The world's largest technology event, CES (International Consumer Electronics Show), will officially open on January 9, local time in the United States.
CES has always focused on consumer electronics, and on this year's CES stage, generative AI has become the absolute protagonist. The industry not only pays attention to how AI is implemented on consumer terminals, but also pays great attention to the upgrading of the underlying computing power chips, such as GPUs and CPUs.
The day before the opening, the chip giant has started the prelude. On January 8, local time, NVIDIA released the GeForce RTX 40 Super series, a desktop GPU with high-performance generative AI capabilities, including RTX 4080 Super, RTX 4070 Ti Super, and RTX 4070 Super.
These new products will be the first to be used in PC products such as AI laptops, and at the same time, NVIDIA also announced new developments in autonomous driving computing platforms and robotics scenarios on the same day.
Li Auto's next-generation electric vehicles will use NVIDIA's Drive Thor computing platform, and Great Wall Motor, Zeekr, and Xiaomi Auto will use Drive Orin to develop AI-enabled vehicles. Boston Dynamics, Collaborative Robotics, Covariant, Sanctuary AI, Unitree, and others in the robotics field are using GPU-accelerated large language models.
There's no doubt that generative AI is being integrated into every industry. Among them, PC products, as the world's general computing power scenario, are one of the mainstream terminals on the C-side, if the PC graphics card ecology is the basic disk of NVIDIA today, then the smart car and robot scene is the rising star that NVIDIA bets on. CES has always focused on consumer electronics, and on this year's CES stage, generative AI has become the absolute protagonist.
As generative AI makes its way into computer applications, the wave of AI PCs is surging. The GPU is a key role in AI, and NVIDIA is also accelerating its layout.
First of all, look at the new series of GPUs,**In terms of aspects,RTX 4080 Super will be available on January 31,The price starts at 8099 yuan; The RTX 4070 Ti Super was launched on January 24, with prices starting at 6499 yuan; The RTX 4070 Super went on sale on January 17, with prices starting at 4,899 yuan. At a time when high-performance computing power is in short supply, NVIDIA's GPUs will continue to create profits for it.
From the point of view of parameters, the new GPU has more cores and faster video memory, and Nvidia also emphasized that the new product further upgrades AI performance. For example, dedicated AI Tensor Cores can provide up to 836 AI Tops, which can enhance the gaming experience, improve creative capabilities, and improve daily work efficiency.
According to Nvidia, the RTX 4080 Super supports 4K ray-traced games, and its generation** is 1. faster than the previous RTX 3080 Ti5 times more and 1 faster to generate images7 times more; The RTX 4070 Ti Super delivers enhanced gaming performance at ultra-high frame rates of 1440p and even 4K. The RTX 4070 Super has 20% more cores than the RTX 4070, surpassing the performance of the previous generation 3090 with less power consumption.
It is reported that brands such as Acer, Asus, Dell, HP, Lenovo, MSI and other brands will release AI laptops based on RTX new GPUs. Nvidia says RTX AI laptops can deliver 20 to 60 times faster performance compared to using an NPU.
"Generative AI is the most important platform shift in computing history that will transform all industries, including gaming," said Nvidia founder and CEO Jensen Huang, who currently has a user base of more than 100 million RTX AI PCs and workstations. ”
In addition, NVIDIA works with developers and partners to release generative AI-powered PC applications and services. For example, NVIDIA ACE microservices, which include generative AI-powered voice and animation models that enable developers to add intelligent, dynamic virtual digital characters to their games; NVIDIA RTX Remix, for example, provides generative AI tools that transform basic textures from classic games into physically accurate, high-precision 4K materials.
Talking about the popularity of AI PC on CES, Li Yayu, IT research director of Qunzhi Consulting, told the 21st Century Business Herald reporter: "The main PC manufacturers and chip manufacturers will attach great importance to this exhibition, Intel, AMD, NVIDIA, Qualcomm will have a special new product launch conference to introduce the performance and power consumption of new products such as AI processors and GPUs, which is to promote the technical productivity of AI PC core, and it is also the cornerstone of the first generation of AI PC products brought by AI machine manufacturers." ”
It can be seen that from hardware, software to application ecology, NVIDIA continues to consolidate its computing power position in the PC field and continues to expand in generative AI.
