With the introduction of NVIDIA’s revolutionary GH200 Grace Hopper Superchip designed for extensive AI and high-performance computing applications, a multitude of cutting-edge supercomputers tailored for medical technology are poised to come online rapidly.
By expediting complex AI and HPC tasks that involve processing terabytes of data, the NVIDIA GH200 empowers scientists and researchers to tackle some of the world’s most challenging problems effectively.
During the SC23 supercomputing event, NVIDIA disclosed that the superchip would be integrated into more systems globally, including those developed by Dell Technologies, Eviden, Hewlett Packard Enterprise (HPE), Lenovo, QCT, and Supermicro.
The GH200 functions as the cornerstone of medical computing facilities worldwide by merging the Arm-based NVIDIA Grace CPU and Hopper GPU architectures with NVIDIA NVLink-C2C interconnect systems.
Collectively, these GH200-powered facilities represent approximately 200 exaflops of AI capabilities, driving technological progress in various domains.
NVIDIA Grace Hopper Integration in HPE Cray Supercomputers
At the Denver event, HPE revealed its collaboration with NVIDIA to incorporate the NVIDIA Grace Hopper Superchip with HPE Cray EX2500 microprocessors. This integrated solution aims to provide unparalleled supercomputing flexibility and accelerated AI training by leveraging multiple GH200 chips, scaling up to tens of thousands of Grace Hopper Superchip networks. This configuration will also be a key component of HPE’s newly introduced supercomputing solution for generative AI.
Justin Hotard, Executive Vice President and General Manager of HPC, AI, and Labs at HPE, emphasized that “organizations are swiftly embracing generative AI to drive business transformations and technological advancements. Through our partnership with NVIDIA, we are thrilled to deliver a comprehensive computing solution for generative AI, powered by technologies like Grace Hopper, facilitating seamless large-scale AI model training and deployment at unprecedented levels of efficiency.”
Next-Generation AI Supercomputing Centers
NVIDIA Grace Hopper technologies are driving a diverse array of computing facilities worldwide. At SC23, several prominent centers announced plans to integrate GH200 systems into their supercomputers.
The Jülich Supercomputing Center in Germany is considering the utilization of GH200 superchips in its upcoming system, Mercury, which is anticipated to be Europe’s foremost large-scale computer. This supercomputer will play a pivotal role in addressing critical medical challenges such as mitigating climate change, combating pandemics, and fostering the advancement of sustainable energy solutions.
In Japan, the Joint Center for Advanced High Performance Computing, a collaborative initiative between the Centers for Computational Sciences at the University of Tsukuba and the Information Technology Center at Tokyo, is spearheading advancements in advanced mathematical sciences, data analysis, AI, and machine learning. Their forthcoming computer will harness the power of NVIDIA Grace Hopper.
The Texas Advanced Computing Center, based in Austin, is renowned for creating and managing some of the most powerful computing resources globally. By leveraging NVIDIA GH200 for low-power, high-bandwidth memory, the center aims to enhance computational efficiency and facilitate the operation of larger models on its Vista computer.
DeltaAI, an advanced technology and information resource poised to triple the AI-focused processing power at NCSA, may leverage NVIDIA Grace Hopper superchips with the support of the National Center for Supercomputing Software at the University of Illinois Urbana-Champaign.
Furthermore, the UK government has allocated funding to the University of Bristol for the development of Isambard-AI, intended to be the nation’s most potent computer, driving AI-driven advancements in robotics, big data analytics, weather research, and pharmaceutical discoveries. The program will incorporate over 5,000 NVIDIA GH200 Grace Hopper Superchips, provided by HPE, delivering 21 exaflops of AI supercomputing prowess capable of executing 21 quintillion AI calculations per second.
These initiatives complement existing endeavors by SoftBank Corp., the Swiss National Supercomputing Center, Los Alamos National Laboratory, and recently announced next-generation Grace Hopper technologies.
Global GH200 Deployment and Expedited Access via CSPs
Initial access to GH200 is being offered by select cloud service providers, including Lambda and Vultr. CoreWeave has outlined plans to introduce its GH200 instances in Q1 2024, while Oracle Cloud Infrastructure has confirmed its intentions to provide access to these resources.
By the year’s end, systems equipped with the GH200 superchips will commence shipping from various technology firms such as ASRock Rack, ASUS, GIGABYTE, and Ingrasys.
Over 100 businesses, organizations, and government entities globally, including the NASA Ames Research Center for Aeronautics Research and multinational energy conglomerate TotalEnergies, have embraced NVIDIA Grace Hopper through early access programs for supercomputing initiatives.
Moreover, the GH200 will soon be accessible via NVIDIA LaunchPad, an online platform offering complimentary access to NVIDIA hardware and software solutions.
