Advanced Computing in the Age of AI|Monday, March 1, 2021
  • Subscribe to EnterpriseAI Weekly Updates:  Subscribe by email

China Unveils Its First 7nm Chip 

Source: Tianshu Zhaoxin

Shanghai Tianshu Zhaoxin Semiconductor Co. is claiming China’s first 7-nanometer chip, described as a leading-edge, general-purpose cloud computing chip based on a proprietary GPU architecture.

Dubbed “Big Island” GPGPU, the data center processor is described as a neural network training chip aimed at AI and HPC applications as well as general-purpose computing “at the cloud server level,” Tianshu Zhaoxin said this week.

While touted as China’s first 7nm training chip, the fabless company did not specify where Big Island was fabricated. Chinese chip makers were previously believed to be at least two generations behind leading-edge process technologies. Most 7nm devices are made by Taiwan Semiconductor Manufacturing Co. (NYSE: TSM). TSMC is listed on the Chinese company’s website as a partner, along with leading chip design software vendors such as Synopsys (NASDAQ: SNPS) and Mentor Graphics.

There are other clues to the chip’s manufacturing provenance. Tianshu Zhaoxin said its Big Island chip packs 24 billion transistors on a multichip packaging technology first introduced by TSMC called a 2.5 chip-on-wafer-on-substrate. Among other attributes, the packaging technology significantly boosts memory bandwidth.

By comparison, Nvidia’s A100 Ampere-based GPU houses 54 billion transistors on 826mm2 of silicon, making it the world’s largest 7nm chip.

At top speed, Big Island delivers 147 teraflops of 16-bit floating point performance. It supports multi-precision, mixed-data training and integrates high-speed interconnections. The new GPU also supports other numerical formats, including FP32, 32-bit integer and bfloat16.

The data center GPU is also aimed at “solv[ing] the core computer power computing problem,” said the Chinese chip maker, an assertion that indicates the company expects to compete against market leader Nvidia (NASDAQ: NVDA), AMD (NASDAQ: AMD) and others.

Big Island “can complete the artificial intelligence processing of hundreds of camera video channels per second, and the performance is twice that of mainstream products in the market,” it further claimed.

Development of the 7nm GPGPU chip commenced in 2018. According to reports, the device was first taped out in May 2020, and the chip maker is expected to ramp volume production sometime this year. Along with AI-based video processing, the company said Big Island would be aimed at applications ranging from autonomous driving to medical research.

While not specifying general availability, Tianshu Zhaoxin said it expects to “make steady progress in the whole product cycle, and make solid and steady preparations for large-scale production and market launch of future products.”

The fabless chip maker was launched in 2013 as a joint venture between the Shanghai municipal government and Taiwan’s VIA Technologies (TPE: 2388).

About the author: George Leopold

George Leopold has written about science and technology for more than 30 years, focusing on electronics and aerospace technology. He previously served as executive editor of Electronic Engineering Times. Leopold is the author of "Calculated Risk: The Supersonic Life and Times of Gus Grissom" (Purdue University Press, 2016).

Add a Comment

EnterpriseAI