AWS GPU Instances Now More Affordable Than Ever


Serverless computing and GPU instances are two areas of the cloud that are poised for rapid growth, but both fields come with their own unique set of challenges.

For serverless, enterprises are cautious to jump in due to the lack of local testing of functions, though Amazon is changing that with SAM Local for Lambda, a feature that gives iterative development a boost allowing developers to write some code, test it to make sure that it works, and then write more.

However, app streaming via GPU instances is something that has been held back because of cost — high prices have made the technology prohibitive for many a company.

All that has started to change with AppStream 2.0, where a new AWS GPU instance type has made the process more affordable.

Amazon AppStrema

This cloud offering enables enterprises to stream desktop applications from Amazon Web Services to a web browser that is compatible with HTML5, delivering graphics intensive applications for workloads like creative design, gaming and engineering that rely on certain technologies for hardware acceleration.

Technologies like DirectX, OpenGL, and OpenCL.

Amazon introduced the Graphic Design instance for AppStream 2.0, which cut the cost of streaming graphics applications up to 50%. These instances make use of the AMD FirePro S7150x2 Server GPUs with AMD Multiuser GPU technology.

And come in sizes from 2 to 16 virtual CPUs, and anywhere from 7.5 to 61 gibibytes of system memory.

Ideal for any type of graphics intensive work, and making the notion of working remotely for designers a much more affordable proposition than ever before.

The cloud giant lists the complete pricing details for these instance types here.

Developers now also have the option to select between two types of Amazon AppStream instance fleets in a streaming environment — Always-On and On-Demand.

The Always-On fleets provide instant access to apps, charging a fee for every instance in the fleet, while On-Demand fleet only charge fees for instances when end users are connected, plus an hourly fee, though users will experience a delay when they access the first application.