lower cost to train when compared to P3 (previous-gen) instances.

The due upgrade comes after NVIDIA launched its next-generation GPUs this year which took the computer industry by storm. The release included the RTX 3000 series GPUs for personal computers and ampere architecture based data center GPUs for AI, data analytics, and high-performance computing (HPC).

Each cluster will be powered by an Intel Cascade Lake CPU and 8 A100 NVIDIA tensor core GPUs connected with NVLink and support GPUDirect. These instances offer 2.5 PetaFLOPS of floating point performance and 320 GB of high-bandwidth GPU memory. They include 1.1 TB of system memory and 8 TB of NVME-based SSD storage that can deliver up to 16 gigabytes of read throughput per second. According to AWS, the customer gets access to 100 Gbps network connections to a dedicated network designed specifically for P4, along with 19 Gbps of EBS bandwidth that can support up to 80K IOPS.

Currently, the Cluster GPU instances are only available in one size: p4d.24xlarge, which includes 1 CPU and 8 A100 GPUs. They are available in US East (Northern Virginia) and US West (Oregon) regions.

AWS also announced the ultra-powerful EC2 UltraClusters, which comes with 4,000 or more A100 GPUs. This is an AWS ParallelCluster, which is mostly targeted for large businesses and corporations. “These clusters can take on your toughest supercomputer-scale machine learning and HPC workloads: natural language processing, object detection & classification, scene understanding, seismic analysis, weather forecasting, financial modeling, and so forth,” wrote Amazon, in P4 press release.

The market for machine learning and AI is rapidly growing and expanding, and some experts even go as far as to say that AI is the future of technology. On the other hand, with the growth in size of available data, data processing tasks are becoming more and more power hungry. With AWS Cluster GPU instances powered by the latest GPUs and CPUs built for data centers, customers can remotely harness all the power they require to train their AI models or to process and analyze large amounts of data without having to invest in the hardware themselves.

Customers can buy the available P4d instances as On-Demand Instances, Reserved Instances, Spot Instances, Dedicated Hosts, or as part of Savings Plan. As of now, the on-demand price is $32.77, but 1-year reserved instance will cost $19.22 and 3-year reserved $11.57.