The AMD booth at the Computex Taipei 2014 expo in Taiwan.
Photograph by Chris Stowers — Bloomberg/Getty Images

It's late...but probably not too late.

By Aaron Pressman
December 12, 2016

Advanced Micro Devices unveiled a new line of computer graphics cards on Monday, aimed at catching up to competitors in the nascent market for artificial intelligence and machine learning tasks.

AMD’s Radeon Instinct line won’t be pitched to video gamers who run Assassin’s Creed on an ultra high-definition screen. Rather, the target market is companies, research labs, and cloud computing platforms that want to analyze big data sets and train computers to perform tasks like image recognition and financial fraud detection.

Nvidia has dominated the machine learning market so far, followed by Intel. AMD’s new cards will try to match the performance of competitors on such specialized tasks. The company will also need to match Intel intc and Nvidia nvda on software compatibility, which can be just as important for data analysis tasks as raw performance.

Get Data Sheet, Fortune’s technology newsletter.

Although AMD amd is late to bring out its own specialized machine learning cards, the market is just getting started according to Mizuho Securities analyst Vijay Rakesh. Less than one in 1,000 servers are currently dedicated to artificial intelligence tasks like machine learning, but the market will grow 12 times larger over the next few years, Rakesh predicted last month. Intel announced its own forthcoming line of specialized chips at a conference last month as well.

AMD CEO Lisa Su promised the new cards would deliver outstanding performance. “We are the only company with the GPU and x86 silicon expertise to address the broad needs of the datacenter and help advance the proliferation of machine intelligence,” Su said in a statement.

The company is developing open-source software tools as well as seeking to integrate with existing software platforms for deep learning. One AMD software tool set, known as ROCm, is already compatible with several popular machine learning software frameworks and will work with Google’s goog Tensorflow platform starting in January.

SPONSORED FINANCIAL CONTENT

You May Like