Elastic GPUs are GPU resources that can be attached to an Amazon EC2 instance to accelerate the graphics performance of applications. Elastic GPUs come in several different memory capacities and offer the flexibility to choose the right compute, memory, and storage balance for your application. They’re best suited for applications that require a small or intermittent amount of additional GPU power.
Amazon says this new offering can save significant amounts of money. Under previous arrangements the least expensive graphics workstation cost about US$0.76 per hour. That now falls to as little as $US0.12 per hour. Elastic GPUs support up to and including the OpenGL 3.3 API standards with expanded API support coming soon.
For applications requiring accelerated graphics this looks like a worthwhile tool that can save money. If reflects AWS’ continuing shift to make services more granular and accessible on demand.
The blog post announcing the introduction of elastic GPUs includes a worked example of how the new service can be used. It is currently provisioned through the us-east-1 and us-east-2 AWS regions.