While it isn’t clear yet how Samsung specifically plans to apply such technology to the Galaxy S10, new reports suggest that the company is working on a neural processing unit (NPU) for its next Exynos chipset. This is a development that reportedly could lead to the introduction of new artificial intelligence features on the upcoming flagship phone.
In a report published Saturday, Phone Arena cited the LinkedIn profile of a former Samsung senior hardware design engineer, which stated that the employee worked with Samsung from June 2016 to March 2018 and was working on a second-generation NPU from December 2017 up until the time they left the company. The profile also shows that the engineer was responsible for “architectural changes” on the NPU for “performance and power optimization” purposes, and had designed parts of the unit’s microarchitecture.
According to Phone Arena, it’s possible that Samsung’s second-generation NPU design was unfinished at the time the engineer left. However, since the project started close to one year ago, the publication also speculated that there’s a chance the work will be complete right in time for the unveiling of next year’s Galaxy S10.
Phone Arena also explained the reason why the ex-Samsung engineer referred to the company’s NPU as a “second-generation version,” noting that the South Korean tech giant might have developed an early version before work on the new one started. This, according to Phone Arena, suggests that Samsung wasn’t happy with the original design, but also hints that the company might be under greater pressure to integrate the new NPU into its upcoming Exynos processor.
NPUs first became a key part of many smartphone processors last year, when Huawei introduced the Kirin 970 and hyped it up as the first chipset to come with such feature. Similarly, Apple introduced the A11 Bionic chip with last year’s iPhone 8, 8 Plus, and X, noting that the processor features a neural engine “purpose-built for machine learning.” As explained by Engadget, the main benefit of last year’s NPUs and neural engines was their ability to make machine learning computations normally processed in the cloud more efficient, allowing for smoother multitasking on a mobile device.
Although Samsung has typically released its Galaxy flagships with Exynos processors for most international markets and Qualcomm processors for the U.S. and some other parts of the world, UberGizmo wrote that it’s also been rumored that Qualcomm’s next-generation processor will also come with an NPU, thus making it possible that the Galaxy S10 will support AI processing, regardless of the chip found under the hood.