.Ted Hisokawa.Aug 23, 2024 14:53.NVIDIA engineers to reveal improvements in the Blackwell platform, liquid cooling, and AI-driven potato chip layout at Hot Chips 2024.
At the upcoming Hot Chips 2024 conference, NVIDIA developers are actually set to present groundbreaking innovations focused on improving records facility functionality as well as power effectiveness. The occasion, scheduled for August 25-27 at Stanford University, will definitely include 4 speaks by senior NVIDIA engineers, according to NVIDIA Weblog.NVIDIA Blackwell Platform.NVIDIA's discussions will certainly highlight the NVIDIA Blackwell platform, which integrates various chips, bodies, and NVIDIA CUDA software program to drive the future generation of artificial intelligence applications. The NVIDIA GB200 NVL72, a multi-node, liquid-cooled answer, will additionally be actually showcased for its own capability to link 72 Blackwell GPUs and 36 Compassion CPUs, setting a brand new criteria for AI system layout. In addition, the NVLink relate innovation will be actually discussed for its own high throughput and also low-latency capabilities, crucial for generative AI.Fluid Cooling Innovations.A notable section of the conference will definitely concentrate on fluid air conditioning innovations. Ali Heydari, supervisor of data center air conditioning and also structure at NVIDIA, are going to present designs for hybrid-cooled information centers. These designs strive to retrofit existing air-cooled records facilities with liquid-cooling units, using an even more efficient and space-saving service. Heydari's group is actually also collaborating along with the USA Division of Energy on the COOLERCHIPS course to create state-of-the-art air conditioning modern technologies. Using the NVIDIA Omniverse system, they are actually generating electronic twins to model electricity consumption and cooling down productivity.AI-Driven Potato Chip Layout.NVIDIA is actually also leveraging artificial intelligence to enhance semiconductor design. Mark Ren, director of style automation research study at NVIDIA, will certainly review just how AI versions strengthen layout quality as well as performance through automating lengthy duties. These models feature prophecy as well as optimization resources that help engineers analyze and improve styles swiftly. AI agents powered by big foreign language models (LLMs) are being established to autonomously complete duties, interact along with professionals, and gain from a data source of human as well as representative experiences. Ren is going to discuss examples of AI applications in time report evaluation, tissue cluster optimization, as well as code production.These discussions highlight NVIDIA's commitment to pressing the limits of information facility computing with cutting-edge modern technologies in artificial intelligence as well as cooling options. The firm's initiatives intend to provide unparalleled functionality, performance, as well as marketing across various sectors.Image resource: Shutterstock.