GTC19 | NVIDIA announces multiple innovative technologies and collaborations to accelerate AI implementation

Latest update time：2019-12-19

Reads：

From December 16 to 19, GTC CHINA 2019 (GPU Technology Conference) was held in Suzhou. GTC China is NVIDIA's most important annual conference in China. This year, it attracted more than 6,000 scientists, engineers and entrepreneurs . NVIDIA founder and CEO Huang Renxun delivered a keynote speech at the GTC China conference, announcing a number of innovative technologies and cooperation, and explained how NVIDIA helps AI accelerate into large-scale applications.

As an authorized distributor of NVIDIA , China Electronics Port appeared at GTC CHINA 2019 with a new distribution business model , which received unanimous praise from the professional audience of the parliament. NVIDIA is a leading manufacturer in the field of GPUs today. It has been committed to technological innovation and has worked with partners to lead the trend of the industry. As a partner of NVIDIA , China Electronics Port leads the model innovation of the component industry with the concept of "new distribution". In the future, we will jointly provide customers with more services and solutions in the fields of big data and AI.

TensorRT 7: Enabling smarter AI human-computer interaction

During his speech, Huang announced the release of NVIDIA's breakthrough inference software, NVIDIA TensorRT™ 7. With this software, developers around the world can implement conversational AI applications and significantly reduce inference latency. As NVIDIA's seventh-generation inference software development kit, TensorRT™ 7 opens the door to more intelligent AI human-computer interaction, enabling real-time interaction with applications such as voice agents, chatbots, and recommendation engines.

TensorRT 7 includes a new deep learning compiler that automatically optimizes and accelerates recurrent neural networks and converter-based neural networks. Conversational AI components are more than 10 times faster than when running on a CPU, reducing latency to below the 300 millisecond threshold required for real-time interaction.

“We have entered a new era of AI where machines can understand human language in real time,” said Jensen Huang. “TensorRT 7 makes this possible, providing developers around the world with the tools to build and deploy faster, smarter conversational AI services, enabling more natural AI-machine interactions.”

Accelerating automotive innovation

NVIDIA DRIVE™, an open-source deep neural network for autonomous driving

In the automotive and transportation space, Huang announced that NVIDIA will open source NVIDIA DRIVE™ deep neural networks for autonomous vehicle development to the transportation industry on the NVIDIA GPU Cloud (NGC) container registry.

NVIDIA DRIVE has become the industry standard for autonomous vehicle development and is widely adopted by automakers, truck manufacturers, self-driving taxi companies, software companies, and universities. Today, NVIDIA is open sourcing its pre-trained AI models and training code to autonomous vehicle developers. Through a suite of NVIDIA AI tools, developers within the NVIDIA ecosystem can freely extend and customize the models to improve the robustness and capabilities of their autonomous driving systems.

“AI self-driving cars are software-defined cars that must be based on large data sets to operate globally,” said Huang. “We are open-sourcing our deep neural networks to self-driving car developers and providing them with advanced learning tools that enable them to optimize these networks based on different data sets. In this way, we are enabling shared learning across companies and countries while protecting data ownership and privacy, ultimately accelerating the deployment of self-driving cars around the world.”

Release of Orin™, a software-defined autonomous driving platform

In addition, Huang Renxun also released NVIDIA DRIVE AGX Orin™, a highly advanced software-defined platform for autonomous driving and robotics. The platform has a built-in new Orin system-level chip, which consists of 17 billion transistors and integrates NVIDIA's next-generation GPU architecture and Arm Hercules CPU cores as well as new deep learning and computer vision accelerators. It can run 200 trillion calculations per second, which is almost 7 times the performance of NVIDIA's previous generation Xavier system-level chip. Orin can handle a large number of applications and deep neural networks running simultaneously in autonomous vehicles and robots, and has achieved system safety standards such as ISO 26262 ASIL-D.

As a software-defined platform, DRIVE AGX Orin can empower compatible architecture platforms for the development of fully autonomous vehicles from L2 to L5, helping OEMs develop large and complex software product lines. Since both Orin and Xavier can be programmed through open CUDA, TensorRT APIs and various libraries, developers can use products across multiple generations after a one-time investment.

Cooperation with Didi

In terms of customer cooperation, Huang Renxun announced that NVIDIA and Didi Chuxing, the world's leading mobile travel platform, have reached a cooperation, and Didi will use NVIDIA GPUs and other technologies to develop autonomous driving and cloud computing solutions. Didi will use NVIDIA® GPUs in data centers to train machine learning algorithms and use NVIDIA DRIVE™ to provide reasoning capabilities for its L4 autonomous driving cars. Didi upgraded its autonomous driving department to an independent company in August and has carried out extensive cooperation with industry chain partners.

