AMD demonstrated its XDNA AI core at Computex, joining Intel in showcasing its artificial intelligence (AI) capabilities. The XDNA AI core was featured in the Phoenix Accelerated Processing Unit (APU), which forms the basis of AMD’s Ryzen 7040 series laptop processors and the Z1 chip in the Asus RoG Ally. This is the first time the XDNA AI core has been shown to the public. AMD has not announced any plans to put the XDNA engine into its desktop CPUs. The engine is currently only available for laptops and handheld devices.
XDNA AI Core Functionality
During the demo, the XDNA engine was tested on an Asus Strix Scar 17 with a Ryzen 9 7940HS chip. The demo involved accelerating a facial recognition task, but no comparative numbers were generated. Therefore, there is no measure of Phoenix’s performance compared to running on a CPU, GPU, or a combination of the two. AMD claims that its XDNA engine is faster than the equivalent Neural engine in Apple’s M2 chip but has not made any claims compared to Intel’s AI tile in Meteor Lake, also known as a VPU or “Versatile Processing Unit”.
The XDNA AI engine is designed to accelerate light AI inferencing workloads, including audio, video, and image processing. It is intended to accomplish these tasks faster and more efficiently than a CPU or GPU. The net result should be lower latency for real-time audio processing or background blurring and better battery life while doing it.
AMD announced a new set of tools to help developers code for the XDNA engine. However, there are no current examples of software or apps that can utilize XDNA for now.
The XDNA AI engine is a new word salad that developers and tech enthusiasts will have to get used to. The usefulness of the AI cores remains to be determined, but the possibilities are endless.