NVIDIA Unveils Upgraded Hopper GPUs With HBM3e Memory, Grace Hopper Superchips
NVIDIA has recently announced a series of groundbreaking advancements in the field of artificial intelligence and supercomputing. Most notably, the tech giant revealed its latest addition to the Hopper GPU series, the HGX H200, which is set to revolutionize the landscape of AI computing.
The new H200 Tensor Core GPUs are powered by NVIDIA’s latest Hopper optimizations in both hardware and software, all while delivering the world’s fastest memory solution to date. These GPUs come equipped with Micron’s HBM3e memory solution, boasting memory capacities of up to 141 GB and up to 4.8 TB/s of bandwidth. This represents a significant leap in performance compared to NVIDIA’s previous A100 GPU, nearly doubling the AI inference performance.
The H200 GPUs will be available in a wide range of HGX H200 servers with 4 and 8-way GPU configurations, providing up to 32 PetaFLOPs of FP8 compute performance and 1.1 TB of memory capacities in an 8-way configuration. Moreover, these GPUs will be compatible with existing HGX H100 systems, streamlining the upgrade process for customers. Leading tech partners like ASUS, Dell, Lenovo, and more will offer updated solutions when the H200 GPUs become available in the second quarter of 2024.
In addition to the upgraded Hopper GPUs, NVIDIA also announced a major supercomputer win with its Grace Hopper Superchips powering the Exaflop Jupiter supercomputer, located at the Forschungszentrum Jülich facility in Germany. The Jupiter supercomputer, a part of the EuroHPC Joint Undertaking, will leverage NVIDIA’s cutting-edge superchips for applications in material science, climate research, and drug discovery. With a total of 24,000 NVIDIA GH200 Grace Hopper Superchips, interconnected using Quantum-2 Infiniband, the Jupiter supercomputer is expected to deliver 90 Exaflops of AI training and 1 Exaflop of high-performance compute.
These latest developments underscore NVIDIA’s commitment to pushing the boundaries of AI computing and supercomputing, solidifying its leading position in the industry. As the company continues to innovate and drive forward powerful hardware and software technologies, the future of AI and supercomputing looks more promising than ever.