Deep Learning Mainboard

   AIO-1684JD4 deep learning mainboard is equipped with SOPHON's third-generation TPU processor BM1684 and eight-core Cotex-A53. Its basic frequency is up to 2.3GHz, the INT8 computing power is up to 17.6Tops, 12GB memory, and can support up to 32 channels of 1080P@25fps video decoding.  It can process more than 16 channels of HD video at the same time and support mainstream deep learning frameworks such as cafe / tensorflow / mxnet / pytorch / PaddlePaddle. It is integrated with dual 1000Mbps Ethernet ports, expandable WiFi, Bluetooth and 5G/4G LTE. It also has rich interfaces such as RS485, RS232, USB3.0, USB2.0 and HDMI. It supports Ubuntu, Debian, Linux and other operating systems; The mainboard can be directly applied to automatic driving, intelligent security, medical treatment, city, transportation, fire protection to speed up the product commercialization.


Specification and Characteristic

Product Model
Application Scenarios

Face recognition, automatic driving, urban brain, intelligent security,  medical care, city, transportation,  construction site and fire control

Deep Learning Processor
Deep Learning Performance
17.6T(INT8)/ 2.2T(FP32)/ 35.2 T(INT8,enable winograd)
Master Processor
8-core ARM A53, basic frequency up to 2.3GHz
Video Decoding Capability
1080P@960fps, support H264 & H265
Video Encoding Capability
1080p@50fps video encoding
Picture Codec Capability
MJPEG image encoding and decoding up to 1080P@480fps
16GB/32GB/64GB/128GB eMMC,1 * 2242 SATA3.0 SSD(M.2 SATA), 1 * TF Card
Network Interface
2 * 1000Mbps(RJ45)Ethernet
USB Interface
2 * USB3.0(Current Limit 1A)
Display Interface
1 * HDMI2.0(4K@60fps)
Audio Interface
1 * HDMI2.0 Audio output
Wireless Network
WiFi module supported through the M.2 E-KEY interface:
- Support 2.4GHz / 5GHz dual-band WiFi, WiFi6, 802.11a/b/g/n/ac/ax protocol
- Support Bluetooth 5.0 expandable 4G LTE (Mini PCIe interface), 5G (M.2 interface)
Other Interface
1 * Debug(3P-2.0mm), 1 * FAN(12V,4P-1.25mm),1 * RS232 1×RS485
DC12V (DC5.5×2.1mm)
Other Parameters

1 * HDMI, 1 * TF, 1 * M.2, 2 * miniPCIE, 1 * RS-485, 1 * RS-232, 1 * WiFi, 1 * 5G, 1 * 4G

Operating Temperature
Storage Temperature
Operating Humidity
-40%~70 %
Operating System
Supports Ubuntu desktop and server, Debian
Deep Learning Framework
Support TensorFlow / Caffe / PyTorch / Paddle / ONNX / MXNet / Tengine / DarkNet

Partner Introduction

Zhongshan TianQi Technology Official Website

   Zhongshan TianQi Technology (hereinafter referred to as" TianQi "or" the company ") was established in 2005. It is a national high-tech enterprise, Guangdong Internet+ enterprise and Zhongshan engineering technology center. It is an industry-leading service provider for intelligent hardware design, manufacturing and deep learning application solutions.

   TianQi  has focused on the development of intelligent hardware for more than 10 years and has a deep cooperation with a number of first-tier deep learning processor manufacturers. To meet the development needs of deep learning development. it has focused on the deep learning edge computing field, and launched deep learning edge computing modules, deep learning edge computing boxes, edge intelligent servers and IOT gateways, providing a full rage of deep learning edge computing products and solutions for the IOC, intelligent security, medical care, smart city, intelligent transportation, construction site, fire safety and other fields.