(Electronic Enthusiast Network report article / Zhang Ying) “GPU can lead the computing power era of the Internet of Everything, from natural language processing, machine vision and even to biomedicine, business finance, AI is now everywhere. In the era of intelligent manufacturing of digital twins, GPU can achieve the ultimate in graphics rendering, not only can be visually realistic, but even the real world can be affected by the virtual world. It can promote urban intelligence, industrial automation and unmanned vehicles. GPU can help you unlock the mysteries of nature, We simulate trillions of atoms and molecules, which can develop new drugs and new materials.” Dr. Lv Jian, Chief Technology Officer of Shanghai Tianshu Zhixin Semiconductor, expressed his forward-looking views on AI changing the world to the media.
On July 13, 2022, the company announced that Shanghai Tianshu Zhixin Semiconductor Co., Ltd. announced the completion of more than 1 billion yuan in C rounds and C++ rounds of financing. The C+ round of financing was led by Financial Street Capital, and the C++ round was led by Hopu Investment and its subsidiary Houan Innovation Fund. Zhongguancun Science City Science and Technology Growth Fund, Shanghai Guosheng, Xicheng Zhiyuan, Emerging Assets, Dingxiang Capital, Dingli Capital, Guangdong-Hong Kong-Macao Industry and Finance, Shanghai Free Trade Zone Equity Fund and other well-known enterprises and institutions participated in the investment. This financing event has become the largest investment event in the second week of July.
Why is Tianshu Zhixin favored by many investment institutions? Where is the latest development and mass production of their GPGPU products?
GPU market: two major international manufacturers occupy 80% of the market, and domestic GPU manufacturers have a breakthrough point
According to Verified Market Research data, the independent GPU market in mainland China in 2020 will be US$4.739 billion, and it is expected that the GPU market in mainland China will exceed US$34.557 billion in 2027. In the era of big data, all walks of life need stronger computing power, so the market demand for GPGPU (general-purpose GPU) has increased significantly.
At present, the two giants of Nvidia and AMD in the United States occupy more than 80% of the global general-purpose GPU market. How can local manufacturers break this monopoly in the Chinese market? A group of players such as Moore Thread, Tianshu Zhixin, Biren Technology, and Muxi Integrated Circuit are aiming at the GPU track to compete with NVIDIA.
“The basic development cost of a large GPU chip requires 600 million yuan, involving a large number of engineers.” Senior experts in the chip industry revealed. Shanghai Tianshu Zhixin Semiconductor Co., Ltd. has achieved breakthroughs in related fields. In March 2021, Shanghai Tianshu Zhixin released the Tiangai 100 chip.
Tianshu Zhixin official micro shows that Tiangai 100 is the first fully self-developed general-purpose GPU under the GPU architecture of 7nm process in China.
The Tiangai 100 launched by Tianshu Zhixin broke the long-term situation of domestic enterprises as foreign IP agents. Its fully self-developed architecture, computing core, instruction set and basic software stack are not only free from foreign constraints, but also can respond quickly. Rapidly changing market demands. Secondly, in terms of application, Tiangai 100 benefits from the completely self-developed chip architecture and software stack, which can realize multi-dimensional technological innovation, and has many characteristics such as independent controllability, high performance, versatility and flexibility.
Tiangai 100 has made Tianshu Zhixin the only hard technology enterprise in China that has achieved mass production of general-purpose GPUs, and its product development and commercial applications are at least 1 year ahead of its domestic counterparts. This general-purpose GPU has won orders from customers in security, artificial intelligence, supercomputing centers and other fields.
In addition, Moore Thread released the first-generation all-round GPU “Suti” not long ago, and launched the GPU graphics card MTT S60 and MTT S2000, achieving the first breakthrough in the domestic graphics card field.
Tianshu Zhixin GPU products continue to evolve, corresponding to the needs of new scenarios such as Metaverse and Digital Twins
Why can GPU break through traditional graphics rendering and enter other fields of application? Dr. Lv Jian, Chief Technology Officer of Tianshu Zhixin Semiconductor, pointed out that GPUs are composed of a large number of parallel processing layers in three major fields: AI inference training, graphics rendering and scientific computing. The GPU itself is characterized by a basic comprehensive architecture with single-instruction and multi-threading, and the applications of the three major computing in the future are perfectly matched with the GPU. GPUs are especially suitable for many areas of increased computing, which is why GPUs have far more cores than CPUs.
GPU architects hope that graphics processing can be programmed, and every graphics application developer hopes to win in terms of effects, so this direction is particularly important. On the graphics pipeline, the architect identifies certain nodes that he can implement in a programmable way using general-purpose computing. Since there are general computing pools, why not find new sources of applications outside of graphics? The general purpose GPU was born.
The general-purpose GPU has two meanings: first, it does not use hardware for graphics professionals, and directly uses its general-purpose computing for applications other than graphics; second, Tianshu Zhixin developed GPU hardware, omitting graphics-specific hardware, only Keep the general part. The term GPGPU-specific appears.
Dr. Lv Jianping emphasized that AI needs to continue to innovate general-purpose GPUs. He said that the general-purpose GPU has been selected as the final winner in the AI market. As new scenarios such as metaverse and digital twins put forward new requirements for cloud integration of AI and graphics, general-purpose GPUs also need to continue to innovate and progress.
Dr. Lv Jianping proposed an innovative path: DSA generalization, which can sustain general advantages; graphics computing, which can bridge advanced graphics; computing graphics, which can improve computing performance; hardware differentiation, which can promote AI graphics integration.
He emphasized that Tianshu Zhixin will follow the trend of graphics to general computing, take the successful mass-produced general-purpose GPU product – Tiangai 100 as a solid foundation, adhere to the road of AI and graphics integration innovation, and lead the development trend of general-purpose GPUs in China.
Figure: Zhijia 100 chip from Tianshu Zhixin WeChat
By the end of March 2022, Tiangai 100 products have achieved sales orders of nearly 200 million yuan, assisting customers to implement more than 200 application scenarios. In May 2022, the second product of Tianshu Zhixin, the 7nm cloud-edge reasoning chip “Intelligent Armor 100”, was successfully lit up, and the product iterative development and commercial application were ahead of domestic counterparts. Dr. Lv Jianping also said that, based on the successful foundation of Tiangai 100, we will continue to improve product performance, develop Tiangai series follow-up products that catch up with the international market and match national conditions, Tiangai 200 and Tiangao 300.
Why win the favor of so many capitals? Lin Sihan, head and managing director of Houan Innovation Fund, said: “GPU is a key element to promote the development of computing power. Tianshu Zhixin has a very deep technical accumulation in this field and has successfully achieved commercialization. We look forward to Tianshu Zhixin. It can empower industrial development with computing power and promote the intelligent transformation of all walks of life.