NVIDIA has officially launched its groundbreaking generation of AI supercomputer chips, poised to revolutionize the landscape of deep learning and supercomputing. The latest technology promises remarkable advancements, notably in the realm of large language models (LLMs) such as OpenAI’s upcoming GPT-4. The star of the show is the HGX H200 GPU, built on the innovative “Hopper” architecture, marking a significant leap forward from its predecessor, the H100.
This new chip introduces HBM3e memory, offering unprecedented speed and capacity. With 141GB of memory at 4.8 terabytes per second, the H200 doubles inference speed on large language models like Llama 2, boasting a 70 billion-parameter model. Its adaptability across various data centre types — on-premises, cloud, hybrid, and edge — sets a new standard, with major players like Amazon Web Services, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure gearing up for its deployment by Q2 2024.
Complementing the HGX H200 is the GH200 Grace Hopper “super chip,” a fusion of the HGX H200 GPU and the Arm-based NVIDIA Grace CPU. This powerhouse promises to elevate the capabilities of supercomputers, facilitating accelerated AI and high-performance computing applications processing massive datasets. Its adoption by renowned entities like Dell, Eviden, Hewlett Packard Enterprise (HPE), Lenovo, QCT, and Supermicro solidifies its potential impact.
The stage is set for the GH200 to take centre stage in over 40 AI supercomputers worldwide, including the formidable JUPITER at the Jülich facility in Germany, projected to be the “world’s most powerful AI system” upon its installation in 2024. With liquid-cooled architecture and an impressive interconnected network, this system is primed to drive significant advancements in climate and weather prediction, drug discovery, quantum computing, and industrial engineering.
NVIDIA’s aim to dominate the AI and data centre segments receives a significant boost from these cutting-edge innovations. As the company continues to set new records in revenue from these sectors, the introduction of these new chips is anticipated to further cement NVIDIA’s position as a frontrunner in the industry.
In a nutshell, these new chips are not just a leap forward for NVIDIA but a giant leap for AI and supercomputing as a whole, setting the stage for unprecedented advancements and breakthroughs in various fields.