Sophon Smart Server SA3

Based on the self-developed artificial intelligence SOC chip and the intelligent server customized for the intelligent analysis field, it provides a wider range of video intelligent applications for users in security, video and other industries, with the characteristics of high performance and low power consumption.

Video customization, energy saving and efficient

Sophon SA3 intelligent server can be equipped with a third-party algorithm to run video structured, face detection and recognition, vehicle detection and recognition and other types of security, video intelligent analysis applications. SA3 intelligent server has the characteristics of high performance, high computing density, low power consumption, and is equipped with easy-to-use system software and tool chain to help customers to carry out rapid and convenient secondary development.

Video structure

Face detection and recognition

Vehicle detection and identification

high performance

A single device supports up to about 90 channels of video structure and about 600 face images per second

High energy efficiency ratio

Based on AI special chip with high energy efficiency ratio, low power consumption main control processor, reduce the power consumption of the whole machine

High computing density

High-density integration, close deployment, single built-in 18 BM1682, reduce TCO

High scalability

Multiple servers can be deployed in clusters, Ethernet connections between servers, unified management services, and cloud deployment

Wide application and rich scenes

Aggregate multiple intelligent algorithms to perfectly support edge computing such as face recognition, vehicle recognition, object recognition, video structure, behavior analysis, etc

Easy to use and convenient, full stack efficient

BMNNSDK (BITMAIN Neural Network SDK) one-stop toolkit provides a series of software tools such as the underlying driver environment, compiler, inference deployment tool and so on. Easy to use and convenient, covering the model optimization, efficient runtime support and other capabilities required for the neural network inference stage, providing easy-to-use and efficient full-stack solutions for deep learning application development and deployment. BMNNSDK minimizes the development cycle and cost of algorithms and software. Users can quickly deploy deep learning algorithms on various AI hardware products of Fortune Group to facilitate intelligent applications.

Support mainstream programming framework




Video structuring ability

About 30 channels 1080P

About 90 channels 1080P

Face image analysis ability

About 300 images per second

About 800 images per second

Floating point performance

18T Flops

54T Flops

Master processor

Quad-core processor

Quad-core processor

Number of intelligent units



Number of BM1682



AI computing memory




Redhat / CentOS / Ubuntu

Redhat / CentOS / Ubuntu

Maximum power consumption



System storage



Network Interface




1*USB 2.0, 1*RS232, 1*LCD

1*USB 2.0, 1*RS232, 1*LCD

Power supply

AC 220V, 1+1 Redundant power supply, HotPlugIn

AC 220V, 1+1 Redundant power supply, HotPlugIn

Cooling Method

Air cooling, adjustable fan speed

Air cooling, adjustable fan speed

Chassis Specifications

19-inch 1U standard case

19-inch 2U standard case

Size of server case (Depth*Width*Height/mm)

436.4*610*43.6 (Without handle)

436.4*610*86.1 (Without handle)