NVIDIA, a global leader in AI computing, continues to revolutionize the AI industry with the launch of GH200-powered systems. These systems join an extensive lineup of over 400 configurations, combining various combinations of NVIDIA’s latest CPU, GPU, and DPU architectures, including NVIDIA Grace, NVIDIA Hopper, NVIDIA Ada Lovelace, and NVIDIA BlueField. The purpose of this diverse range of systems is to meet the ever-increasing demand for generative AI applications.
During the recent COMPUTEX event, Jensen Huang, the founder and CEO of NVIDIA, unveiled the GH200 Grace Hopper Superchip, providing more insights into the technology, partners, and systems associated with it. The GH200 chip brings together the power of the Arm-based NVIDIA Grace CPU and the Hopper GPU architectures using the cutting-edge NVIDIA NVLink-C2C interconnect technology. This groundbreaking combination results in an impressive total bandwidth of up to 900GB/s, which is seven times higher than the bandwidth offered by traditional accelerated systems utilizing standard PCIe Gen5 lanes. With this incredible compute capability, GH200-powered systems are poised to address the most demanding generative AI and high-performance computing (HPC) applications.
Ian Buck, the Vice President of Accelerated Computing at NVIDIA, emphasized the transformative impact of generative AI on businesses across various industries, such as healthcare, finance, and business services. He highlighted how the accelerated infrastructure provided by GH200 Grace Hopper Superchips will enable enterprises worldwide to build and deploy generative AI applications that leverage their unique proprietary data.
GH200-powered systems will be accessible to a wide range of customers, including global hyperscalers and supercomputing centers in Europe and the U.S., further underscoring the widespread adoption and significance of this technology.
The NVIDIA ecosystem comprises an extensive network of system manufacturers and cloud partners who are instrumental in bringing GH200-powered systems to market. Notable manufacturers include AAEON, Advantech, Aetina, ASRock Rack, ASUS, GIGABYTE, Ingrasys, Inventec, Pegatron, QCT, Tyan, Wistron, and Wiwynn. Additionally, global server manufacturers such as Cisco, Dell Technologies, Hewlett Packard Enterprise, Lenovo, Supermicro, and Eviden (an Atos company) offer a diverse range of NVIDIA-accelerated systems. Leading cloud partners, including Amazon Web Services (AWS), Cirrascale, CoreWeave, Google Cloud, Lambda, Microsoft Azure, Oracle Cloud Infrastructure, Paperspace, and Vultr, also support the NVIDIA H100 architecture.
The upcoming portfolio of systems accelerated by NVIDIA’s Grace, Hopper, and Ada Lovelace architectures ensures comprehensive support for the NVIDIA software stack. This includes NVIDIA AI, the NVIDIA Omniverse platform, and NVIDIA RTX technology. NVIDIA AI Enterprise, the software layer of the NVIDIA AI (Human Like Video games)platform, offers over 100 frameworks, pretrained models, and development tools, simplifying the development and deployment of production AI, including generative AI, computer vision, and speech AI. The NVIDIA Omniverse development platform facilitates collaboration and seamless work across multiple software suites in real time, utilizing the power of the Universal Scene Description framework. Finally, the NVIDIA RTX platform integrates ray tracing, deep learning, and rasterization, enabling content creators and developers to unleash their creativity and enhance graphics, video, and image processing with real-time photorealistic rendering and AI-enhanced capabilities.
GH200 Grace Hopper Superchips are expected to be available in systems starting later this year, marking a significant milestone in the advancement of generative AI and computational capabilities. NVIDIA’s relentless commitment to innovation continues to push the boundaries of AI technology, empowering businesses and researchers to unlock new possibilities and accelerate their path to success.
