EC-A3399C(AI Edition) Six-core Al Embedded Computer

EC-A3399C(AI Edition) Six-core Al Embedded Computer
Based on AIO-3399C (Al version) high-performance open-source platform, industrial-grade metal case and fan-off design configured, with high-efficiency cooling ability, complete interface, strong applicability and computing power and ultrahigh performance. It can be applicable to high-performance edge computing to accelerate in-depth learning computation and Al algorithm. it can be rapidly applied in the field of mobile edge computing, smart home, face recognition, etc

64-bit High-performance Core
ARM Cortex-A72 architecture and six-core 64-bit high-performance processor carried, frequency 2.0GHz reached, integrated quad-core Mali-T860 GPU, H.265 HEVC, VP9 H.264 encoding and 4K HDR supported, with strong ability of hard decoding as high as 4K.

Ultrahigh Performance
Onboard Al embedded neural network processor NPU, with 28000 parallel neural calculation cores, on-chip parallel and in-situ computation supported, with a peak performance reaching up to 5.6 Tops and 2.8 Tops computing performance, with an efficiency energy-consumption ratio reaching up to 9.3 Tops/W. It has strong computing power. In the meantime, extremely low energy consumption is also maintained, making it extremely advantageous in the application of edge computing of terminal device.

Unique Al Architecture
Dedicated ME matrix engine and APiM (Al processing in Memory) framework for Al adopted, local parallel Al computation integrated with storage and calculation. one-time upgrade network preload with no need for instructions, bus, and external DDR cache. This configuration improves the processing speed hugely and lowers processing energy consumption a lot compared with processors built in traditional architecture approaches.

Supporting Model Training Tool
Complete easy-to-use model-training tool PLAI (People Learn Al) based on PyTorch is offered. It can be developed on Windows 10 and Ubuntu 16.04 system, and user-defined network model is added easier, greatly lowering the technical threshold of using Al.

Provide Network Training Model
Three kinds of network training model sample including GNet1, GNet18 and GNetfc based on VGG are supported. Subsequently, network samples will be continuously added, contributing to easy test of a large amount of in-depth learning application on device.

Custom Metal Case
High-quality metal case and fanless design configured, heat-conducting aluminum alloy structure, high-efficiency cooling, 60°C high-temperature aging, stable operation for 7x24 hours, various installation ways contributing to its embedding into various smart devices more convenient.

Rich External Interfaces
HDMI2.0, LAN, Type-C, USB3.0, RS485, R$232 and other interfaces owned to make it possible to be widely applied in various industries.