WPG World Peace Group launches OP-Killer AI prototype development board solution based on NXP products

On April 21, 2022, WPG Holdings, a leading semiconductor component distributor dedicated to the Asia-Pacific market, announced that its subsidiary WPI has launched the OP-Killer AI prototype development board solution based on the NXP i.MX8M Plus chip .

Figure 1- A display board of WPG’s OP-Killer AI prototype development board solution based on NXP products

Technology is advancing with each passing day, allowing AI applications to quietly integrate into people's lives. In the near future, edge computing will bring another peak to this field. With the birth of the Neural Processing Unit (NPU), AI computing power has grown by leaps and bounds. It also drives the rapid implementation of applications such as machine learning and artificial intelligence in various fields such as industry, medical care, and the Internet of Things, thereby providing people with smarter and more complete services. In order to help developers quickly develop AI applications, WPG launched the OP-Killer AI prototype development board solution based on the NXP i.MX8M Plus chip. The solution provides a set of AI-related application product development prototypes and technical resources based on the OP-Killer EVM Board. Even developers who are new to this field can quickly get started designing machine learning related applications.

Figure 2-Application scenario of WPG World Peace's OP-Killer AI prototype development board solution based on NXP products

This solution consists of two development boards: SOM Board and I/O Board . Among them, SOM Board is the main development board. It is based on the NXP i.MX8M Plus platform and provides two specifications of 4x Cortex-A53 processors at 1.6GHz or 1.8GHz, as well as a powerful architecture including a 12-megapixel image processor ISP, a neural network processor NPU with a computing power of 2.3TOPS, a graphics accelerator 2D/3D GPU, and a sound digital signal processor HiFi 4 DSP. Among them, the neural computing processor NPU is the core highlight of this chip. With its unique hardware architecture characteristics, it can provide extremely high computing performance in a low-power environment, thereby greatly improving the reasoning efficiency of machine learning.

The I/O Board is an expansion board with rich interfaces that can provide diverse application support, making it more flexible in actual application. With this development board, the concept of edge computing can be implemented in fields such as the Internet of Things (IoT), Industry 4.0, and autonomous cars to create better application value.

Figure 3-Block diagram of WPG's OP-Killer AI prototype development board solution based on NXP products

In addition, this solution is also equipped with NXP's machine learning development environment eIQ (Edge Intelligence), which can quickly use deep learning frameworks such as TensorFlow Lite, ONNX, ArmNN, and DeepViewRT. With this advantage, users only need to authorize the corresponding data such as video and sound to any deep learning framework for inference (Inference), and they can quickly parse the neural network architecture to get the results. And the framework will optimize the acceleration calculation through the OpenVX database and NPU. After actual testing, the MobileNet-SSD object detection inference speed can reach about 80 FPS per second. The MobileNet object classification inference speed can reach 329 FPS per second.

Core technology advantages:

• With a computing power of 2.3TOPS, the Neural Processing Unit (NPU) has powerful machine learning reasoning capabilities. It is more power-efficient and efficient than the well-known Graphics Processing Unit (GPU), and is a processor specially designed for deep learning and artificial intelligence.

• The independent SOM development board design, combined with the powerful NXP i.MX8M Plus chip, can provide a minimum boot system and can be used with USB Wi-Fi / BT modules in the future, making it more suitable for applications in IoT and industrial control fields.

• Combined with the I/O development board, it can provide a complete range of peripheral configurations, such as high-definition multimedia interface (HDMI), low-voltage differential signaling technology interface (LVDS), Ethernet, controller area network (CAN bus), asynchronous receiver and transmitter (UART), universal serial bus interface (USB Type A/C), 3.5mm headset audio interface, lens information transmission interface (MIPI-CSI), display information transmission interface (MIPI-DSI), and M.2-PCIe 3.0 transmission interface.

• You can quickly get started with the eIQ / PyeIQ machine learning development environment, which provides application examples of multiple deep learning frameworks such as TensorFlow Lite, ONNX, and DeepViewRT.

Program Specifications:

SOM Borad Specifications

• MPU (NXP i.MX8M Plus, MIMX8ML8CVNKZAB/MIMX8ML8DVNLZAB) specifications:

• Equipped with four Arm Cortex-A53 high-performance processors, with maximum clock speeds of 1.6GHz and 1.8GHZ respectively;

• Equipped with a neural processing unit (NPU) with 2.3TOPS computing power for machine learning applications;

• Equipped with two image signal processing units (ISP), it can resolve 120,000 pixels and reach 375 MPixels/s;

• Supports two sets of MIPI-CSI lens interfaces

• Support low-power sound accelerator: Cadence® Tensilica® HiFi 4 DSP at 800MHz;

• Powerful 3D/2D graphics accelerator (GPU, GC7000UL);

• Powerful video decoder and encoder support 1080p at 60 frame video streaming.

PMIC (PCA9450C) specifications:

• A set of bidirectional buck regulators is provided;

• Five sets of linear regulators are provided;

• Provide 400mA active load switch;

• Supports ESD protection mechanisms: +/- 2000V HBM and +/-500V CDM.

eMMC 5.1 (MTFC32GAPALBH-IT) specifications:

• Storage capacity is 32GB;

• The operating voltage is 2.7V to 3.6V;

• The operating temperature is -40℃ to 85℃.

External Memory LPDDR4 (MT53D1024M32D4DT-046 AAT:D) Specifications:

• The maximum operating clock is 2133MHz;

• The storage capacity is 4GB;

• The operating voltage is 1.1V;

• The operating temperature is -40℃ to 105℃.

NOR Flash (IS25WP256E-JLLE) specifications:

• The storage capacity is 32MB;

• The operating voltage is 1.7V to 1.95V;

• The operating temperature is -40℃ to 105℃.

I/O Board Specifications

• 1x PCIe M.2 Key M transmission interface;

• 1x Expansion Connector expansion interface (I2C, GPIO, UART, PWM, SPI, PDM);

• 2x LVDS low voltage differential signaling technology interface;

• 1x USB Type A 3.0 universal serial bus interface;

• 1x USB Type C 3.0 universal serial bus interface;

• 1x Debug port (Micro USB);

• 2x CAN Bus controller area network;

• 1x MIPI-DSI display data transmission interface;

• 3.5mm headset audio source interface;

• 2x Gigabit Ethernet;

• 1x HDMI high-definition multimedia interface;

• 2x MIPI-CSI lens information transmission interface.