Nvidia Hopper GH100 GPU could have 140 billion transistors, 160% more than GA100

Unlike previous years where we’ve seen a single architecture used for both desktop and servers, this year we’ll see two new architectures from Nvidia’s side: Ada Lovelace for Gaming, and Hopper for Servers. New information about this latest architecture reveals that its largest chip, GH100, could have a huge increase in transistors compared to GA100, the largest Ampere-based chip, being extremely more powerful.

According to a new rumor published in Chiphell, GH100 would have about 140 billion transistors, 160% more than the 54 billion that GA100 has. According to previous rumors, these would be placed in a huge DIE of almost 1000mm², manufactured under TSMC’s 5nm process, compared to the 826mm² DIE manufactured under TSMC’s 7nm GA100 process.

In this way, GH100 will have an extremely higher raw power than any current chip, being a huge improvement to accelerate deep learning, machine learning, and other AI-related tasks that require a lot of computing power.

It will be interesting to see what new technologies Nvidia implements in these next-generation cards to stay ahead of the game after the release of new GPUs from AMD and Intel, which will also have huge computing capabilities according to what both companies anticipate. We may not have official news until the second half of the year, although data will continue to leak that can give us an idea of ​​what to expect.

What do you think about the incredible size of the Hopper-based Nvidia GH100 GPU?

Source: WCCFTech