AWS EC2’s New GPU Instances for Customers That Need High-Performance Graphics

As soon as you think about cloud computing, high-performance 3D graphics are probably not the first thing that comes to attention, how much ever computing power the streaming graphics-intensive applications and 3D visualizations require, this GPU instance is a logical step for AWS.

AWS launched elastic Graphic Processing Unit or GPU with existing EC2 instances. Like EBS that can easily attached & detached with EC2, Elastic GPU is connect to EC2 Instance as a network component. Amazon Elastic GPUs are created as soon as you launch EC2 instances and state the size of Elastic GPUs which you want to use. Using these new instances, Amazon claims, its users can now build high-performance DirectX, OpenGL, CUDA, and OpenCL applications and services without making any up-front capital investments.

Elastic GPUs are made in the same Availability Zone as your EC2 instances, and could be of 1 GiB to 8 GiB in GPU memory size. Once it is created and attached, the GPU driver will identify the presence of Elastic GPUs and launch a connection between your EC2 instances and Elastic GPUs. At this point, the instance can interact with the Elastic GPUs’ OpenGL library in much the same way as locally attached GPUs.

Using Elastic GPU’s

Elastic GPUs are not a part of any hardware instance, they are actually attached through an elastic GPU network interface in the subnet which is created when the instance is launched with an Elastic GPU. An Elastic GPU can simply attach to one EC2 instance, and each EC2 instance can have only one Elastic GPU attached to it. Elastic GPUs can be used by merging many desktop streaming protocols, like Virtual Network Computing (VNC), Desktop Cloud Visualization (DCV), and Remote Desktop Protocol (RDP).

Key Features:

  • Can add GPU just like EBS with exiting EC2 instances.
  • Can provide right amount of GPU as per our EC2 instance requirement
  • Elastic GPUs makes them preferably suitable for any workload that needs a small amount of additional GPU like virtual desktops, gaming, industrial design, or HPC visualization.
  • An elastic GPU is not form part of the hardware of your instance. However, the elastic GPU is network-attached through a network interface, known as the elastic GPU network interface
  • AWS use OPenGL API to connect Elastic GPU network Interface that connect Elastic GPU at backend

Supported Regions:

  • US-East-1 (N.Verginia)
  • US-East-2 (Ohio)

Supported OS:

It is supported on limited Windows OS at the moment.

  • Server 2012
  • Server 2012 R2
  • Server 2016

Elastic GPU Types:

  • eg1.medium (1GB)
  • eg1.large (2GB)
  • eg1.xlarge (4GB)
  • eg1.2xlarge (8GB)

EC2 Instances which are not supported:

Elastic GPU supported all instance types & Families except below. In simple words Elastic GPU does is not covered in AWS Free Tier services.

  • T2 Micro
  • T2 Nano
  • T2 Small

Other GPU Instance types:

  • GPU Graphics - (g3.4xlarge, g3.8xlarge, g3.16xlarge)
  • GPU Instances - (g2.2xlarge, g2.8xlarge)
  • GPU Compute - (p2.xlarge, p2.8xlarge, p2.16xlarge)

Royal Cyber & AWS

Royal Cyber is an Amazon Partner, continuing to build greater experiences for all our customers globally. To know more about the AWS & Royal Cyber’s expertise logon to or email us at [email protected]

Leave a Reply