More and more industries today are shifting toward intelligence, and the century-old automotive industry is no exception. Automakers, suppliers, startups and others are embracing this wave and moving forward into the age of autonomous driving.
Azera, a global smart electric vehicle company, was officially launched on November 25, 2014. Azera is committed to creating an enjoyable lifestyle for its users by providing high-performance smart electric vehicles with the ultimate user experience. Azera uses NVIDIA DRIVE Orin system-on-chip to build its self-developed, next-generation autonomous driving platform, which remains a leader in autonomous driving research and development in the smart electric vehicle space, meeting the AI infrastructure and in-vehicle computing needs of software-defined vehicles.
NVIDIA helps Azera build a mobile computing platform
Autonomous Driving Faces Computing Challenges
The development of self-driving vehicles is a compute-intensive endeavor that requires significant acceleration of the computing infrastructure. In order to provide the ultimate driving experience for users in more scenarios and to implement higher-level autonomous driving algorithms, software-defined vehicles need to implement redundancy, configure multiple high-resolution sensors, and apply deep neural networks. All of these factors will create a need for powerful computing power to be able to handle all AI workloads.
In addition, the data collection vehicles used for R&D will generate a hundred petabytes of data and billions of images per year, and mass production vehicles will need to return massive amounts of road data for optimizing algorithm models, which means that vehicle companies’ data centers need to have the ability to return, filter and label massive amounts of data for AI model training.
A robust technical infrastructure is essential for successful technology application at scale, including an AI platform and large-scale machine learning system covering the entire scenario chain. NVIDIA GPUs have been used to develop a complete set of autonomous driving infrastructure system.
NVIDIA Enabled Supercomputer
Azure’s supercomputer is built with NVIDIA HGX A100 8-GPU and NVIDIA Mellanox InfiniBand ConnectX-6. The supercomputer cluster consists of supercomputers and high-performance NVME SSD servers with NVIDIA Mellanox InfiniBand ConnectX-6, connected via the NVIDIA Mellanox InfiniBand high-speed network, and growing in size.
By using NVIDIA Multi-Instance GPUs wisely, the efficiency of algorithm engineers in the development process and the utilization of cluster resources have been greatly improved. Meanwhile, Azure has built a deep learning training framework that supports multi-card and multi-machine training based on NVIDIA NVLink and NVIDIA Mellanox InfiniBand, using NVIDIA Mellanox InfiniBand and RDMA technology based on NVIDIA Mellanox InfiniBand. built a distributed, high-performance file read acceleration suite. In data-driven application scenarios, it enables large amounts of deep learning training data to be transferred to supercomputer memory or NVIDIA A100 video memory at ultra-high speed of 200Gbps with zero copies, further solidifying the infrastructure service architecture of the autonomous driving platform.
In addition, four NVIDIA DRIVE Orin system-on-chips with 8096 CUDA cores on each end provide over 1000 TOPS of computing power for production vehicles. Orin can handle the large number of applications and deep neural networks running simultaneously in a self-driving car, and meets system safety standards such as ISO 26262 ASIL-D.
With the performance benefits of DRIVE Orin, Azera is able to provide the redundancy and flexibility necessary for safe autonomous driving functions and a comfortable and secure driving experience.
Azera’s first sedan, the ET7, is the flagship smart electric sedan with autonomous driving capability, and the ET7 utilizes the latest Azera autonomous driving technology, NIO Autonomous Driving (NAD). Azera’s supercomputing platform Adam is equipped with four NVIDIA DRIVE Orin, with an arithmetic power of 1016 TOPS. Azera has established a full-stack autonomous driving technology capability, which will gradually realize easy and safe point-to-point autonomous driving experience in high-speed, urban, parking and power-up scenarios.
Accelerating Autonomous Driving Product Landing
NVIDIA HGX A100 is an acceleration server platform for AI scenarios with large data sets, complex models, etc. The HGX A100 integrates NVIDIA Ampere GPUs, NVIDIA NVLink, NVIDIA Mellanox InfiniBand network, and a fully integrated GPU in NGC (NVIDIA GPU Cloud). The HGX A100 integrates NVIDIA Ampere GPUs, NVIDIA NVLink, NVIDIA Mellanox InfiniBand network and the fully optimized NVIDIA AI software stack in NGC (NVIDIA GPU Cloud). At the same time, it sets a new benchmark for compute density, condensing 5 PETAFLOPS of AI performance and replacing traditional siloed infrastructures for various AI workloads with a single platform. With HGX A100, Azera is free to build and apply AI platforms in a vertically scalable and horizontally scalable way, which also makes Azera’s model development 20 times more efficient, helping Azera land self-driving products faster and evolve to newer and faster architectures.
On the other hand, NVIDIA DGX POD reference architecture combines NVIDIA accelerated computing architecture, Mellanox network architecture and system management software. Based on this, Azera achieves a solution that provides the ability to make supercomputing power accessible to the masses, easy to access, install and manage, thus meeting the needs of complex architectures for autonomous driving.
High-performance AI computing is a key component of Azera’s autonomous driving technology, and with the strong support of NVIDIA DRIVE end-to-end solutions, Azera can evolve and design new autonomous driving features from vast amounts of real-world data on the road to continue to provide owners with a peace of mind driving experience.
NVIDIA is committed to providing leading technologies and solutions to help automakers accelerate the development and production implementation of software-defined vehicles.
NVIDIA’s high performance computing solutions have become an accelerator for Azera on the road to autonomous driving,” said Yuli Bai, head of Azera’s AI platform. NVIDIA’s high performance computing solutions become the gas pedal for Azera on the road to autonomous driving.”
ASUS ROG-STRIX-RTX3060-O12G-V2-GAMING NVIDIA Ampere Streaming Multiprocessor, Axial Fan Design, GPU Tweak II
[Product Price] RMB 9999
Intel Core i7 12700K
[Product Price] $3199
DELL G15 15.6-inch gaming laptop (11th generation Intel Core i7-11800H 16G 512G RTX3060 165Hz 100% sRGB) Black
[Product Price] 8999RMB