Covering Scientific & Technical AI | Wednesday, December 11, 2024

Nvidia Unveils Two Edge AI Offerings for ‘Smart Everything’ 

Nvidia EGX Jetson Xavier NX

Extending latency-sensitive AI workloads to the edge is the objective of two new products from Nvidia, announced today at the company’s annual (virtual) GPU Technology Conference (GTC), based on its new Ampere GPU architecture and built to run on Nvidia’s EGX Edge AI platform announced last October.

Converged with Nvidia Mellanox’s new ConnectX-6 Dx SmartNIC over 5G networks, Nvidia said the new edge entries offer HPC-class throughput for massive amounts of streaming data from edge sensors, bringing about a fusion of IoT and AI for “smart everything” environments of the future.

The two products, the EGX A100 converged accelerator and EGX Jetson Xavier NX micro-edge server, were built for different size, cost and performance requirements. Nvidia said servers using the larger EGX A100 can manage hundreds of cameras in, for example, airports while the smaller EGX Jetson Xavier NX is built to manage a few cameras in convenience stores.

“Large industries can now offer intelligent connected products and services like the phone industry has with the smartphone,” said Jensen Huang, Nvidia founder and CEO. “Nvidia EGX Edge AI platform transforms a standard server into a mini cloud-native, secure, AI data center. With our AI application frameworks, companies can build AI services ranging from smart retail, robotic factories, to automated call centers.”

Nvidia said the EGX A100, available by the end of the year, is the first edge AI chip based on the NVIDIA Ampere architecture (see related story) and is designed for conducting real-time processing of large amounts of streaming data from edge sensors. When joined with Mellanox ConnectX-6 Dx network card, EGX A100-based servers can receive up to 200

Gbps of data and send it directly to the GPU memory for AI or 5G signal processing, Nvidia said. “With the introduction of NVIDIA Mellanox’s time-triggered transport technology for telco (5T for 5G), EGX A100 is a cloud-native, software-defined accelerator that can handle the most latency-sensitive use cases for 5G,” the company said.

Nvidia’s smaller new entry, the credit card-sized EGX Jetson Xavier NX, available now, is built for microservers and edge AIoT boxes constrained by size, weight, power budget or cost. Designed for embedded edge-computing devices, it delivers roughly the power of Nvidia’s Xavier SoC with throughput of up to 21 TOPS at 15W, or 14 TOPS at 10W, according to the company.

The EGX Edge AI platform has a cloud-native architecture that allows it to run containerized software for GPU-accelerated workloads, including application development frameworks for healthcare, telcos, conversational AI, robotics and smart cities, retail and transportation.

“NVIDIA Jetson and NVIDIA EGX are helping us transform retail, making the self-checkout experience quicker and more secure,” said Matt Scott, co-founder/CEO of Malong Technologies, an AI computer vision company. “It is now possible to accurately recognize hundreds of thousands of products in real time to create more seamless and protected shopping experiences, easily deployable at large scale. We’re continuing to explore NVIDIA’s powerful lineup to discover new ways to increase customer satisfaction and decrease retail shrink, by bringing more intelligence to the edge.”

AIwire