Product details

product description 

NVIDIA GH200 Superserver AI AI Trends

At COMPUTEX 2023, NVIDIA announced the NVIDIA DGX GH200, which marks another 

breakthrough for GPU-accelerating computing to power the most demanding giant AI 

workloads. In addition to describing key aspects of the NVIDIA DGX GH200 architecture, 

this article discusses how to use NVIDIA Base Command to enable rapid deployment, 

speed user onboarding, and simplify system administration.

The unified in-memory programming model for Gpus has been the cornerstone of various 

breakthroughs in complex accelerated computing applications over the past seven years. In 

2016, NVIDIA introduced NVLink technology and a unified memory programming model 

with CUDA-6 designed to increase the memory available for GPU-accelerated workloads.

Since then, the core of every DGX system has been a GPU complex on a substrate interlinked 

with NVLink, where each GPU can access the memory of the other at the speed of NVLink. 

Many DGXS with GPU complexes are interconnected through high-speed networks to form 

larger supercomputers, such as the NVIDIA Selene supercomputer. However, an emerging 

class of trillion-parameter giant AI models either require months of training or cannot be 

solved even on today's best supercomputers.

To empower scientists who need an advanced platform capable of solving these extraordinary 

challenges, the NVIDIANVIDIA Grace Hopper Superchip with the NVLink switching system 

integrates up to 256 Gpus in the NVIDIA DGX GH200 system. In the DGX GH200 system, the 

GPU shared memory programming model can access 144 TB of memory at high speed via NVLink.

Compared to a single NVIDIA DGX A100 320 GB system, the NVIDIA DGX GH200 provides nearly 

500 times more memory for the GPU shared memory programming model via NVLink, resulting 

in a huge data-center sized GPU. The NVIDIA DGX GH200 is the first supercomputer to break the 

100 TB barrier of GPU-accessible memory on NVLink.