Edge computing solutions
The NPU computing power integration solution combines RK3399 and multiple RK1808 core boards to form an AI cluster with ultra-high computing power, which can perform AI computing processing at the same time. The NPU computing power integrated solution can not only increase computing power, but also control costs flexibly and has a higher cost performance.
For example, with one RK3399 as the main board and 8 RK1808s, the object recognition analysis in 8 channels of videos can be performed simultaneously under the Ubuntu system.
Features
(1) Configure AI dedicated chip: RK1808
The built-in NPU computing power of RK1808 can reach up to 3.0 TOPs, supports INT8/INT16/FP16 mixed operation, takes into account performance, power consumption and operation accuracy to the greatest extent, supports network model conversion of TensorFlow, Caffe, ONNX, Darknet and other frameworks, open source code , Tutorials and other technical information.
(2) Flexible increase of computing power and cost control
One RK3399 can be combined with multiple RK1808 core boards to form a high-performance AI acceleration cluster. The NPU computing power can increase by 1+N, and the computing efficiency will be greatly improved. The number of RK1808 to be used can be flexibly selected according to the needs of the scene, effectively controlling the cost.
(3) Extremely low power consumption
The chip adopts 22nmFD-SOI process, the power consumption can be reduced by about 30% compared with the mainstream 28nm process under the same performance; built-in 2MB system-level SRAM, which can realize the operation of always-on equipment without DDR; has hardware VAD function, supports low-power listening Far-field wakeup.
(4) Rich interfaces for AIoT applications
With abundant peripheral interfaces, it is convenient for application expansion. Video supports MIPI/CIF/BT1120 input, supports MIPI/RGB display output; has a series of sensor input and output interfaces such as PWM/I2C/SPI/UART; has high-speed device interfaces such as USB3.0/USB2.0/PCIE, supports Gigabit Ethernet and external WiFi/BT module; audio supports microphone array input and audio output at the same time.