MXM AI inference acceleration card SV100

deep learning

⚫ uses an SV200 that provides up to 100TOPS of computing power. ⚫ Supports overcurrent and overvoltage protection. ⚫ Intelligent fan speed control is supported. ⚫ The device status information, such as temperature, can be obtained in-band. ⚫ The comprehensive command line management function enables users to manage devices by running various commands. ⚫ Supports in-band asset management and provides accelerator card information to facilitate asset management. H1 accelerator card is a circuit board for reasoning applications in DNN and SNN fields developed for SV100 brain computing chips, which is used to demonstrate the powerful computing power of SV100. The H1 is connected to the MXM slot of the computer server via a goldfinger to form a basic development system. SV100 is a high-performance, low-power AI chip designed specifically for image recognition, video processing, inference computing and machine learning, with integrated memory and computing and multi-core parallel. Built-in high-speed image preprocessing unit provides efficient and easy-to-use pre-and-post processing module for video and image application development, improves the overall application capability, and supports a variety of deep learning frameworks, including Tensorflow, Caffe, Pytorch, MxNet and ONNX.