As part of Didi’s autonomous driving AI processing, NVIDIA DRIVE uses multiple deep neural networks to fuse data from various sensors (cameras, lidar, radar, etc.) to achieve a 360-degree understanding of the car’s surroundings and plan a safe driving path.

NVIDIA GPUs Accelerate Singles’ Day Shopping Festival

When it comes to the power of AI, there’s no better example than Alibaba, where Huang announced that Alibaba recently deployed AI at scale using NVIDIA GPUs to serve customers during the annual shopping festival, Singles’ Day.

During this year's "Double Eleven" shopping festival, Alibaba generated sales of $38 billion, an increase of nearly a quarter compared to last year's $31 billion, and more than double the combined online sales of "Black Friday" and "Cyber Monday".

Recommendation systems are a critical application for Alibaba and other leading online retailers. They show users products that match their preferences, thereby increasing click-through rates, which have always been an important driver of sales in the e-commerce industry. Every small improvement in click-through rate directly affects the user experience and the merchant's revenue. Using advanced recommendation system models that can run in real time and at scale can increase click-through rates by 10%, which is currently only possible with GPUs.

The Alibaba team uses NVIDIA GPUs to support three optimization strategies: resource allocation, model quantization, and graph transformation, thereby improving throughput and responsiveness. This enables NVIDIA T4 GPUs to accelerate Alibaba's extensive and deep recommendation models, processing 780 queries per second. This is far ahead of CPU-based reasoning, which can only process 3 queries per second.

Cooperated with Tencent to launch START cloud gaming service

On the gaming front, Huang announced a collaboration between NVIDIA and Tencent Games to bring PC games to the cloud, further increasing NVIDIA’s influence in cloud gaming.

NVIDIA’s GPU technology powers Tencent Games’ START cloud gaming service, which has been in beta since earlier this year. START enables gamers to play AAA games anywhere, anytime, even on underpowered devices. Tencent Games plans to expand its cloud gaming offerings to provide millions of players with a gaming experience that is consistent with local gaming devices.

NVIDIA continues to expand its footprint in cloud gaming by supporting the START platform. Earlier this year, NVIDIA announced a partnership with Japan's SoftBank and South Korea's LG U+ to develop cloud gaming solutions. In addition, NVIDIA also provides GeForce NOW™ cloud gaming services in North America and parts of Europe, providing gamers around the world with the best exclusive cloud PC gaming experience.

About China Power Port

China Electronics Port is an industrial component application innovation platform under China Electronics Information Industry Group (CEC). Through its three major business segments, namely, China Electronics Equipment and AiSiYi, which focus on component distribution, Firefly Factory, which serves the design chain, and Yiancang, which develops supply chain collaborative supporting services, it provides the industry with comprehensive industrial supply chain service solutions with electronic components as the core.

China Electronics Port adheres to the business philosophy of "serving customers and sharing with partners". While fulfilling its social responsibilities, it explores a "new distribution" business model driven by application innovation and supported by supply chain services. It has begun to transform from a single enterprise platform to a platform enterprise serving the industry, and strives to build a component supply chain ecosystem to help the development of China's electronic components industry.

Latest articles about

■Event Preview｜Xinchacha will meet you at the ADI Product Application Technology Salon

■NXP & China Power Port Power Energy and Internet of Things Technology Seminar was successfully held in Xiamen

■[Domestic self-developed] Office scene test! Firefly Factory CEM5855H domestic 24GHz millimeter wave radar module!

■New Product Release | Renesas Launches New RA8 Entry-Level MCU Product Group, Providing Cost-Effective, High-Performance Arm Cortex-M85 Processor

■China Electric Port has been named "China Brand Distributor of the Year" for 23 consecutive years

■On November 14, the NXP & China Power Port Power Energy and Internet of Things Technology Seminar will meet you in Xiamen!

■[Review and get gifts] Get a free trial of the domestic open source hardware Feiteng Pi kit for RMB 699, and get cultural and creative gifts!

■Xinchacha Report | September 2024 Component Supply Chain Monitoring Report - Xinchacha Index 1+7 Analysis

■Read the full white paper on 48V-12V DC-DC converters

■[Word of Mouth Review] There is a surprise at the end of the article! Review of the domestic human presence detection millimeter wave radar module CEM5825F from Firefly Factory!