NVIDIA B100: Transforming Data Center Applications

More performance and efficiency for artificial intelligence and high-performance computing with the NVIDIA B100 “Blackwell” GPU built on TSMC’s 3nm technology

Following the conclusion of negotiations between the two companies, NVIDIA will have its next generation of artificial intelligence graphics processing units (AI GPUs) built using TSMC’s 3nm technology. It is anticipated that the B100 “Blackwell” GPUs will become available on the market around the latter half of 2024. These GPUs will provide a substantial increase in performance when compared to the previous generation.

Once again, the graphics processing units (GPUs) of the Blackwell GB100 series feature a chiplet architecture, in which a single chip is constructed from a collection of smaller chips known as chiplets. This offers a variety of benefits, such as increased rates of production, more versatility in chip design, and the opportunity to combine a variety of chiplets, each of which performs a distinct function.

However, because to the more complicated structural design of the Blackwell GB100 GPUs, there is a possibility that the manufacturing costs may be increased. On the other hand, this may wind up having an effect on the price that customers pay for GPUs.

The fourth quarter of 2024 is when NVIDIA’s next-generation B100 GPUs, which will employ TSMC’s 3nm family semiconductors and have a monopoly on more than 80 percent of the market for AI graphics processing units (GPUs), are expected to be released. It will exploit commercial possibilities related to AI deployment while they are still available and reduce competition from AMD, Intel, and other companies.

In any case, NVIDIA has only recently made an announcement stating that the company will proceed with the development of Blackwell AI GPUs. According to a report published by DigiTimes, NVIDIA has already inked a contract with TSMC to build the Blackwell GB100 GPUs using the 3nm manufacturing technology that TSMC has developed.

The 3nm technology makes it possible to achieve better levels of performance and efficiency at a lower overall cost. Despite this, TSMC is now still having trouble manufacturing in the 3nm process. Among them are poor utilisation rates and an absence of cutting-edge industrial technology and facilities. As a consequence of this, it is feasible that the Blackwell GB100 GPUs will not be able to be built in 3 nm until much later in the year 2024. This coincides with the anticipated retail availability date of the GPUs.

This acquisition is more indication that NVIDIA is aiming to further consolidate its leading position in the artificial intelligence (AI) industry. The corporation will be able to further expand its dominance in a variety of fields, including machine learning, computer vision, and natural language processing, with the assistance of the Blackwell GB100 GPUs.

AMD and Intel, two of NVIDIA’s biggest rivals, are also working on AI graphics processing units for the next generation. On the other hand, the Blackwell GB100 GPUs are far more sophisticated than these GPUs.

NVIDIA is presently in a position to be considered the industry leader in the area of AI. The business has projected that sales of AI-based goods would bring in $300 billion by the year 2027.

In order for NVIDIA to accomplish this objective, it is necessary for the company to further demonstrate its advantage over the competitors. As a result, the launch of the Blackwell GB100 GPUs is a significant step in the right direction. NVIDIA will have a huge performance advantage over its rivals as a result of the GPUs, which will enable the business further expand its leading position in artificial intelligence. It’s a very other matter whether or not you’ll like this as a final user of regular graphics cards, however; it depends entirely on you.

News source