Smart City Ascend Atlas 300I Duo Inference Card 256-channel real-time HD 280 TOPS 48GB Memory video applications
Ascend Hua Wei Atlas 300I Duo Inference Card
In the Smart City 2.0 city governance scenario, algorithms such as image classification, object detection, and image comparison are used to identify illegal roadside stalls, monitor food safety, and implement water governance, meeting the requirements of government departments for intelligent city governance and providing important support for service decision-making.the architecture of the Smart City 2.0 system. The system consists of front-end HD webcams or snapshot cameras, video analysis/inference clusters, resource scheduling servers, streaming media servers, platform application servers, and resource scheduling servers. The Hua Wei Ascend Atlas 300I Duo inference card is deployed in the video analysis/inference cluster to implement inference functions such as video decoding/preprocessing and identification of illegal roadside stalls, kitchen staff not wearing masks, and floating substances and illegal fishing in rivers.
Performance of Hua Wei Atlas 300I Duo Inference Card
Supreme computing power
– Up to 280 TOPS INT8 computing power per card, better support for central inference
– 16-core x 1.9 GHz CPU computing power
Ultra-high energy efficiency
Industry-leading energy efficiency ratio of 1.86 TOPS/W
High-performance video analysis
– Real-time analysis of 128-channel HD videos
– JPEG and video hardware encoding/decoding, boosting image and video applications.
Specifications of the Hua Wei Ascend Atlas 300I Duo Inference Card
AI processor | 2 x Ascend 310P AI Processors, including: |
Form factor | FHFL single-slot (10.5 inches) |
Half-precision (FP16) | 140 TFLOPS |
Integer precision (INT8) | 280 TOPS |
Memory | 48 GB LPDDR4X with a total bandwidth of 408 GB/s Error checking and correcting (ECC) |
Encoding/ Decoding capability | H.264 hardware decoding, 256-channel 1080p 30 FPS (32-channel 3840 x 2160 60 FPS) |
Virtual instance specifications | One Ascend AI Processor can be divided into several virtual NPUs in virtualization mode. Each virtual NPU supports 1, 2, or 4 AI Cores, and other hardware resources (such as memory and codec module) are divided proportionally. |
Power consumption | 150 W |
Dimensions (W x H x D) | 266.7 mm × 18.46 mm × 111.15 mm |
Weight (g) | 910g |
PCIe port | x16 lanes, compatible with x8/x4/x2 PCIe Gen4.0, compatible with 3.0/2.0/1.0 |
Operating Temperature | 0°C to 55°C (32°F to 131°F) |