Google LLC is equipping Chrome with an implementation of WebGPU, a new technology that allows browsers to render graphics and run machine learning models faster.

The company announced the update on Thursday. Google’s WebGPU implementation is available with Chrome 113, an upcoming release of the browser currently undergoing beta testing. The WebGPU integration will initially roll out to Windows, macOS and ChromeOS with support for more platforms set to arrive later.

Tasks such as rendering a webpage’s visual assets can be done faster by a computer’s graphics card than its central processing unit. However, using graphics cards for processing has historically been a challenge for browsers. Until the start of the previous decade, the task often required downloading plugins.

In 2011, a consortium backed by Google, Apple Inc. and other industry players released a technology called WebGL. It gave browsers the ability to render web pages using a computer’s graphics processing unit without having to rely on plugins as before. WebGPU, the new technology that Google has implemented in the latest version of Chrome, is the successor to WebGL.

A web application running in Chrome can send a computing task to WebGL, which will then forward it to the GPU of the user’s machine. WebGL sends instructions to the GPU via specialized application programming interfaces. In Windows, the API for accessing graphics cards is called Direct3D, while Apple’s macOS API is known as Metal.

WebGL, the previous-generation software that WebGPU replaces, worked in a similar manner. But WebGL doesn’t support many of the newest features offered by graphics card APIs such as Direct3D and Apple’s Metal. WebGPU does, which allows it to render graphics faster. 

“WebGPU is a new web graphics API that offers significant benefits such as greatly reduced JavaScript workload for the same graphics,” Google engineers François Beaufort and Corentin Wallez detailed in a blog post. “This is possible due to more flexible GPU programming and access to advanced capabilities that WebGL does not provide.”

The second tentpole feature of WebGPU is its extensive support for GPGPU, or general-purpose GPU, computations. Those are tasks that are carried out with a graphics card but don’t involve rendering. One such task is running machine learning models. 

When WebGL was released, consumer GPUs were rarely used to run artificial intelligence applications. As a result, the technology is not well-optimized to perform GPGPU computations. WebGPU, in contrast, was designed specifically to power GPGPU use cases such as browser-based machine learning. 

According to Google, the technology can significantly speed up web applications that include AI features. The WebGPU-powered version of Chrome performs machine learning inference more than three times faster than was possible with WebGL.

Chrome is the first major browser to implement WebGPU. According to Google, the technology will be added to Firefox and Safari as well further down the road. The search giant, for its part, plans to enhance the WebGPU implementation in Chrome by adding features that will help developers make better use of consumer GPUs’ processing capacity. 

Image: Google