The new Nvidia chip comes to enhance AI even more
NVIDIA has become a gigantic company, in fact, it is valued at billions of dollars -yes, with a B-, it has achieved this status thanks to the most sought-after artificial intelligence chip on the market: the H100. This technology has not only established its position as the undisputed leader in the sector, but has also managed to surpass industry titans such as Alphabet (Google’s parent company) or Amazon. Now they have presented what is considered the largest AI chip in the world by a company that today is intractable in this field.
NVIDIA goes on the attack
NVIDIA, which could still exploit the H100 chips, has decided jump into the pool and expand your horizon even further technological by launching two new chips designed specifically for artificial intelligence tasks: the Blackwell B200 and the impressive GB200 superchip.
The B200 chip features 200 billion transistors, capable of delivering up to 20 petaflops of FP4 power. On the other hand, the GB200, there is no doubt that it is the current masterpiece in terms of AI technology, integrates two B200 GPUs and a single Grace CPU, managing to increase performance by inference of large-scale language models by up to thirty times. This advancement means that training a model with 1.8 billion data points has passed from requiring 8,000 GPUs to only 2,000, marking a milestone in efficiency and capacity.
Given all this, the launch of a second generation transformer motor that doubles the computing capacity, bandwidth and size of the models. Furthermore, the new NVLink switch The latest generation facilitates communication between up to 576 GPUs, reaching a bidirectional bandwidth of 1.8 terabytes per second.
With all this, it is not surprising that NVIDIA is currently one of the most important companies in the world thanks to its development of AI chips. In fact, the company uses AI itself to boost its performance.
To summarize:
- NVIDIA is a company valued in trillions of dollars thanks to the most sought-after AI chip on the market, the H100.
- This has made it one of the most powerful companies on the market, surpassing giants like Alphabet, but it has not stopped there.
- The company has decided to expand its catalog with the launch of two new AI-oriented chips, the Blackwell B200 and the superchip GB200.
- The B200 has 200 billion transistors and offers up to 20 petaflops of FP4 power.
- The GB200 combines two B200 GPUs with a single Grace CPU, increasing performance for large-scale language model inference by up to 30x.
- Training a model with 1.8 trillion data now requires 2,000 GPUs, up from 8,000 in the past.
- NVIDIA has introduced novel technologies such as the arrival of a second-generation transformer engine, doubling the computing power, bandwidth and model size.
- To all this we must add a new next-generation NVLink switch that allows 576 GPUs to communicate with each other with a bidirectional bandwidth of 1.8 terabytes per second.
To always be up to date with the latest in technology, subscribe to our official and verified Andro4all channel on WhatsApp.