www.industryemea.com
19
'24
Written on Modified on
Acceed News
New AI platform supports assistance systems
German distributor Acceed is expanding its NRU series for Edge AI applications, boasting a fanless chassis for high-temperature operation and robust inference performance.
The NRU series controllers are pre-destined for Edge computing installations in confined spaces, e.g. for AI-based video analysis, for monitoring and steering tasks or for manufacturing controls. The robust and fanless NRU-220S and Nuvo-222S controllers have six RJ45 or M12 Ethernet connections, two of which have 2.5 GbB bandwidth. PoE with an overall capacity of 100 W is available for the remaining four interfaces.
Based on the latest Jetson AGX Orin 32/64GB module, the new NRU-220S box controller provides up to 275 INT8 TOPS AI capacity as a fanless GPU computer. Thanks to its optimised design for operation at temperatures from -25 to +70 °C, two 2.5 GbE connections and four further GbE connections with PoE, the devices form the perfect AI inference system for use on the road, in vehicles and for intelligent video analyses in Smart city applications, to name just a few application examples.
In particular, for use in vehicles, the NRU-220S provides important functions such as IPC (Ignition Power Control), it is impact and vibration-proof and has safe connection mechanisms for connected components. The model NRU-222S is equipped with M12 connectors and also meets the European standards EN 50155 and EMC, in order to guarantee stable operation in difficult surroundings. This enables seamless use of the controller in industrial vehicles, robotics and rail vehicles and can be used there, for example, for ADAS (driver assistance systems).
NRU-220S and NRU-222S act primarily as AI-NVRs (network video recorders) with real time inference and as video transcoders. They are driven by the SOM Jetson AGX Orin 32/64GB from Nvidia. Jetson-Orin modules provide up to 275 billion operations per second (TOPS) and 8 times the performance of the previous generation for several simultaneous AI inference pipelines. The ampere GPU architecture with 2048 CUDA cores, 64 tensor cores and two NVDLA 2.0 engines supports up to 22 simultaneous video flows (H.265).
The four PoE connections (802.3at) share the bandwidth of 1 gigabit; each interface can supply IP cameras with up to 25.5 W power. The additional two 2.5 GbE connections are ideal for monitoring applications which require even more bandwidth to the backend. In addition to 64 GB eMMC on the Orin module and an M.2 socket for fast SSD read/write processes, the NRU-220S and NRU-222S are equipped with two 2.5 inch plug-in units for storage expansion which are accessible from the front. Moreover, they have two mini PCIe connections for CAN / COM / WiFi modules and an M.2-B key connection for mobile communication (4G/5G).
To complement the hardware, the Nvidia system also provides a wide range of AI tools and modern deep learning frameworks. NRU-220S and NRU-222S provide real time video data for monitoring, preventive maintenance and intelligent transport systems (ITS). The NRU-22xS series is the response to current performance requirements from Edge AI, with a compact form factor and fanless, also when operated at high environmental temperatures.
www.acceed.com
