Exclusive interview with Biren Technology executives: Deconstructing the company’s first 7nm GPU

Recently, BiRen Technology, a domestic high-end GPU chip company, disclosed its first general-purpose GPU chip, BR100, at a product launch conference held recently.

According to Zhang Wen, founder, chairman and CEO of BiRen Technology, this chip built with 7nm technology has set a global computing power record, with a peak computing power that is more than three times that of the flagship products sold by international manufacturers, setting a domestic interconnection bandwidth record. It is also the first general-purpose GPU chip in China to adopt Chiplet technology, the first to adopt the new generation host interface PCIe 5.0, and the first to support the CXL interconnection protocol.

"The official release of BR100 marks the first time that a global general-purpose GPU computing power record has been set by a Chinese company. China's general-purpose GPU chips have officially entered a new era of 'petaflops of calculations per second.'" Zhang Wen emphasized.

Because of its strong technical team and strong capital appeal, BiRen Technology has received great attention from the industry since its establishment three years ago. Now, after more than a thousand days and nights, they have finally brought their highly competitive products. As mentioned above, BiRen Technology also has many innovative designs on this chip.

In order to help everyone understand the chip of Biren Technology and the logic of the team’s product design, Semiconductor Industry Observer reporters recently interviewed many executives of Biren Technology, and together they deconstructed Biren Technology’s first 7nm GPU, and tried to reveal to everyone This GPU upstart’s thoughts on future product planning.

According to BiRen Technology, the BR100 manufactured on TSMC's 7nm process has a chip area of ​​1074mm2 and an integrated number of 77 billion transistors. Such data results are closely related to the company's use of Chiplet and 2.5D CoWoS advanced packaging technology in chip design with a powerful original architecture, while taking into account high yield and high performance. Thanks to this design, the performance of BR100 is comparable to Nvidia's 4nm chip H100 released in 2022. Compared with the latter's 7nm chip A100 released in 2020, BR100 can achieve a three-fold performance improvement.

*Disclaimer: This article is original by the author. The content of the article is the personal opinion of the author. The reprinting by Semiconductor Industry Watch is only to convey a different point of view. It does not mean that Semiconductor Industry Watch agrees or supports the view. If you have any objections, please contact Semiconductor Industry Watch.

Today is the 3128th content shared by "Semiconductor Industry Observation" with you. Welcome to pay attention.