Introducing Amazon EC2 P2 Instances, the largest GPU-Powered virtual machine in the cloud
We are excited to announce the availability of P2 instances, a new instance type designed for compute-intensive applications that require high-performance GPU coprocessors and massive parallel floating point performance. P2 instances are ideal for deep learning, computational fluid dynamics, computational finance, seismic analysis, molecular modeling, genomics, and rendering workloads. For graphics workloads requiring DirectX and OpenGL, you can continue to use G2 instances.
P2 instances are the most powerful GPU virtual machine in the cloud, with up to 16 NVIDIA Tesla® K80 GPUs, 192GB of total video memory, 40 thousand parallel processing cores yielding 70 teraflops of single precision floating point performance and over 23 teraflops of double precision floating point performance. P2 instance GPUs are connected to the same PCI fabric, reducing the latency of GPU to GPU transfers by up to 70%, and enabling GPUDirect™ (peer-to-peer GPU communication) among up to 16 GPUs in a virtualized environment for the first time. The new instance type also features 732GB of host memory, up to 64 vCPUs using custom Intel Xeon® E5-2686 v4 ( Broadwell ) processors, and enhanced networking with the Amazon EC2 Elastic Network Adaptor featuring up to 20Gbps of aggregate network bandwidth within a Placement Group.
New P2-optimized AMIs come pre-installed with MXNet and Caffe, and are now available in the AWS Marketplace to help you get started quickly on deep learning and other GPU-intensive workloads. To get started, visit the AWS Management Console, AWS Command Line Interface (CLI), AWS SDKs, and third-party libraries. To learn more, visit the Amazon EC2 page.
P2 instances come in three sizes and are available in the US East (N. Virginia), US West (Oregon), and EU (Ireland) Regions. Visit the Amazon EC2 Pricing page for more details.