Cloud provider Vultr has added Nvidia's GH200 Grace Hopper to its cloud GPU offering for AI training and inference.
The Nvidia GH200 chips, a combination of CPU and GPU that uses the Nvidia Grace and Hopper architecture, will be available at all 32 of Vultr's cloud data center locations. The GH200 is designed for giant-scale AI and high-performance computing (HPC) workloads, and is up to seven times faster than the PCle Gen5.
This latest Nvidia addition to Vultr's GPU offering joins the HGX H100, A100 Tensor Core, L40S, A40, and A16 GPUs.
“The Nvidia GH200 Grace Hopper Superchip delivers unrivaled performance and TCO for scaling out AI inference. We are excited to be one of the first cloud providers to deliver global access to this essential technology,” said J.J. Kardwell, CEO of Vultr’s parent company, Constant.
“With our global rollout of the Nvidia GH200 Grace Hopper Superchip, AI innovators now have access to the most powerful GPU for AI inference and the ability to access it worldwide to support their local market latency, data sovereignty, compliance, and privacy goals.”
“The Nvidia GH200 Grace Hopper Superchip was purpose-built to meet today and tomorrow’s most data-intensive problems head-on,” said Matt McGrigg, director, global business development, Nvidia cloud partners.
The Nvidia GPUs are integrated with Vultr's managed Kubernetes, virtualized cloud computing, and bare metal offerings.
The updated GH200 was announced by Nvidia in August 2023 and is the first GPU to include HBM3e memory. It comprises a single server with 144 Arm Neoverse cores, eight petaflops of AI performance, and 282GB of HBM3e memory.
Nvidia has also recently revealed its HGX H200 GPU server boards, its latest high-end GPU featuring 141GB of HBM3e and a 4.8TBps memory bandwidth. Each server board will be available in four and eight-way configurations. Vultr is set to deploy H200 instances starting next year.
Vultr is a cloud provider with a presence in at colocation facilities in 32 locations spread across North America, Latin America, Europe, Asia, Australia, and Africa.
More in Cloud & Hyperscale
Episode Is AI an investment hype cycle?