New processors provide neural processing for human-like cognition at the edge

The i.MX 8M Plus applications processors in the EdgeVerse portfolio integrate a Neural Processing Unit (NPU) that unleashes the benefits of edge-based Machine Learning (ML) and Artificial Intelligence (AI) in IoT and industrial devices.

The i.MX 8M Plus features an NPU capable of delivering 2.3 TOPS, a powerful quad-core Arm Cortex-A53 processor running up to 2 GHz and an Image Signal Processor (ISP) for effective vision systems, all of which enables state-of-the-art image processing and machine learning inferencing to empower the development of smart devices with the ability to become ‘aware’ of their surroundings and make decisions on the input received with little or no human intervention. The NPU ML processing capability not only provides the ability to recognize people with accuracy comparable to human vision, it also goes beyond by predicting behavior and making decisions with the same intelligence.

The NXP i.MX 8M Plus is designed to execute multiple, highly-complex neural networks simultaneously, such as providing multi-object surveillance, recognizing 40,000 English words, detecting human emotions and moods, and numerous other machine learning capabilities. The new System-on-a-Chip (SoC) features an integrated ISP with two high-definition camera inputs to deliver real-time stereo vision and includes HDR and fisheye lens correction features.

Additionally, its intelligent vision system supports up to 12-MPixel resolution, enabling real-time algorithmic image processing for various applications (including surveillance, smart home AI server, home health monitor, smart retail and robot vision).

The i.MX 8M Plus applications processor combines advanced machine learning with exceptional vision capabilities, enabling industrial machines to inspect, measure, and precisely identity objects to improve automation in factories. Integrating Time-Sensitive Networking and an Arm Cortex M7 core for real-time processing, the i.MX 8M Plus provides deterministic processing and networking for Industry 4.0 control applications.

By including a powerful machine learning accelerator, the i.MX 8M Plus enables advanced IoT applications that continuously monitor and analyze factory data to provide advanced warning of machine failures, prevent downtime, and increase quality and yield.

To provide the high reliability required for industrial applications the i.MX 8M Plus is designed with Error Correction Code (ECC) in internal memories and the external DDR controller, and leverages the 14nm LPC FinFET process for its excellent combination of high performance and low power consumption. The i.MX 8M Plus is ideal for a number of industrial applications, such as industrial gateways, autonomous robots and anomaly detection.

Technical Specifications
• Quad-Core Arm Cortex-A53 up to 2.0GHz and Cortex-M7 up to 800 MHz
• 32/16-bit DDR3L, DDR4, LPDDR4 up to 4.0GT/sec
• Neural Processing Unit (NPU) up to 2.3 TOPS
• Image Signal Processor (ISP): Resolution up to 12 MP, 2x MIPI-CSI camera interfaces
• Video Encode/Decode including H.265/4
• 3D/2D Graphics supporting OpenGL ES 3.1, Vulkan, Open CL 1.2 FP and OpenVG 1.1
• Multiple audio interfaces: TDM, DSD512, S/PDIF, PDM mic input, eARC and ASRC.
• Low-power voice accelerator: Tensilica HiFi4 DSP @ 800 MHz
• Display: MIPI-DSI, HDMI 2.0b Tx, LVDS (4/8-lane) Tx
• 2x Gigabit Ethernet with AVB, IEEE 1588, EEE and 1x w/ TSN and 2x CAN-FD
• 2x USB 3.0/2.0 OTG with PHY type C and PCIe Gen3
• Error Correction Code (ECC) on internal memories and DDR bus
• advanced 14nm LPC FinFET process

NXP is expected to begin sampling i.MX 8M Plus application processors to customers in 1Q 2020.

NXP Semiconductors, 3501 Ed Bluestein Blvd, Austin, TX 78721, www.nxp.com