While AI used to be a buzzword sprinkled around casually to describe new levels of machine or device automation, there’s little question now that true AI-based machine learning and intelligence is making a meaningful impact in more and more consumer, commercial and industrial devices. However, the AI edge (or end point devices) has a unique set of requirements versus AI in the data center or cloud. At the edge, robust, low latency connectivity (5G) is critical, as well as low power consumption and the ability to offer capable AI processing on device, with as little reliance on cloud data centers as possible, in order to hit latency and performance targets.
One major player that has been driving the advancement of the connected, AI-powered intelligent edge is Qualcomm. The company has been innovating in this space for years and several generations of products now, such that it has delivered over 1.8 billion AI-enabled chips to date, when you consider the number of products it has shipped with an AI engine on board.
Measuring AI Performance – Enter MLPerf
Having recently submitted MLPerf results for its Snapdragon 8+ Gen 1 Mobile Platform SoC — demonstrating a notable performance lead in AI workloads like natural language processing, image segmentation and object detection — Qualcomm continues to execute in the space enabling more capable, intelligent mobile and edge devices, from smartphones to the IoT, automotive and industrial automation.
MLPerf is a widely-respected machine learning benchmark put forth from MLCommons, which is consortium of founding members that established a set of industry-standard metrics to measure machine learning performance back in 2018. Since then, MLCommons and MLPerf has been adopted and contributed to by virtually all the major heavyweights, from Intel, AMD, NVIDIA and Arm, to Facebook, Google, Mediatek and many others, like Qualcomm.
As you can see above, with the exception of offline image classification, even Qualcomm’s previous-gen Snapdragon 8 Gen 1 platform led the field with respect to the various smartphone AI workloads, and its Snapdragon 8+ Gen 1 platform is currently unmatched across the board .
Where Qualcomm AI Lives
Beyond the benchmarks, Snapdragon AI engines power a multitude of devices and platforms, that scale from less than 1 TOP of compute for low power functionality, like noise canceling wireless earbuds, to more powerful devices like AR Glasses, where AI helps with hand and eye tracking to 6 DoF (6 Degrees of Freedom) prediction and spatial awareness.
Scaling up in horsepower, Qualcomm’s Hexagon AI engine can also be found in Snapdragon 8cx Gen 3 powered laptops like Lenovo’s ThinkPad X13s, where it’s harnessed for video conferencing with background blur, AI beautification, as well as noise and echo cancellation for audio feeds.
Robotics and industrial automation and monitoring are also big market opportunities for Qualcomm AI, where machine vision and real-time safety monitoring are keys to making factories run smoothly and maintaining safe work environments for employees. However, perhaps one of the biggest opportunities that Qualcomm Snapdragon AI is strapping into now is the connected, software-defined intelligent car.
Here, Snapdragon Digital Chassis solutions with on board AI engines power everything from infotainment systems to climate control, driver monitoring for safety, ADAS (Advanced Driver Assistance Systems) for self-driving functionality, lane detection, navigation and more.
You might say that Qualcomm’s AI engines are one of the best kept secrets in the tech industry, because they’re on-board so many of the company’s chips and provide sort of an unsung hero capability that allows for more machine intelligence, adaptation and control .
Like many others have noted, AI and machine learning are permeating everything in electronics, and in fact it’s even now helping to design new chips. With Qualcomm’s long legacy of AI development, performance-per-watt low power leadership, and a rich set of developer tools and frameworks like Qualcomm AI Stack, similarly Qualcomm AI engines and silicon solutions are powering more and more devices out on the connected, intelligent edge.