AI computing module SM7 32-EP4-22

SOPHON SM7 32-EP4-22 is an AI computing module with super performance. It is positioning the edge computing scenes with high performance requirements and has AI analysis capabilities of over 32 channels FHD video .

Independent research and development, powerful computing power

SOPHON SM7 32-EP4-22 is equipped with the fourth-generation TPU chip BM1684X, which is independently developed by Sophon. The INT8 computing power is up to 32TOPS, and it can process more than 32 channels of HD video at the same time. It is only the size of a credit card, with rich IO interfaces, and easy to integrate into various edge or embedded devices. The tool chain is complete and easy to use, and the cost of algorithm migration is small.

Chip BM1684X


32 channels HD video hardware decoding

Ultra-high computing power, ultra-strong decoding

32 TOPS INT8 computing power, 8 core A53 flexible application development;32-channel 1080P video decoding, 12-channel 1080P video encoding

Wide temperature and low power consumption, flexible deployment

-40℃~+70℃ wide temperature, flexible to cope with different environments;2/3 credit card size, typical power consumption is less than 18W

Dual mode drive, strong expansibility

PCIE Mode can be used as a slave device and SOC Mode as a master device;Rich interface, convenient function expansion

The toolchain is complete

Support AI industry mainstream frameworks Caffe, Tensorflow, Pytorch,Paddle, Mxnet

Wide application and rich scenes

It is applied to intelligent public security, intelligent park, intelligent retail, intelligent power, industrial robot UAV and other visual computing AI scenes.

Easy-to-use, Convenient and Efficient

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



Product model number

SM7 32-EP4-22

Main control CPU

8 cores ARM A53@2.3GHz

AI Computing power







Video/lmageEncoding and decoding

Video decoding capability

H.264 & H.265: 32-ch 1080P@25fps The maximum resolution supported is 7880 x 4320

Video coding capability

H.264 & H.265: 12-ch 1080P@25fps The maximum resolution supported is 7680 x 4320

lmage decodingcapability (JPEG)

400 images's @1080P. The maximum resolution supportedis 32768 x 32768


Standard configuration



Standard configuration




PCIE 3.0 x 4 EP + x 4 RC

Network interface

10/100/1000Mbps RGMII x 2

Bus interface

GPIO / SDIO / PWM / 12C etc


Interface specification

144-PIN BTB Connector

Operating temperature

Temperature range

-40°C ~ +70°C

Power consumption

Typical value


Dimensions of structure

Length * width * height

62 x 58 x 33.45 (mm)