Outside of PCs, automobiles have been the biggest focus at CES in recent years, and this year is no exception. In the transformation of automotive intelligence, NVIDIA is also deeply involved, launching NVIDIA Drive, an AI computing platform for intelligent driving, and has iterated multiple series of autonomous driving chips.
NVIDIA's DRIVE series chip platform is to be the strongest brain of the car, and at present, it has a market share of more than 80% in automotive-grade high-computing power autonomous driving chips.
In 2022, NVIDIA released a new generation of chip computing platform Drive Thor, which is an upgraded version of NVIDIA's Drive Orin chip. According to public information, this SoC chip has 77 billion transistors inside, which can achieve 2000 TOPS of AI computing power, and is planned to be put into production in 2025.
Mass-produced in 2022, the DRIVE Orin has a computing power of up to 254TOPS and can support L2+ to L5 autonomous driving functions. Since 2022, Drive Orin has been widely used by leading automakers, trucking companies, robotaxis, and shuttles.
If you look at the computing data, Drive Tor is nearly 8 times that of Drive Orin, which can be described as a "performance monster" and contains greater ambitions. According to NVIDIA, Drive Thor, a next-generation centralized in-vehicle computing platform, can integrate multiple intelligent functions such as autonomous driving, automatic parking, driver and occupant monitoring, and AI cockpit on a single AI computing platform.
Today, Nvidia announced that Li Auto has chosen to use the Drive Thor platform in its next-generation models, and it also announced that Great Wall Motor, Zeekr, and Xiaomi Auto have adopted the Drive Orin platform in their next-generation autonomous driving systems.
Specifically, Li Auto has launched an intelligent driving system AD MAX on its L-series models, equipped with dual Drive Orin processors and a total computing power of 508 TOPS. The system provides functions such as full-scene intelligent driving (NOA), full-scene assisted driving, intelligent parking, and active safety. AD Max 3 after the upgradeThe 0 system will switch to the end-to-end algorithm architecture based on large models, and deploy the occupancy network, spatio-temporal joint planning algorithm and MPC model ** control algorithm, which improves safety performance and intelligent driving experience.
Xiaomi Auto's first electric vehicle, SU7, will use NVIDIA's dual-drive Orin system-on-chip to build its intelligent driving system, support high-speed NOA function, and rely on a large model of perception decision-making, aiming to achieve the Chinese city navigation function regardless of city, level, and road, and Xiaomi SU7 will be officially launched in the first half of 2024.
Great Wall Motor will use DRIVE Orin's self-developed high-end intelligent driving system Coffee Pilot, which can open up parking, high-speed and urban scenarios in the future, and realize full-scene assisted driving functions without high-definition maps. The first model using this system is scheduled to be unveiled in the first half of 2024, when high-end intelligent driving functions such as urban NOA and cross-level memory parking will be installed at WEY first.
Geely's ZEEKR 007 pure electric luxury sedan will be equipped with a new full-stack self-developed intelligent driving system, equipped with dual DRIVE ORIN system-on-chips, which can realize intelligent parking and intelligent assisted driving on highways and urban roads.
It can be seen that the rise of new energy vehicles has also opened up more emerging markets for NVIDIA. In the field of autonomous driving chips, in addition to NVIDIA, car manufacturers are also developing their own chips.
Danny Shapiro, vice president of NVIDIA's automotive business unit, told the 21st Century Business Herald: "Cross-industry cooperation is essential to ensure the safe development of software-defined vehicles, and the more participants, the better it is for the entire transportation industry." When automakers work with key technology partners like NVIDIA, there are substantial benefits to rolling out advanced hardware and software technologies faster and more efficiently, as well as reduced R&D and engineering costs compared to trying to develop the technology on its own, in addition to a number of cloud services. Author: Ni Yuqing.