AI Accelerator Card SC7 HP75

SC7 HP75 is an AI inference accelerator card designed for efficient adaptation to mainstream AI algorithms on the market. It enables applications such as video structuring, facial recognition, behavioral analysis, and status monitoring, providing empowerment for various fields including smart cities, intelligent transportation, smart energy, smart finance, smart telecommunications, and smart industry.

96 TOPS peak performance INT8

48 TFLOPS peak performance FP16/BF16 computing power

Supports Mixed Precision Calculations

96 Channels 25fps 1080P Video Hardware Decoding

36 Channels 25fps 1080P Video Hardware Encoding

Up to 8K Resolution of Video and Image Decoding

Compatible with various servers

Compatible with various operating systems and mainstream algorithm frameworks

Wide Range of Applications and Scenarios

The SC7 HP75-I acceleration card boasts powerful performance and can be paired with various servers to form intelligent analysis clusters through the stacking of multiple cards. It can be utilized for large-scale, multi-channel AI computing and analysis in a variety of scenarios, such as audio and video auditing, video structuring, and industrial visual inspection and analysis.

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.

Support mainstream programming framework




AI Accelerator Card SC7 HP75


24-core ARM A53@2.3GHz

AI Computing Power

96 TOPS peak performance
FP16 BF16
48 TFLOPS peak performance
6 TFLOPS peak performance

Video/JPEG Codec

Video Decoding
H.264 & H.265: 1080P @2400fps
7680*4320 /8K / 4K / 1080P / 720P / D1 / CIF
Video Encoding
H.264 & H.265: 1080P @900fps
4K / 1080P / 720P / D1 / CIF
Image Decoding
JPEG: 1200张/秒 @1080P (interleave off)
Max. Resolution
32768 * 32768


48GB LPDDR4x 384bit 205GB/s


Physical/Power Interface
PCIe Gen3 X16
Data Link
PCIe Gen3 X8

Power Consumption



0℃ ~ 55℃
Heat Dissipation


Half Height, Half length, Single slot

Deep Learning Framework

TensorFlow / PyTorch / Paddle / Caffe / ONNX / MXNet / DarkNet

Operation System Support

Ubuntu / CentOs / Debian / UOS / Kylino

Main CPU Support

Compatible with Intel / AMD / and other CPUs