Micro Server SE8-416

SOPHON micro server SE8 is a new generation of edge computing product with high-performance, low-power consumption and small size. Equipped with the fourth-generation tensor processor BM1684X launched by SOPHGO, SE8 supports mixed-precision inference, while computing power is up to 416 TOPS @ INT8 and 208 TFLOPS @ FP16/BF16. Compared with SE6, SE8 doubles the computing power and density, improving the encoding capacity by nearly 5 times as well. It can process up to 416 channels of HD video deep learning intelligent analysis at the same time, widely used in large concurrent application scenarios.

Features

Surging Deep Learning Performace
Peak Deep learning performance up to 416TOPS @ INT8, 208TFLOPS @ FP16/BF16, mixed-precision inference supported
Excellent Computing Performance
Support 416-channel HD video whole-process processing and 416-channel full HD video hardware decoding and 156-channel encoding
High Computing Density
High-density integration and tight deployment, 13*BM1684X in a single device to reduce TCO
Easy Migration
Developer-friendly SOPHON SDK, supporting Caffe/TensorFlow/PyTorch/ MXNet/Paddle Lite and other mainstream deep learning frameworks
Strong Adaptability
Working temperature ranging from 0 ℃ to +40 ℃ with portable design
Rich Algorithms
Support algorithm migration of pedestrian/vehicle / non-vehicle / object recognition, video structurization, trajectory behavior analysis, etc.
Various Scenarios
Widely used in public security / traffic / industrial parks / producing / warehousing, etc.
Flexible deployment in multiple fields and scenarios
Flexible Scalability
Multi-server cluster deployment, Ethernet connection, Unified management services and cloud deployment

Application Scenarios

Intelligent storage

Intelligent Public Security

Intelligent Governance

Safety production

Intelligent storage

Automatic monitoring of violations and abnormal events such as smoke and fire, material stacking, blockage of fire escapes, lack of safety signs in hazardous chemical warehouses, video obscuring, vehicle intrusion, item movement, personnel sleeping, smoking, dress code violations, intrusion in restricted areas, etc.

Intelligent Public Security

Full target Video Structuring including human, vehicle and face-recognition, behavior analysis, target capture, traffic event analysis. Widely used in counter-terrorism, anti-drug, security management, police case handling, intelligence command etc.

Intelligent Governance

City appearance, street order, municipal facilities, social events, outdoor advertising, urban firefighting, natural disaster-related monitor, alarm and closed-loop event disposal control

Safety production

Supervision of production safety in hazardous chemical enterprises, oil and gas sites, three-small sites, and construction sites

Specifications

Product Model

SE8 416-BA2-61

Deep learning Capacity

Supports FP32/FP16/BF16/INT8, capable of handling up to 416 channels of high-definition video intelligent analysis

Video Decoding

H.264 & H.265: 1080P @10400fps

Video Decoding Resolution

8K / 4K / 1080P / 720P / D1 / CIF

Video/JPEG Codec

Video Encoding

H.264 & H.265: 1080P @3900fps

Video Encoding Resolution

4K / 1080P / 720P / D1 / CIF

JPEG Decoding

JPEG: 7800 pages/second @1080P

Device Memory

192GB LPDDR4X

Deep learning Co-processor

ARM A53 96-core @2.3GHz

Dimension&Weight

148mm(W)*140mm(H)*300mm(D),5.6kg(weight)

Host Processor

ARM A53 8核 @2.3GHz

Host Memory

16GB LPDDR4X

Storage

832GB

Interface

3x 10/100/1000Mbps,Expandable 1 x 2.5-inch SATA hard drive,1x UART(USB Type C) ,1x Micro SD,1x HDMI,2x USB3.0

Power Supply

1 x 500W gold power supply, 100-240V~, 7-3.5A, 60-50Hz

Working Temperature

0°C ~ 40°C

Operation System

Ubuntu/Kylin

Deep Learning Framework

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