Graphic processing unit manufacturer NVIDIA announces smallest GPU based powerful computing device for Edge computing. The NVIDIA Jetson Xavier NX delivers terra operations per second (TOPS) at 10W or 21 terra operations per second at 15W. Yet the size of this device is smaller than credit card. This makes it a future device for large scale Edge computing which involves running of machine learning model for predictions in real time on the Edge devices.
Company is expected to ship GPU based NVIDIA Jetson Xavier NX in Mach 2020. The cost of this Edge computing device will be around USD $399 per unit.
Company announced this device on 6th November. This device will be another addition to the Jetson family of NVIDIA devices. In the beginning of this year NVIDIA launched Jetson Nano, a smallest and powerful edge computing device.
The NVIDIA Jetson Xavier NX is powerful edge computing device that can be used to run machine learning models developed in major AI frameworks including TensorFlow, PyTorch, MXNet, Caffe and others.
The processing power of NVIDIA Jetson Xavier NX is very good and it can operate on 14 terra operations per second (TOPS) at 10W or 21 terra operations per second at 15W. The NVIDIA Jetson Xavier NX is designed to run multiple neural networks in parallel. This Edge computing device is so powerful that it can process data from multiple sensors simultaneously in Nano form factor.
The NVIDIA Jetson Xavier NX Edge device also supports CUDA-X AI software and this helps the developer to take full advantage of this feature. Developers can develop optimized deep learning models to run in production efficiently. Xavier NX is powered by JetPack software which includes all the components and libraries for training/inferring neural network machine learning models.
The NVIDIA Jetson Xavier NX Edge device is a complete system that comes with CPU, RAM and GPU combination. This device is fitted with a 6-core Carmel Arm 64-bit CPU, 6MB of L2 and 4MB of L3 cache CPU. Device comes with 8GB of 128-bit LPDDR4x RAM, which is capable of fast data transfer at a speed of 51.2GB/second. This device is so powerful that it can support up to six CSI cameras over 12 MIPI CSI-2 lanes.
As far as operating system is concerned this device is powered with Ubuntu-based Linux operating system.
According to the recent MLPerf inferencing benchmark results announced by company, the Xavier ranked as the highest performer under various scenarios.
The NVIDIA Jetson Xavier NX Â is CUDA compatible that insures that pre-trained neural network with mainstream framework will work on this device. Developers can run the neural network trained in main stream frameworks such as TensorFlow, Caffe, PyTorch, and MXNet on Â NVIDIA Jetson Xavier NX Edge device.
It is also possible to use NVIDIA SDK and other tools to convert the pre-trained models into TensorRT before deploying on the Edge device. It is noted that TensorRT is a software layer used optimized for inferencing. As of now the Jetson Nano and Xavier NX are the most affordable Edge computing available in market today. Next year once Xavier NX is released, developers will be able to use the device for high performance Edge computing.
564 total views, 2 views today