Domestic AI Edge Computing Box Based on Sophon, with Powerful Built-in TPU | 32 TOPS INT8 Computing Power
Edge Computing Box
Powerful Built-in TPU | 32 TOPS INT8 Computing Power
● TPU platform box supporting floating-point operations, with high computing power of 32 TOPS@INT8, 16 TFLOPS@FP16, and 2 TFLOPS@FP32.
● Single chip supports real-time decoding of up to 32 channels of H.264 & H.265.
● Supports the domestic AI framework PaddlePaddle, and is compatible with mainstream deep learning frameworks such as Caffe /TensorFlow/MxNet/PyTorch/ONNX.
● Supports Docker containerized deployment and Python development environment.
● Features OTA (Over-The-Air) upgrade capability.
● Low-power design, combined with passive cooling from the enclosure.
● Supports wide operating temperature range of -20℃ to 60℃.
● Compatible with the domestic Kylin operating system.
This highly integrated smart workstation offers ultra-strong computing performance with a powerful built-in TPU, delivering up to 32 TOPS of INT8 computing power. It features an 8-core ARM Cortex-A53 CPU processor clocked at up to 2.3GHz. A comprehensive development and migration toolchain is provided, enabling rapid algorithm porting and project implementation. The smart workstation can operate in wide temperature environments and can be flexibly deployed in various AI scenarios, finding extensive applications in smart factories, smart construction sites, smart urban management, smart gas stations, and other fields.
| Item | Type | Model/Parameter | Description | | :--- | :--- | :--- | :--- | | Core Configuration | Processor | CPU | 8-core ARM Cortex-A53 @ 2.3GHz | | | INT8 | 32 TOPS | | | | FP16 | 16 TFLOPS | | | | FP32 | 2 TFLOPS | | | | Memory | LPDDR4 | Default configuration 16 Gbyte | | | Flash Storage | eMMC | Default configuration 64 Gbyte | | Hardware Features | Codec Performance | Video Decoding | 32 channels 1080P @ 25fps, H.264 & H.265 | | | Video Encoding | 12 channels 1080P @ 25fps, H.264 & H.265 | | | | Image Codec | 1080P 600 frames/second | | | | Ethernet Port (Ethernet) | x2 | Supports 10/100/1000M network access | | | RS-232 | x1 | Can connect to alarm input devices such as smoke detectors, infrared detectors, access control systems, or alarm output devices such as alarm bells (Note: RS-232 cable length is recommended not to exceed 10m) | | | RS-485 | x1 | Can connect to alarm input devices such as smoke detectors, infrared detectors, access control systems, or alarm output devices such as alarm bells | | | HDMI_OUT | x1 | Can output 1080P@60fps video source to a display terminal | | | Relay | x1 | Can be used for current loads, voltage below 36V | | | RST | x1 | Standard GPIO button, can be used for application software reset | | | USB3.0 | x2 | Can connect to USB flash drives, USB mice, USB keyboards, and other devices | | | DEBUG | x1 | System debug port (Type-C) | | | TF Card Slot | x1 | Can insert TF card to expand storage space, Class 10 or higher specification is recommended | | | GPIO | x2 | Can be used to output control signals to external expansion devices | | | Power Interface | x1 | Supports DC 12V / 4A | | Expandable Features | mSATA | x1 | Supports expansion with SSD (Solid State Drive) to increase storage space | | | WIFI | x1 | Expandable WIFI function, supports IEEE 802.11 a/b/g/n/ac/ax standard protocols | | | M.2 | x1 | Expandable 4G/5G module, supports wireless transmission of edge data | | | SIM Card Slot | x1 | Used with 4G/5G wireless communication module | | Item | Type | Model/Parameter | Description | | Linux | File System | ext4 | Ubuntu 20.04 LTS | | Software Development | Media Processing | BMCV, OpenCV, FFMPEG, BMLIB | | | | AI Development | TensorFlow, Caffe, PyTorch, MxNet, Paddle Lite, and other quantization to offline tools. | | | Basic Modules | Network Settings | Command Execution | Supports static and DHCP network parameter settings | | | Operating Status | CPU, Memory, Disk | | | | Device Information | Device serial number, software version number | | | | Log Management | Operating status, runtime errors, etc. | | | | Time | NTP, manual time synchronization | | | | Upgrade Management | Firmware Upgrade | Supports TF card upgrade, supports TFTP upgrade |
Application Areas Widely applied in various behavioral analysis fields such as facial recognition, smart construction sites, smart parks, smart transportation, and smart campuses.
