The latest iteration of Samsung Electronics Co. Ltd.’s mobile processor is designed with artificial intelligence in mind.

The Exynos 9820, which debuted today, is based on an eight-nanometer, eight-core architecture and features a specialized “neural processing unit.” Samsung said that the module enables the chip to run AI software about seven times faster than the Exynos 9810 inside its flagship Galaxy S9 smartphone.  

Mobile apps that run AI algorithms directly on the user’s device are currently something of a rarity. Because of handsets’ limited processing power, developers usually opt to offload machine learning tasks to the cloud. The spread of AI-optimized chips such as the Exynos 9820 should over time make local execution a more feasible option, potentially opening the door to new use cases that require low latency or offline access.

Samsung isn’t the first major handset maker to address the rise of machine learning. Apple Inc.’s latest iPhones are powered by a chip called A12 Bionic that likewise has a dedicated AI module, as did the A11 the company introduced last year.

The neural processing unit in the Exynos 9820 is joined by several other major improvements. To start, Samsung has added an enhanced central processing unit that it said provides 20 percent more single-core performance and 15 percent more multicore performance than the Exynos 9810. Alternatively, the chip can trade off this speed to reduce power draw by up to 40 percent compared with its predecessor.

Samsung has also revamped the onboard graphics processing unit. It uses Arm Holdings plc’s latest Mali-G76 GPU design, which the handset maker said can provide either a 40 percent performance boost or a 35 percent reduction in power consumption.

Samsung expects to start mass-producing the Exynos 9820 by the end of the year. That means that the chip could very well make its way into the company’s next Galaxy S phone and the recently unveiled Galaxy F foldable tablet, which are rumored to be on track for an early 2019 launch.  

Photo: Samsung