Nvidia s AI chip B200 is expected to be released in 2025, and the power consumption soars to 1000W

Mondo Technology Updated on 2024-03-05

NVIDIA's upcoming next-generation artificial intelligence (AI) chip, B200**, has attracted widespread attention in the industry. According to Dell, the B200 will consume up to 1000W and is expected to be released in 2025. This news means that Nvidia is moving towards higher performance and more powerful AI computing, however, it also raises concerns about energy consumption.

Nvidia's next-generation AI chip, the B200, codenamed Blackwell, will be the latest addition to its B series. Compared with the current H100 products, the B200 is not only more powerful in computing performance, but also brings amazing power consumption, which is expected to reach up to 1000W, an increase of more than 40% compared with the H100. It is said that the B200 will use the Hopper architecture and be equipped with HBM3E high-bandwidth memory, which is considered to be one of the most powerful AI computing chips in the industry.

To accommodate the high power consumption of the B200 chip, Jeff Clarke, chief operating officer of Dell, said that Dell will offer a flagship PowerEdge Xe9680 rack server designed for B200 server GPUs. It also means that Dell is going all out to cool these power-hungry GPUs through its innovative engineering technology.

Compared with the previous H100 product, the B200 not only has a significant improvement in computing performance, but also shows higher power consumption. Currently, Nvidia's H100 consumes 700W overall, while the B200 will consume up to 1000W, an increase of more than 40%. This shows that while NVIDIA is pursuing higher performance, it is also facing the challenge of energy efficiency.

Dell's revelation shows that the B200 will not be officially released until 2025, which also means that Nvidia is working to improve its technology and showcase its engineering innovation in liquid cooling systems. It is expected that the B200 chip will be NVIDIA's first AI GPU with a dual-chip design to handle more heat generated. Considering that AMD and Intel have adopted GPU architectures with multi-chip designs, this is also in line with the development trend of the industry.

With the rapid development of artificial intelligence technology, the surge in demand for chips will inevitably bring a huge demand for power resources. AI currently accounts for a small percentage of global data center electricity consumption, but this share is expected to increase rapidly to 10% by 2025. This poses a challenge for power** but also brings new opportunities for the energy industry. Some analysts believe that the large demand for large-scale AI servers will promote the growth of investment in the energy fields such as power and oil and gas, and even attract attention to nuclear energy.

Disclaimer: The above content is compiled from the Internet and is only for communication and learning. If you have any content or copyright issues, please leave a message and contact us to delete it.

Related Pages