Core-1808-JD4 AI Core Board
With on-board RK1808 AI chip which inbuilt high-efficiency NPU, it possesses a strong calculation
capacity and supports multiple AI framework and AI application SDK, being easy to develop. With rich
interfaces for AI applications, it is convenient for extension and can be applied to voice wake-up,
speech recognition, face recognition and other application scenarios.
High-efficiency AI Chip
The CPU of RK1808 AI chip adopts the dual-core Cortex-A35 framework with the maximum frequency of 1.6
GHz. Its VPU supports the encoding and decoding of 1080P regular video format as well as the camera
video signal input with the inbuilt ISP. It provides various options for storage configuration, thus the
users only need to extend the function baseboard so as to realize project research and output quickly.
Strong AI Calculation Capacity
The inbuilt NPU has a calculation capacity up to 3.0 TOPs, supports INT8/INT16/FP16 mixed operation, and
coordinates performance, power consumption and operational precision to the greatest extent. It also
supports the network model conversion of frameworks like TensorFlow, Caffe, ONNX and Darknet.
Calculation-accelerated Unit
Coupled with RK3399, it can form the high-performance AI neural network acceleration cluster with NPU
being 1+1 or 1+N.
AI Application Development SDK
AI Application Development SDK supports C/C++, Python and fast-developed Rock-X SDK. It also supports
the running of RKNN-Toolkit on PC and RK3399, which is convenient for customers to establish the neural
network operation platform rapidly. Therefore, it’s highly convenient for software development.
Open SDK
Equipped source code, tutorial, technical materials, and development tools are provided, making
development easier and more convenient.
Stable And Reliable
Adopts SODIMM 260P interface which brings the data transmission and extension performance into full
play. Through the immersive craft, it is corrosion-resistant, solid and reliable.
Rich Extension Interfaces
With high-speed device interfaces like 1×USB2.0 and 1×USB3.0 (TX and RX used along with PCIE), and a
series of input and output interfaces like PWM, I2C, SPI and UART. It supports MIPI/CIF/BT1120 video
input, MIPI/RGB display output, Gigabit Ethernet and external WiFi/BT modules.
Application
It has wide applications in voice wake-up, speech recognition, face detection, attributive analysis,
face recognition, posture analysis, target detection and recognition, image processing, etc.
Specification
|
Specification |
| SoC |
Rockchip RK1808
|
| CPU |
Dual-Core Cortex-A35，up to 1.6GHz
|
| NPU |
Up to 3.0TOPs，support OpenCL/OpenVX，support INT8/INT16/FP16
support TensorFlow、Caffe、ONNX、Darknet
|
| GPU |
Support 2MP， AE/AWB/AF
|
| VPU |
1080p @ 60FPS H.264 decoding
1080p @ 30FPS H.264 encoding
|
| RAM |
1GB DDR3 (2GB Optional） |
| Storage |
8GB high-speed eMMC 4.51 (16GB/32GB/128GB Optional)
Support TF card expansion, support PCIe 2.1
|
|
Hardware Features |
| Ethernet |
10/100/1000Mbps Ethernet
|
| WiFi |
Extending WiFi & Bluetooth via SDIO3.0
|
| Display |
1 x MIPI-DSI, support single channel 1080P@60fps output
1 x RGB, 18-bit (RGB666) up to 1280x800@60fps
|
| Audio |
1 x I2S for audio input and output
1 x Headphone
1 x Speaker (8Ω, 1.3W)
2 x MIC
|
| Canmera |
1x MIPI-CSI (built-in dual hardware ISP, support up to single 13Mpixel or dual
8Mpixel)
1xDVP
|
| USB |
1 x USB3.0 (multiplexed with PCEI) 、1 x USB2.0 |
| Other |
Support USB3.0 / PCIe2.1
Built-in 1-ch & 8-ch I2S & 8-ch PDM，Built-in VAD
7 x UART、3 x SPI、5 x I2C、10 x PWM、4 x SARADC
|
|
OS/Software |
| OS |
Linux+QT
|
|
Appearance |
| Size |
69.6mm × 50 mm
|
| Interface |
Gold finger (SODIMM 260P , 0.5mm pitch)
|
| PCB |
8-layer board design
|
