Nvidia CEO Jensen Huang introduced the new AI technology in the GTC 2021 keynote speech and
announced the launch of the new Jetson AGX Orin robotic computer for use in robots, autonomous machines, medical devices, and other forms of edge embedded computing, and the
new generation of InfiniBand network platform NVIDIA Quantum-2, providing cloud computing providers and supercomputing centers with extreme performance, extensive access capabilities and strong security.
The Jetson AGX Orin robot computer is a small, powerful, and energy-efficient AI supercomputer, using Arm Cortex-A78 architecture CPU and Ampere architecture GPU, and joined a new generation of deep learning and visual accelerators, which can perform 200 trillion operations per second (TOPS), which is comparable to servers equipped with GPUs. Compared with the previous generation Jetson AGX Xavier, the processing power is increased by 6 times, but the same size and pin compatibility are maintained.
Users can use NVIDIA CUDA-X accelerated computing stack, NVIDIA JetPack SDK, and the latest NVIDIA tools for application development and optimization, the pre-trained model from the NVIDIA NGC catalog has been optimized and can be fine-tuned using the NVIDIA TAO tool suite and customer data sets. This reduces the deployment time and cost of AI, while cloud-native technology enables seamless updates throughout the product life cycle.
NVIDIA Quantum-2 platform is a 400 Gbps InfiniBand network platform, including NVIDIA Quantum-2 switch, ConnectX-7 network card, BlueField-3 data processor DPU (data processor), and all software supporting this new architecture. This is the most advanced end-to-end network platform so far, tailored for supercomputing centers and cloud service providers.
The core of the platform is the new Quantum-2 InfiniBand switch, which uses a 7nm process and contains 57 billion transistors, which is even higher than the NVIDIA A100 GPU. It features 64 ports at 400Gbps or 128 ports at 200Gbps and will be offered in a variety of switch systems up to 2,048 ports at 400Gbps or 4,096 ports at 200Gbps — more than 5x the switching capability over the previous generation, Quantum-1.
Nvidia said that Jetson AGX Orin modules and developer kits will be available in the first quarter of 2022. The NVIDIA Quantum-2 platform has been supported by many infrastructure and system vendors, including HP, IBM, Inspur, Lenovo, Dell, and Super Micro.
