SAN JOSE, Calif. - NVIDIA (NASDAQ:NVDA) has announced the launch of its Blackwell GPU architecture, designed to significantly enhance computing power and efficiency for organizations using large language models (LLMs) and generative AI. The Blackwell GPU is expected to reduce the cost and energy consumption of operating LLMs by up to 25 times compared to its predecessor.
The Blackwell platform includes six new technologies aimed at accelerating AI training and real-time inference for models with up to 10 trillion parameters. Among these technologies is the world's most potent chip, containing 208 billion transistors and a second-generation Transformer Engine that supports new 4-bit floating point AI inference capabilities.
Another feature is the fifth-generation NVLink, which provides 1.8TB/s bidirectional throughput per GPU, enhancing performance for complex LLMs. Additionally, a dedicated RAS Engine for reliability, availability, and serviceability, along with a decompression engine to accelerate database queries, are part of the Blackwell architecture.
The NVIDIA GB200 Grace Blackwell Superchip is a notable component of the new architecture, connecting two B200 Tensor Core GPUs to the NVIDIA Grace CPU, facilitating AI performance with a 30x increase in efficiency over NVIDIA H100 Tensor Core GPUs for LLM inference workloads.
The Blackwell platform has garnered widespread interest, with major cloud providers, server makers, and leading AI companies expected to adopt it. Notable endorsements have come from tech leaders such as Sundar Pichai of Alphabet (NASDAQ:GOOGL) and Google, Andy Jassy of Amazon (NASDAQ:AMZN), Michael Dell of Dell Technologies (NYSE:DELL), and Elon Musk of Tesla (NASDAQ:TSLA) and xAI.
NVIDIA's Blackwell-based products will be available through a global network of partners later this year, with cloud service providers like AWS, Google Cloud, Microsoft (NASDAQ:MSFT) Azure, and Oracle (NYSE:ORCL) Cloud Infrastructure among the first to offer instances powered by Blackwell.
The Blackwell product portfolio is also supported by NVIDIA AI Enterprise, an operating system for production-grade AI that includes inference microservices, frameworks, libraries, and tools deployable across NVIDIA-accelerated clouds, data centers, and workstations.
This announcement is based on a press release statement from NVIDIA.
This article was generated with the support of AI and reviewed by an editor. For more information see our T&C.