MediaTek announces i700 SoC for A.I. IoT applications

Thanks to its dual NPU architecture, the i700 SoC is able to provide advanced A.I.-enabled face recognition and AR features for IoT devices. (Source: MediaTek)

Thanks to its dual-NPU design, the i700 SoC provides 5X increased A.I. acceleration when compared to the already available i500 chip. It also features eight ARM Cortex cores, a PowerVR GM 9446 GPU and supports up to 8 GB of LPDDR4x-1866 RAM. The chip is targeted at smart IoT devices, fitness and entertainment devices that enable AR experiences and warehouse autonomous vehicles.

Bogdan Solca, 07/11/2019

Working For Notebookcheck

Are you a techie who knows how to write? Then join our Team! Especially English native speakers welcome!

Currently wanted:
German-English-Translator - Details here

MediaTek is mostly known for its competitive Helio P-series smartphone SoCs, but the company is also looking to expand into the A.I. NPU market with chips like the newly announced i700, which is targeted at IoT applications including smart displays, smart building security systems, facial recognition for payment authentication, fitness and entertainment devices that benefit from augmented reality applications, and even warehouse autonomous vehicles.

The i700 integrates two ARM Cortex-A75 cores clocked at 2.2 GHz and six Cortex-A55 cores clocked at 2.0 GHz, plus two additional NPU cores for A.I. acceleration and face recognition that are up to five times faster than the NPU integrated in the i500 SoC. Maximum supported amount of RAM is 8 GB LPDDR4x-1866. MediaTek offers support for its proprietary NeuroPilot SDK as well as for Google’s Android neural Network API, TensorFlow, TF Lite, Caffe and Caffe 2. Connectivity specs include dual WiFi 5, Bluetooth 5.0 and 4G LTE Cat. 12 support.

As far as imagining support is concerned, the i700 can be paired with a single 32 MP camera or a dual-cam setup with 24 + 16 MP resolutions, offering 30 fps video capture. Additionally, there is a 120 fps mode dedicated to applications that require accurate identification of fast-moving objects. Graphics acceleration is handled through the IMG PowerVR GM 9446 GPU clocked at 970 MHz.

Expected release date is set for early 2020.