Nvidia announced its new DGX GH200 supercomputer at Computex 2023. The DGX GH200 is designed to handle terabyte-class models for massive recommender systems, generative AI, and graph analytics.
The DGX GH200 is powered by 256 Grace Hopper Superchips, each of which contains four Hopper GPUs and 1 terabyte of HBM3 memory. The system also includes 160 GB of NVLink interconnect bandwidth, which allows the GPUs to communicate with each other at speeds of up to 64 GB/s.
Nvidia claims that the DGX GH200 can deliver up to 1.5 exaflops of performance on AI workloads. This makes it the most powerful AI supercomputer on the market.
The DGX GH200 is available now for purchase. Pricing starts at $1.99 million.
Table of Contents
Here are some of the key features of the DGX GH200:
256 Grace Hopper Superchips
1 terabyte of HBM3 memory per super chip
160 GB of NVLink interconnect bandwidth
Up to 1.5 exaflops of AI performance
Benefits of Using the DGX GH200
The DGX GH200 is the most powerful AI supercomputer on the market.
The system is capable of delivering up to 40 exaflops of performance, making it ideal for a wide range of AI workloads.
The DGX GH200 is also capable of training and deploying large language models, such as GPT-3.
Here are some of the applications that the DGX GH200 is designed for:
Massive recommender systems
The DGX GH200 is a significant step forward for AI computing. It is the most powerful AI supercomputer on the market and is capable of handling some of the most demanding AI workloads. The DGX GH200 is a major investment, but it is a valuable tool for organizations that need to accelerate their AI initiatives.
Join Telegram and WhatsApp for More updates
Follow us on social media