laitimes

Flintstone Technology: "Yun Flint T20" and "Yun Flint i20" acceleration cards will be mass-produced in the first half of 2022

Flintstone Technology: "Yun Flint T20" and "Yun Flint i20" acceleration cards will be mass-produced in the first half of 2022

Jiwei network news, on January 25, Flint yuan technology posted on the WeChat public account that the company's cluster equipped with the first generation of weis chips is landing on a large scale, and the first generation of training and reasoning products have generated more than 100 million orders and revenue, and the second generation of products "Yun Flint T20" and "Yun Flint i20" acceleration cards will be mass-produced in the first half of 2022.

It is understood that "Iris 2.0" is the latest release of the second generation of artificial intelligence training chip by Flintera Technology, its single-precision computing power FP32 reaches 40TFLOPS, and its tensor hash rate TF32 peak reaches 160TFLOPS, which is the first in China. Compared with the first generation of chips, Flintwon Technology has increased the entire computing power by 1.6 times, and at the same time, "Flex 2.0" has also adopted the architecture 2.0 of the new generation of GCU, which is deeply optimized for the characteristics of AI and fully supports various computing accuracies.

With a size of 57.5 mm × 57.5 mm, "Tsai Si 2.0" has become the largest AI single chip in China, and its AI computing power is also the first in China. In the matching with high-performance computing power, in terms of storage and bandwidth, Flintwon Technology is the first chip in China to adopt high-end storage such as HBM2E, and the combination of storage and data and the matching of computing power ultimately ensure that "Iris 2.0" can provide very high computing power, which is 3.5 times higher than the first generation. "Iris 2.0" can support the interconnection of thousands of cards in clusters, achieving excellent acceleration of large models.

Based on the "Yunsi 2.0" chip, the "Yunfu T20" and "Yunfu T21" are aimed at the entire Internet, traditional industries and new infrastructure three major business sectors, and Flint Technology said that they can create such customized solutions for customers based on differentiated scenarios and the whole.

In addition, the first generation of artificial intelligence chip "Yun Flint T11" has been landed on the customer's side of the kilocal-level high-performance cluster, and this landing is characterized by the entire liquid cooling cluster, which plays an important role in creating a green and intelligent computing center project.

Zhang Yalin, founder and COO of Flintewon Technology, said in an interview that since its establishment in April 2018, Flintyuan Technology has created the original intention of "making chips bigger and fighting hard technology". Their entire target market is to do AI training acceleration chips in the cloud. Because that is the "crown" of the entire AI acceleration chip, and it is also the market of the blue ocean. At the same time, he also said that FlintYuan Technology has always adhered to the original and innovative technical route, they pay attention to product landing and ecological construction, pay attention to self-hematopoiesis, and ensure the sustainable development of enterprises. With the support of the state and the care of the government, Flintwon Technology has won the award of "China Core Annual Major Breakthrough Innovation" by the Ministry of Industry and Information Technology for two consecutive years, making Flinthara Technology leading the domestic AI big chip track. (Proofreading | Andy)

Read on