in ,

Jen-Hsun Huang personally went to OPENAI to deliver the world’s first NVIDIA DGX H200 AI acceleration server


NVIDIA is currently working hard to produce more artificial intelligence accelerator cards suitable for data center levels. The current flagship model is the NVIDIA H200 GPU. While artificial intelligence companies such as OPENAI have already pre-ordered, artificial intelligence companies that pre-order later will have to wait longer. to get delivered.

Today, NVIDIA founder and CEO Jensen Huang personally brought the world's first NVIDIA DGX H200 accelerator card to OPENAI's office in San Francisco, USA, for the delivery process. OPENAI CEO Sam Altman and OPENAI co-founder and president Greg Brockman met with Huang in person to hand over the event, and also posted photos of the scene on X/Twitter.

Jen-Hsun Huang personally went to OPENAI to deliver the world’s first NVIDIA DGX H200 AI acceleration server

It should be reminded that NVIDIA H200 is a GPU launched by NVIDIA suitable for artificial intelligence computing, while DGX H200 is a server that contains multiple H200 GPUs. The DGX series servers are designed and built by NVIDIA themselves, and contain some NVIDIA proprietary technologies. Designed to improve performance by stacking multiple GPUs at the same time.

Compared with H100, H200 has upgraded memory capacity and memory bandwidth. The memory capacity is increased by 1.8 times, and 141GB of memory can be installed. The memory bandwidth is increased by 1.4 times, and the total memory bandwidth is as high as 4.8TB/second.

The price of NVIDIA H200 is rumored to be between US$25,000 and US$40,000. The DGX H200 server can accommodate 4 to 8 H200 accelerator cards. Of course, there is also the premium of DGX itself, so the price of this server should be very exaggerated.

Of course, even if the prices are very exaggerated, these accelerator cards and servers are currently in short supply. NVIDIA previously revealed in its financial report that these accelerator cards will be distributed fairly, but no matter how the distribution is, it may have to be in accordance with the order of the order, so the artificial intelligence companies that order later may still Need to wait a long time.

Copyright statement: Thank you for reading. Unless the source website name or link is indicated in the article, it is the original content of Blue Dot.com.When reprinting, please be sure to indicate: Source: bluedot.com, author andFull link to this article,Thank you for understanding.

What do you think?

Leave a Reply

Your email address will not be published. Required fields are marked *

GIPHY App Key not set. Please check settings

Manual backup using Beyond Compare

Qualcomm is accused of cheating on X Plus/Elite chip benchmark test, actual performance is far from as good as advertised