share_log

英伟达推出仅249美元的AI超级电脑 生成式AI性能提高达1.7倍

NVIDIA has launched an AI supercomputer priced at only $249, with generative AI performance improved by 1.7 times.

wallstreetcn ·  Dec 17 23:55

NVIDIA launched a generative AI supercomputer called Jetson Orin Nano Super on Tuesday. NVIDIA CEO Jensen Huang stated that this developer kit achieved higher performance and a Low Stock Price through software upgrades, selling for only $249. It is compact, palm-sized, and can enhance generative AI performance by up to 1.7 times.

Author: Zhao Yuhua

Source: Hard AI.

NVIDIA launched a generative AI supercomputer called Jetson Orin Nano Super on Tuesday. NVIDIA CEO Jensen Huang stated that this developer kit achieved higher performance and a Low Stock Price through software upgrades, selling for only $249. It is compact, palm-sized, and can enhance generative AI performance by up to 1.7 times.

big

The company stated that this product is suitable for commercial AI developers, enthusiasts, and students. Compared to its predecessor, its generative AI inference performance has improved by 1.7 times, with a performance increase of 70%, capable of executing up to 67 trillion INT8 operations per second (67 INT8 TOPS), and memory bandwidth has also increased by 50%, reaching 102GB/s.

Jensen Huang also specially recorded a video for this product, similar to when the Hopper H100 was released in 2021, demonstrating NVIDIA's emphasis on this product. In the video, Huang stated that since the Xavier, NVIDIA has created a new type of processor, which is a robot processor, and at that time, no one knew what they were doing.

The new Jetson Orin Nano Super has nearly 70 million operations capability, with a power of 25 watts, priced at $249, and can perform all functions of HGX, even executing large language models.

big

He stated that whether creating a large language model (LLM) chatbot based on Retrieval-Augmented Generation (RAG), building a visual AI agent, or deploying AI-based robots, Jetson Orin Nano Super is an ideal choice. This kit can provide powerful computing performance at a lower cost, mainly due to its support for low-power design, capable of running on a power consumption of 7 watts to 25 watts, suitable for various edge device cloud computing scenarios, whether it be smart cities, smart agriculture, or robot development.

Continuously improving performance and experience.

Through software updates, Jetson Orin Nano Super will also enhance the generative AI performance for existing users of the Jetson Orin Nano developer kit.

Jensen Huang mentioned that Jetson Orin Nano Super is very suitable for those who wish to develop skills in generative AI, robotics, or computer vision. In the trend of shifting from task-specific models to foundational models in the AI field, it also provides developers with an easy-to-use platform to turn their creativity into reality.

Provides strong performance for generative AI.

Jetson Orin Nano Super enhances performance for all popular generative AI models, such as Meta's LLaMA, Alibaba's Tongyi Qianwen, and Google's Gemma. This kit significantly enhances the performance of transformer-based computer vision models.

big

The company stated that the developer kit includes a Jetson Orin Nano 8GB system module (SoM) and a reference carrier board, providing an ideal platform for prototyping AI applications for edge devices.

The system module features an NVIDIA Ampere architecture GPU, equipped with Tensor Cores and a 6-core Arm CPU, capable of handling multiple parallel AI application pipelines and providing high-performance inference capabilities. In addition, it supports up to four cameras, with resolutions and frame rates surpassing previous generations.

NVIDIA emphasized that as AI technology gradually shifts from the cloud to the edge, physical AI will be the next major development focus. The Jetson Orin Nano Super can be used to develop chatbots and is also suitable for applications such as resume visual AI agents or deploying AI robots.

A wide range of generative AI software ecosystem and Community.

Additionally, the NVIDIA Jetson AI Lab provides support for the latest open-source community models and offers easy-to-use tutorials. Developers can also receive support from the broader Jetson community and draw inspiration from projects by other developers.

Jetson supports NVIDIA's AI software, including NVIDIA Isaac for robotics, NVIDIA Metropolis for visual AI, and NVIDIA Holoscan for sensor processing. By generating synthetic data through NVIDIA Omniverse Replicator and fine-tuning pre-trained AI models from the NGC catalog using the NVIDIA TAO Toolkit, development time can be significantly reduced.

Jetson ecosystem partners provide additional AI and System Applications, development tools, and custom Software Development services. They also offer support related to cameras and other Sensors, as well as design services for carrier boards and product solutions.

Unlock performance enhancement for all Jetson Orin users: Super Mode.

NVIDIA also announced that it will provide software updates for the Jetson Orin NX and Nano series system modules, allowing them to enjoy performance improvements in Super mode, with a maximum increase of 70%. These updates will be released through the Jetpack SDK Manager on the NVIDIA website and GitHub.

big

The Jetson Orin Nano Super Developer Kit is expected to be available soon through Global Dealers. This development kit offers an affordable and powerful platform option for students, engineers, or makers looking to get started in AI development.

NVIDIA positions the Jetson Orin Nano Super Developer Kit as a key product to promote the popularization of generative AI technology, hoping it will shine in Intelligent Transportation, retail management, and Internet of Things (IoT) applications.

This article is from the WeChat public account, "YingAI." For more AI news, please click here.

Disclaimer: This content is for informational and educational purposes only and does not constitute a recommendation or endorsement of any specific investment or investment strategy. Read more
    Write a comment