Nvidia’s CEO Jensen Huang just hand-delivered the DGX H200 GPU to OpenAI. The DGX H2000 is being termed the most powerful GPU in the world. The DGX series by Nvidia is the top-end lineup of GPUs mainly used in server-grade computers for deep learning. The DGX H200 is also twice as fast as its predecessor in GPT and Llama tasks and will help OpenAI speed up its AI development. Let’s take a closer look.
What is the Nvidia DGX H200?
The Nvidia DGX H200 is a server-grade GPU based on the Nvidia Hopper architecture. It comes with up to 141GB of HBM3e GPU memory and with speeds of up to 4.8TB/s, a first of its kind for any GPU. In simple words, the DGX 200 is the most powerful GPU released by Nvidia so far, which is specially curated for accelerating generative AI.
The DGX H200 is also 50% more power efficient than its predecessor, the DGX H100. The latest model also offers a performance boost of 40% in Llama2 13B, 60% in GPT-3 175B, and over 90% improvement in Llama2 70B. Each of these is a different version of large language models (LLMs) over which Meta AI and ChatGPT are built, respectively.
The Nvidia DGX H200 should not be confused with the DGX GH200. The DGX H200 is a GPU, whereas the DGX GH200 is a large memory supercomputer. Although both are developed by Nvidia to accelerate AI tasks, there is a significant performance difference between the two. Nvidia has delivered the DGX H200 to OpenAI, and not the DGX GH200.
First @NVIDIA DGX H200 in the world, hand-delivered to OpenAI and dedicated by Jensen "to advance AI, computing, and humanity": pic.twitter.com/rEJu7OTNGT
— Greg Brockman (@gdb) April 24, 2024
Nvidia Delivers DGX H200 to OpenAI: What’s the significance?
The DGX H200 is a high-demand GPU among enterprises due to its AI prowess. However, it is still unavailable to companies due to limited production. Nvidia and OpenAI are currently the top two names in the AI field, with one leading the hardware development while the other is racing the software side. Nvidia’s delivery of the world’s first DGX H200 GPU to OpenAI signifies the partnership between the two entities.
In addition to its face-product ChatGPT, OpenAI is working on several other AI projects, such as the Sora video generator, the alleged Q* AGI model, and many others. These AI tools require huge amounts of computing power, which is sourced from GPUs. The DGX H200, which is by far the most powerful GPU, can significantly help OpenAI accelerate its AI development.
Just a while back, ChatGPT by OpenAI was the only AI chatbot that was publically available. Over the last few months, it quickly received tough competition from Google Gemini, Claude, Perplexity, Meta AI, and many others. Some of these AI services were also better than OpenAI’s ChatGPT and DALL-E.
Hence, it has become necessary for OpenAI to keep up with the evolving AI space. Its partnership with Nvidia to get first access to top-end GPUs is expected to help the company. However, we can also expect the DGX H200 GPU to be made available to other giants like Google, Meta, Amazon, and others by the end of this year.
