Image Compute Engine (ICE)

Visual Computing is transforming the way responsive websites and mobile applications are designed and created.

Web and mobile applications need to look great and render quickly on any size window or display. Many sites create numerous copies of the same image, at different sizes, and choose the right image to transmit and display. These images need to be stored, replicated and made available for users anywhere in the world.

The NVIDIA Image Compute Engine (ICE) makes responsive design possible without the need to cache multiple copies of images in different sizes and formats. Any image needed for the layout of a page can be converted almost effortlessly to whatever size is needed in the blink of an eye.

This provides a variety of benefits such as:

  • More responsive web pages and web applications for diverse screen formats
  • More maintainable sites and applications
  • Lower bandwidth, storage and cloud/datacenter costs

The fundamental difference with ICE is that you no longer have to trade-off between complexity, responsiveness, and the demands of visual design. With ICE, you can have it all.

ICE is a plug-and-play GPU-accelerated web service that takes requests for images to be resized in the form of a HTTP request and returns the properly sized image. It does this by performing the image resize computations on the GPU to achieve lower latency and higher bandwidth than CPU-only solutions.


  • RESTful micro service
  • Rapid Response Time
  • High Bandwidth
  • Supports JPEG images
  • Optimize compression setting while preserving image information

"Our photographers communicate their vision through the photos they share on our platform, and experiencing those images quickly at the highest quality regardless of screen size is critical to their success. NVIDIA’s ICE web service running on the Tesla Platform delivers millions of pixel-perfect images at exactly the right size every day, so every photo looks just right." Don MacAskill, CEO & Chief Geek of SmugMug.


Consider the images you create when you snap a picture with your smart phone. Those are relatively high resolution images that ICE can resize on an NVIDIA Tesla K40 GPU in approximately a quarter of a second (260 ms). That’s 7 times faster than the same kind of round-trip request can be completed with an Intel Xeon E5 CPU. ICE is typically used to resize large numbers of images over time, sustaining a rate of 70 image conversions per second - that’s 14 times faster than a CPU-only server.

Sign up for an ICE evaluation today

You may also be interested in: