By NewsBlogger Nvidia unveiled its next-generation Blackwell GPUs at the GTC 2024 event, promising significant performance gains and energy efficiency for AI processing.
At a glance
- Nvidia unveils next-generation Blackwell GPUs for AI processing with improved energy consumption and cost efficiency.
- Blackwell GPUs promise up to 30 times performance increase for large language model inference workloads.
- Key components of the Blackwell platform include Nvidia GB200 Grace Blackwell Superchip and Project GR00T for humanoid robots.
- DGX SuperPod, powered by Blackwell GPUs, offers 240 terabytes of memory and 11.5 exaflops of AI supercomputing power.
- Nvidia’s advancements in AI technology aim to revolutionize the computing landscape with unparalleled performance and efficiency.
The details
Nvidia has recently unveiled its next-generation Blackwell GPUs, which offer significant improvements in energy consumption and cost efficiency for AI processing.
The Nvidia GB200 Grace Blackwell Superchip, a key component of the Blackwell platform, promises exceptional performance gains, providing up to a 30 times increase in performance for large language model (LLM) inference workloads.
This announcement was made by Nvidia CEO Jensen Huang at the Nvidia GTC 2024 event, where he emphasized that the Blackwell GPUs will usher in a transformative era in computing.
Blackwell GPUs Features
The Blackwell GPUs are specifically designed to power generative AI and enable real-time processing of trillion-parameter large language models at a significantly lower cost and energy consumption compared to its predecessor.
The platform introduces six pioneering technologies across various sectors, including a second-generation transformer engine and a fifth-generation NVLink networking technology.
Additionally, the RAS Engine ensures reliability, availability, and serviceability, while the secure AI solution safeguards AI models and customer data.
Key Components of Blackwell Platform
One of the key components of the Blackwell platform is the Nvidia GB200 Grace Blackwell Superchip, which connects two Nvidia B200 Tensor Core GPUs to the Nvidia Grace CPU via a 900GB/s ultra-low-power chip-to-chip link.
This super chip offers up to a 30 times performance increase for LLM inference workloads compared to the Nvidia H100 Tensor Core GPU. Nvidia is also offering the HGX B200 server board, which can link eight B200 GPUs to support x86-based generative AI platforms.
In addition to the Blackwell GPUs, Nvidia has introduced Project GR00T, a multimodal AI system designed for humanoid robots.
This project utilizes a general-purpose foundation model to process text, speech, videos, and live demonstrations.
Nvidia’s Isaac Robotic Platform tools, including Isaac Lab for reinforcement learning, were instrumental in the development of Project GR00T. The Jetson Thor chip has also been announced to support enterprises running Project GR00T, showcasing Nvidia’s advancements in AI-powered industrial manipulation arms and robots for navigating unstructured environments.
Furthermore, Nvidia has launched the DGX SuperPod, its most powerful system to date, powered by the next-generation Blackwell GPUs.
This system is designed to support AI models with a trillion parameters and integrates the GB200 superchip version of the Blackwell, which includes both CPU and GPU resources.
The DGX SuperPod can be configured with eight or more DGX GB200 systems, offering 240 terabytes of memory and 11.5 exaflops of AI supercomputing power.
It features Nvidia Quantum-X800 InfiniBand networking technology, BlueField-3 DPUs, and the fifth generation of NVLink interconnect.
Overall, Nvidia’s latest announcements highlight the company’s commitment to advancing AI technology and revolutionizing the computing landscape with innovative solutions that offer unparalleled performance and efficiency.
Article X-ray
Sources
Facts attribution
This section links each of the article’s facts back to its original source.
