» Current | 2020 | 2019 | 2018 | Subscribe

Linley Newsletter

Neural Engines Rev Up Mobile AI

January 23, 2018

Author: Mike Demler

Siri and Google Assistant popularized the use of machine learning applications on mobile devices. But they stream audio directly to the cloud, where deep neural networks (DNNs) running on powerful servers parse the spoken words. By contrast, new premium mobile processors can perform similar tasks directly thanks to their integrated neural-network engines (NNEs).

Although CPUs, GPUs, and DSPs can all run pretrained neural networks, they each make performance, power-consumption, precision, and programmability tradeoffs. Therefore, designers must employ heterogeneous-computing architectures that distribute the workload to the core best suited to each task. To further increase efficiency and enable more-complex machine learning, the trend is to augment the CPU/GPU/DSP complex with specialized deep-learning accelerators (DLAs).

Companies employing DLAs in their latest flagship processors include Apple, Huawei, and MediaTek. We expect Samsung will also integrate one in its next-generation Exynos. These integrated neural-network engines enable on-device machine-learning tasks ranging from keyword spotting for voice UIs to more complex biometric apps such as facial-, iris-, and voice-identification. Although the early focus for on-device machine learning has largely been AR, computer vision, and personal digital assistants, locking a smartphone to its owner will increase its use for a plethora of security-sensitive applications such as mobile payments, which can help thwart identity theft.

Subscribers can view the full article in the Microprocessor Report.

Subscribe to the Microprocessor Report and always get the full story!

Purchase the full article


Linley Spring Processor Conference 2020
Coming April 7-8, 2020
Hyatt Regency, Santa Clara, CA
Register Now!
Linley Fall Processor Conference 2020
Coming October 28-29, 2020
Hyatt Regency, Santa Clara, CA
More Events »


Linley Newsletter
Analysis of new developments in microprocessors and other semiconductor products
Subscribe to our Newsletter »