AWS has introduced its P3 instances for EC2, super-charging machine learning for businesses running the AWS cloud platform.
AWS's new instances offer up to 14 times the performance compared to previous Amazon EC2 GPU compute instances, making it faster for businesses to deploy machine-learning applications. The speed of the instances means it's much faster to train the machine learning applications, so they can start doing their job instantly, with as little latency as possible.
Each instance comprises one, four or eight Nvidia Tesla V100 GPUs per virtual machine, providing a 300GB/s second-generation Nvidia NVLink interconnect for extremely fast, low-latency GPU-to-GPU communication. Each is supported by 64 vCPUs based on a modified version of Intel's Xeon E5-2686v4 processor.
The instances can be used for a vast array of applications, including computational fluid dynamics, computational finance, seismic analysis, molecular modeling, genomics, and autonomous vehicle systems - systems that need to analyse vast amounts of data, instantly.
"By offering up to 14 times better performance than P2 instances, P3 instances will significantly reduce the time involved in training machine learning models, providing agility for developers to experiment, and optimizing machine learning without requiring large investments in on-premises GPU clusters," said Matt Garman, vice president of Amazon EC2.
"In addition, high performance computing applications will benefit from up to 2.7 times improvement in double-precision floating point performance."
Businesses wanting to supercharge their machine learning VMs can either set them up manually using the AWS Management Console or use one of the two preconfigured Amazon Machine Images - one with Amazon's own developed AMI, or the alternative using Nvidia's AI Cloud Container Registry. Both include support for Apache MXNet, Caffe2 and TensorFlow.
