Shanghai Tianshu Zhixin Semiconductor Company announced that its flagship 7nm process GPGPU (General Parallel) cloud computing chip "BI" has been successfully lit recently.
This is the first fully self-developed 7nm GPGPU training chip based on GPU architecture in China. After mass production, it will be widely used in AI training, high-performance computing (HPC) and other scenarios, serving education, Internet, finance, and autonomous driving. , Medical, security and other industries.
It is reported that the BI chip started to research and development in 2018, will be taped out in May 2020, returned in November, and successfully "lit" in December. In the past month or so, the technical team has conducted a series of hardware and software tests. Covers nearly a hundred indicators to verify that the actual functions of BI meet design standards.
Compared with the existing mainstream products in the market, the Tianshu Zhixin BI chip can provide flexible programming ability, stronger performance, attractive cost performance, and is a safe domestic chip solution.
It uses 7nm process technology, 2.5D CoWoS packaging technology, integrates 24 billion transistors, supports FP32/16, BF16, INT32/16/8 and other multi-precision data mixed training, supports high-speed inter-chip interconnection, and single-core can perform 1.47 million per second One hundred million times of FP16 floating-point calculations, it 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.
Headquartered in Zhangjiang, Shanghai, Tianshu Zhixin is the first hard technology company in China to focus on high-performance computing systems for GPGPU chips. It is mainly engaged in the research and development of high-end general-purpose parallel computing chips at the cloud server level, aiming at cloud computing, artificial intelligence, and digital transformation. Represents the data-driven technology market.
0 Comments