On November 6th, NVIDIA introduced the latest member of the Jetson family - the Jetson Xavier NX. With the size thats smaller than a credit card, this module packs a punch.

Earlier this year, NVIDIA launched Jetson Nano, the smallest yet the most powerful GPU-based edge computing device. Jetson Xavier NX, much-advanced edge computing device, has the pin compatibility with Jetson Nano making it possible to port the AIoT applications deployed on the Nano. It also supports all major AI frameworks, including TensorFlow, PyTorch, MXNet, Caffe and others.

Jetson Xavier NX

According to NVIDIA, Jetson Xavier NX delivers up to 14 terra operations per second (TOPS) at 10W or 21 terra operations per second at 15W, running multiple neural networks in parallel and processing data from multiple high-resolution sensors simultaneously in a Nano form factor. Like other Jetson products, the Xavier NX runs on CUDA-X AI software which makes it easy to optimize deep learning networks for inference at the edge. Similar to Jetson Nano and Jetson TX2, Xavier NX is powered by JetPack software that comes with all the components need to train and infer neural networks.

When it comes to the CPU, the Xavier NX is powered by a 6-core Carmel Arm 64-bit CPU, 6MB of L2 and 4MB of L3 cache. The device can support up to six CSI cameras over 12 MIPI CSI-2 lanes. It comes with 8GB of 128-bit LPDDR4x RAM, which is capable of performing data transfer at 51.2GB/second. The device runs an Ubuntu-based Linux operating system.

Priced at USD 399, The Jetson Xavier NX device is expected to be shipped in March of 2020.

In the recently announced MLPerf inferencing benchmark results, Jetson Xavier has helped NVIDIA top the inferencing benchmark results. According to NVIDIA, Xavier ranked as the highest performer under both edge-focused scenarios (single- and multi-stream) among commercially available edge and mobile SoCs.

MLPerf Benchmark

Though there are AI accelerators available from Google, Intel, and Qualcomm, what differentiates NVIDIA is the compatibility of the software stack. With an increased number of GPU cores, the Jetson family of devices can not only be used for inferencing but also for performing training on the device. CUDA compatibility ensures that neural networks trained in mainstream frameworks such as TensorFlow, Caffe, PyTorch, and MXNet can be run on these devices with no conversion or optimization. But for increased performance, NVIDIA ships SDK and tools to convert the trained models into TensorRT, a software layer optimized for inferencing.

Jetson Nano and Xavier NX are the most affordable and powerful edge computing devices available in the market. With CUDA-X AI software and support for popular deep learning frameworks, developers are adopting the Jetson family for running AI at the edge.

When it becomes available early next year, Jetson Xavier NX will enable developers and businesses to run sophisticated AI-infused applications at the edge.

NVIDIA is moving fast with its AI strategy. Having captured the majority of the AI training segment with K80, P100, and T4 GPUs, NVIDIA is now eyeing the inferencing segment with the Jetson family of products.

Read the original here:

NVIDIA Jetson Xavier NX Debuts As The Smallest Super Computer For AI At The Edge - Forbes