If you suspect false information in the article, you can use this section to investigate where it came from.
| venturebeat.com |
|---|
| – Nvidia unveiled its next-generation Blackwell GPUs with 25 times better energy consumption and lower costs for AI processing – The Nvidia GB200 Grace Blackwell Superchip promises exceptional performance gains, providing up to 30 times performance increase for LLM inference workloads – Nvidia CEO Jensen Huang unveiled Blackwell at Nvidia GTC 2024, stating it will herald a transformative era in computing – Blackwell GPUs are designed to power generative AI and enable real-time generative AI on trillion-parameter large language models at 25 times less cost and energy consumption than its predecessor – The Blackwell platform introduces six pioneering technologies across various sectors – Blackwell features a second-generation transformer engine and a fifth-generation NVLink networking technology – The RAS Engine ensures reliability, availability, and serviceability, while the secure AI solution safeguards AI models and customer data – The Nvidia GB200 Grace Blackwell Superchip connects two Nvidia B200 Tensor Core GPUs to the Nvidia Grace CPU over a 900GB/s ultra-low-power chip-to-chip link – The GB200 Superchip provides up to a 30 times performance increase compared to the Nvidia H100 Tensor Core GPU for LLM inference workloads – Nvidia offers the HGX B200 server board that links eight B200 GPUs to support x86-based generative AI platforms – Nvidia DGX Cloud, an AI platform co-engineered with leading cloud service providers, will host new Nvidia Grace Blackwell-based instances later this year – Cisco, Dell Technologies, Hewlett Packard Enterprise, Lenovo, and Supermicro are expected to deliver servers based on Blackwell products |
| venturebeat.com |
|---|
| – Nvidia has introduced Project GR00T, a multimodal AI for humanoid robots – Project GR00T uses a general-purpose foundation model to process text, speech, videos, and live demonstrations – Nvidia’s Isaac Robotic Platform tools, including Isaac Lab for reinforcement learning, were used to develop Project GR00T – The Jetson Thor chip has been announced by Nvidia to help enterprises run Project GR00T – Nvidia has made advancements in AI-powered industrial manipulation arms and robots for navigating unstructured environments – Project GR00T stands for Generalist Robot 00 Technology and is designed to understand natural language and human movements – Nvidia demonstrated GR00T-powered humanoid robots completing tasks from various companies – OpenAI is also working on embodied AI and has backed two startups in the domain – Nvidia has launched the Jetson Thor computing platform for humanoids to run complex multimodal models – The Jetson Thor platform includes a high-performance CPU cluster and next-generation GPU based on Nvidia Blackwell architecture – Nvidia used its Isaac Robotics Platform to develop Project GR00T – The Isaac Robotics Platform is getting two use-case targeted offerings: Isaac Manipulator and Isaac Perceptor – Isaac Manipulator offers GPU-accelerated libraries and foundation models for robotic arm manufacturers – Isaac Perceptor guides robots through unstructured environments with multi-camera, 360-degree vision capabilities – Nvidia is working with partners to advance autonomous mobile robot functions in manufacturing and fulfillment – The new Isaac platform capabilities are expected to be available in the second quarter of this year – Project GR00T remains in early access, with Nvidia accepting applications for more humanoid developers to access the technology. |
| venturebeat.com |
|---|
| – Nvidia is launching its most powerful systems yet with the new DGX SuperPod – The DGX SuperPod system is powered by Nvidia’s next generation of GPUs for AI acceleration, known as Blackwell – The Blackwell is being positioned by Nvidia to support and enable AI models that have a trillion parameters – The DGX SuperPod integrates the GB200 superchip version of the Blackwell, which includes both CPU and GPU resources – Existing DGX systems from Nvidia are widely deployed for numerous use cases – The DGX SuperPod can be configured with eight or more DGX GB200 systems and can scale to tens of thousands of GB200 Superchips – The system can deliver 240 terabytes of memory and has 11.5 exaflops of AI supercomputing power – The DGX SuperPod features Nvidia Quantum-X800 InfiniBand networking technology – The DGX also integrates the Nvidia BlueField-3 DPUs and the fifth generation of the Nvidia NVLink interconnect – The new SuperPOD includes fourth-generation Nvidia Scalable Hierarchical Aggregation and Reduction Protocol (SHARP) technology – The new GB200-based DGX systems are coming to the Nvidia DGX cloud service – The GB200 capabilities will be available first on Amazon Web Services (AWS), Google Cloud, and Oracle Cloud – The new GB200 will help to advance the Project Ceiba supercomputer that Nvidia has been developing with AWS, supporting 20,000 GPUs and delivering over 400 exaflops of AI |
