BeiQi RK3399Pro AIoT 96Boards Compute SoM

TB-96AI is a powerful core board for artificial intelligence. Carrier Board developed can form a complete development board or evaluation board; and the board customized by customers according to actual needs can directly form the industry application motherboard, which can meet industrial automation, UAV, image detection, face recognition, edge computing gateway, cluster server, Intelligent Quotient display, automatic driving, medicine. Application needs of market segments such as health care equipment, robots and intelligent retail.
TB-96AI uses RK3399Pro as the main control chip and TB-96AIoT uses RK1808 as the main control chip.

Features
Six-core 64-bit processor, superior general-purpose computing power
Dual-core ARM Cortex-A72 MPcore processor and quad-core ARM Cortex-A53 MPcore processor are high-performance, low-power and cache application processors.
Two CPU clusters. Big cluster with dual-coreCortex-A72 is optimized for high-performance and little cluster with quad-core Cortex-A53 is optimized for low power
Full implementation of the ARM architecture v8-A instruction set, ARM Neon Advanced SIMD (single instruction, multiple data) support for accelerating media and signal processing
Supporting 8 bit/16 bit operation, AI computing power up to 3.0 TOPs (INT8 Inference)
Full load calculation is strong and light load operation power consumption is low.
Compatible with Caffe/Mxnet/TensorFlow model, it can support multiple frameworks, support mainstream layer types, and add custom layer easily
Provide easy-to-use development tools, PC can complete model conversion, performance prediction, accuracy verification.
Provide AI application development interface: support Android NN API, RKNN cross-platform API, Linux support TensorFlow development;
Powerful Multimedia Processing Performance
Integrated quad-core ARM Mali-T860MP4 GPU, support OpenGL ES1.1/2.0/3.0, OpenCL1.2, Directx11.1, etc., with more bandwidth compression technology
Strong hardware codec capability
Support 4K VP9 and 4K 10bits H265/H264 video decoding up to 60fps
Support 1080P multi-format video decoding (VC-1, MPEG-1/2/4, VP8)
Support 1080P video encoding, support H.264, VP8 format
Multiple video input and output interfaces
Dual camera interface: two MIPI-CSI input interfaces with two ISP image processors
Display output interface: Embed two VOPs, support dual-screen simultaneous/dual-screen display, and can choose to output from the following display interface.
Rich expansion interface
Type-C/DP¡Á1,OTG
USB2.0¡Á2,HOST
USB3.0¡Á1,According to the RK3399Pro design, the NPU needs to be mounted on the USB3.0, so the USB3.0 needs to be connected back to the NPU. If you need to extend the USB3.0 interface, you need to plug in the HUB.
SDMMC¡Á1
SPI¡Á1
CPU Debug UART¡Á1,NPU Debug UART¡Á1
UART¡Á1
I2S¡Á1
SDIO¡Á1
I2C¡Á1
PCIe¡Á1
PWM¡Á2
GPIO,For detailed GPIO definition, please refer to interface definition.
ADC¡Á3,One for buttons, one for headset microphone detection, and one for user-definable use
High-speed on-board connector for more stability and reliability
4 Panasonic high-speed onboard connectors for higher speed signal stability
The core board can be fixed by 4 screw posts for various working environments.
Support for multiple operating systems
Support Android, Linux, Ubuntu
Support U disk upgrade through USB interface
Hardware related information
Circuit schematic reference design
Connector PCB package
Core board size
Pin definition, interface package
Software related information
Software development guide.pdf
Tools. RK driver assistant, firmware upgrade tool, etc
Firmware. Android firmware, Linux firmware
Source code. Android SDK source code
Specifications
Basic Parameters
SoC Rockchip RK3399Pro
GPU ARM Mali-T860 MP4 Quad-core GPU
Support OpenGL ES1.1/2.0/3.0/3.1, OpenVG1.1, OpenCL, DX11
Support AFBC (frame buffer compression)
CPU Dual-core Cortex-A72 up to 1.8GHz
Quad-core Cortex-A53 up to 1.4GHz
NPU
Support 8bit/16bit operation, computing power up to 3.0TOPS
Full load computing power, low load operation power consumption is low
Compatible with Caffe/Mxnet/TensorFlow model, support multiclass framework, support mainstream layer type, easy to add a custom layer
Provides easy-to-use development tools, PC-based model conversion, performance estimation, and accuracy verification
Provide AI application development interface: support Android NN API, provide RKNN cross-platform API, Linux support TensorFlow development;
VPU
Support 4K VP9 and 4K 10bits H265/H264 video decoding, up to 60fps
1080P multi-format video decoding (WMV, MPEG-1/2/4, VP8)
1080P video encoding, support H.264, VP8 format
Video post processor: de-interlacing, denoising, edge/detail/color optimization
RAM
3GB LPDRR3(CPU 2GB + NPU 1GB)
Flash
16GB eMMC
Hardware Characteristics
Ethernet Built-in Gigabit Ethernet PHY chip, 10/100/1000Mbps adaptive
Camera Interface
MIPI-CSI¡Á2,Dual camera interface (built-in dual hardware ISP, up to single 13Mpixel or dual 8Mpixel)
Display Interface MIPI-DSI
eDP
DP
HDMI (Support 480p/480i/576p/576i/720p/1080p/1080i/4k, support RGB format)
Audio Port
I2S0£ºSupport user extended use
I2S1£ºSpeaker¡Á1, Headphone¡Á1, MIC¡Á1
I2S2£ºHDMI interface audio output; DP interface audio output;
Type-C USB3.0/DisplayPort 1.2,OTG
USB
USB3.0¡Á1 (according to RK3399Pro design, NPU needs to be mounted on USB3.0, so USB3.0 needs to connect back to NPU, if you need to expand USB3.0 interface, you need external HUB);
USB2.0¡Á2, HOST
Extension Port
SDMMC(TF Card)¡Á1;
SPI¡Á1;
UART¡Á3,One of the CPU Debug UARTs, one NPU Debug UART;
I2C¡Á6;
SDIO¡Á1;
PCIe¡Á1;
PWM¡Á2;
GPIO,For detailed GPIO definitions, please refer to the interface definition;
ADC¡Á3,One for buttons, one for headset microphone detection, and one for user-definable use;
Power input DC 5V