Nvidia to Unveil Blackwell Tech at Hot Chips

Nvidia to Unveil Blackwell Tech at Hot Chips

In Summary

  • Nvidia to reveal Blackwell tech at Hot Chips 2024.
  • Key highlights: Blackwell GPU, Grace CPU, and Quasar Quantization System.
  • Focus on AI hardware, data processing, and sustainable tech.
  • GB200 NVL72 with 72 GPUs and 36 CPUs to be showcased.


New York, Friday, August 30, 2024 – Nvidia said that it would showcase its cutting-edge Blackwell tech stack at the Hot Chips 2024 conference, with demonstrations starting this weekend and continuing through the main event next week.

This presentation is highly anticipated by tech enthusiasts eager to explore Nvidia’s latest advancements in AI hardware and data center technology.

The Blackwell platform represents Nvidia’s ambitious effort to redefine the AI landscape, offering a comprehensive ecosystem that extends beyond traditional GPU capabilities.

Key presentations will feature the Blackwell GPU, Grace CPU, and various advanced components such as the BlueField data processing unit and NVLink Switch. Nvidia’s roadmap, which includes the release of Blackwell Ultra in 2025 and future developments through 2027, will also be highlighted.

Nvidia will unveil its Quasar Quantization System at the conference, which promises significant improvements in data processing efficiency by combining algorithmic advancements with Blackwell’s second-generation Transformer Engine.

This system is designed to enhance FP4 operations while maintaining high-performance standards, marking a significant leap in AI-driven chip design.

Another focal point will be Nvidia’s GB200 NVL72, a liquid-cooled system featuring 72 Blackwell GPUs and 36 Grace CPUs.

The presentation will also cover Nvidia’s progress in sustainable tech solutions, including its participation in the U.S. Department of Energy’s COOLERCHIPS program, which aims to advance liquid cooling technologies.

In addition, Nvidia will discuss its use of agent-based AI systems in chip design, showcasing innovations that recently earned recognition at the IEEE International Workshop on LLM-Aided Design.

Protected by Copyscape