NVIDIA plans to accelerate the release of Blackwell GPU to meet the growing demand for AI and enter the supply chain certification stage

At present, AI has become a big hit in the technology world. Almost every manufacturer hopes to get involved with AI to attract a wave of AI traffic. To achieve better AI performance, it is necessary to have powerful computing power. Power is essential, especially for AI training, and GPU can be said to be an excellent choice for AI training. Major manufacturers have also purchased GPUs to engage in AI training. Nvidia should be considered the leader in GPUs, especially its Tesla computing cards, which are in short supply. Now the company plans to work overtime to develop the next generation of GPUs, hoping to continue to expand its advantages in the AI ​​field and get a bigger share of the pie. .

According to the supply chain, Nvidia currently plans to significantly advance the development and release schedule of the next-generation GPU, the Blackwell B100 chip. It was originally planned to be released in the fourth quarter of 2024, but it has now been advanced to the second quarter. In addition, NVIDIA has also begun to enter the stage of supply chain certification, which means NVIDIA will look for a suitable supply chain to engage in the manufacturing of NVIDIA B100 computing cards.

According to current news, for the B100 computing card, the competition for the base supply chain is already very fierce. Two well-known foundries including Foxconn and Wistron hope to obtain this large order from Nvidia, and through the layer After layer of inspection, Wistron finally won the order because their yield rate was higher and it seemed to be more in line with Nvidia's requirements. In the field of video memory, Hynix won this order from Nvidia to provide the AI ​​giant with ultra-high bandwidth HBM3e video memory.

There is currently very little news about the B100 computing card, but considering the current craze of artificial intelligence, the B100 computing card is expected to have better artificial intelligence computing power than the existing H100 computing card, and may have several times the performance. promote. In addition, with the help of more advanced manufacturing technology, the B100 chip may become the terminal product with the largest number of transistors in NVIDIA's history

Due to years of layout, NVIDIA seems to have no rivals in the field of AI computing power. Basically, the top computing power chips have been rounded up by old people. And in such a market environment, NVIDIA still does not squeeze out toothpaste, but By launching the next-generation GPU ahead of schedule, Nvidia is obviously still pursuing technological innovation. For gamers, the early release of the B100 should be good news. After all, game chips and computing chips are similar in many units. Maybe we can launch it in 2024. You can see the arrival of a new generation of gaming graphics cards in half a year or at the end of the year.

The above is the detailed content of NVIDIA plans to accelerate the release of Blackwell GPU to meet the growing demand for AI and enter the supply chain certification stage. For more information, please follow other related articles on the PHP Chinese website!