The market value of NVIDIA has exceeded the sum of Intel, AMD and domestic GPU. Where does NVIDIA establish itself?


Every reporter Liang Xiao Zhu Chengxiang
Original title: NVIDIA’s new graphics card performance leaps, market value has surpassed the sum of Intel, AMD and domestic GPU?
Huang Renxun, CEO of NVIDIA, still appeared in his familiar black leather suit, but this time, instead of “squeezing toothpaste”, he presented a cost-effective “killer’s mace”.
Last week, NVIDIA launched a new generation of consumer grade GPU products, geforce RTX 30 series, which uses ampere architecture and Samsung 8 nanometer process. Its performance is far better than that of the previous generation geforce RTX 20 series with Turing architecture and TSMC 12 nm process, but the price has not increased significantly.
Since the beginning of this year, NVIDIA’s share price has reached new highs, and the company’s market value has exceeded $300 billion, surpassing the chip giant Intel. Moreover, it has left amd, an old rival in the field of GPU (graphics processing unit), far behind.
At present, the GPU field of PC is monopolized by Intel, NVIDIA and AMD. What is the strength of domestic GPU? Some industry insiders told the daily economic news that the development level of domestic GPU still lags behind that of international large manufacturers, but “the performance of game computing can’t do, which does not mean that the HD decoding function can’t be done”. These companies can still compete with them in differentiation.
Photo source: video capture of NVIDIA official press conference
Share price of new products reaches a new high
“The 3080 has twice the performance of the 2080, but at the same price. Ampere has achieved the biggest performance leap ever. ” At a new product launch in his kitchen, Huang said.
GPU is the core of computer graphics card, which undertakes the task of image processing and output display.
In mid August this year, NVIDIA surpassed Intel in market value and became the third largest chip company in the world (TSMC and Samsung were the top two). As of the end of September 4 local time, NVIDIA’s market value reached US $311.5 billion, more than the sum of Intel and AMD.
Chen Qi, the semiconductor investment manager of pinley fund, said: “3080 claims to have integrated 28 billion transistors. With its powerful performance, it can ensure that some game effects can be fully opened and the speed of 60 frames can be run smoothly under 8K HD screen.”
By Liang Xiao
Chen Qi also introduced NVIDIA’s ray tracing technology to reporters: “the ray tracing technology was first proposed by NVIDIA, and first appeared on 10 series graphics cards with Pascal architecture. It can make the game screen to achieve the degree of authenticity, giving players a stronger sense of reality
“Ray tracing, once the Holy Grail of computer graphics, is now the standard.” At the launch of the new product, Huang Renxun said that the ray tracing performance of ampere architecture is twice that of Pascal architecture. Fortress night, call of Duty: the cold war, cyberpunk: 2077 will all support the next generation of ray tracing technology.
Chen Qi believes that in the future, the performance of NVIDIA’s new GPU will be further improved. “Because TSMC’s N7 + process, that is, 7Nm EUV, is full of orders, NVIDIA chose Samsung’s 8nm process. The 20 series of Turing architecture uses TSMC 12NM process, and the transistor density is about 33.8mtr/mm; this time, the 30 series of ampere architecture uses Samsung 8nm process, with transistor density of 61.2mtr/mm, almost doubling. If TSMC 5 nm process is used in the future, the transistor density will exceed 100 MTR / mm, and it will nearly double, which will inevitably bring about a doubling performance improvement. ” Chen Qi said.
The RTX 30 series is not the first NVIDIA product to adopt the ampere architecture. On May 14, this year, the first gpera of Acer was announced.
Domestic GPU: there is a gap, but differentiated competition can still be carried out
In recent years, GPU has become a new force in semiconductor industry, and the development of domestic GPU has also attracted much attention. According to the estimation of Soochow securities, the market space of domestic substitution in GPU field is more than 5 billion US dollars.
However, there are few people who can share such a big cake. Hangjin Technology (000818, SZ) once said in the record sheet of investor relations activities that domestic GPU chip design enterprises also include jingjiawei, etc.
In other words, in the A-share market, only Hangjin technology and jingjiawei (300474, SZ) are the relevant targets of GPU design.
The original main business of Hangjin technology was the production and sale of caustic soda and other chemical raw materials. After 2017, the listed companies successively acquired Changsha Shaoguang Semiconductor Co., Ltd. (hereinafter referred to as Changsha Shaoguang) and other companies, officially entering the field of military electronics. Among them, Changsha Shaoguang is the most eye-catching, and is one of the few GPU manufacturers in China.
At present, the application of the second generation improved graphics processing chip independently developed by Changsha Shaoguang has been verified in the field of self controllable equipment, and relevant orders have been obtained. The third generation products are expected to be launched by the end of 2021. In March this year, Hangjin technology said that Changsha Shaoguang “benchmarked with the international chip giant NVIDIA”. However, in terms of volume alone, there is no comparability between the two. This statement has also aroused inquiries from the Shenzhen Stock Exchange.
Jingjiawei, with the halo of “the first domestic GPU stock”, landed on the gem at the end of March 2016. In the semi annual report of 2020, the company said that the second generation of graphics processing chip jm7200 was successfully taped in August 2018. Compared with the first generation of graphics processing chip jm5400 in product performance and process design, the company has completed the adaptation work with major domestic CPU and operating system manufacturers. On September 1, jingjiawei disclosed that the performance of the company’s next-generation graphics processing chip will be greatly improved compared with the jm7200, but the specific performance indicators will not be determined until the completion of the tapeing and testing work.
By Liang Xiao
According to Yao Jiayang, a consulting analyst at Jibang, judging from the current product specifications and product blueprints of domestic GPU, there is indeed a big gap with NVIDIA and AMD. On the whole, it is still difficult to catch up in the short term.

Yao Jiayang believes that in the low-level graphics card, for the game field, domestic GPU also needs to have more in-depth cooperation with important game players. However, he also said that in addition to the existing game market, industry, medical, military aerospace and AI have further development space. In the field of industrial control and military, stability and strength are more important, so performance is not the focus of pursuit. From the perspective of domestic substitution strategy, domestic GPU can gradually gain market share in low-level industrial or military markets.
Chen Qi holds a similar view. He told reporters that GPU contains multiple functions, but the performance of game computing doesn’t mean that HD decoding can be done. However, in the face of different application scenarios and consumer needs, there are still opportunities.
The views of Yao Jiayang and Chen Qi are quite similar to the development ideas of the two listed companies. Jingjiawei said in the 2020 semi annual report that the company has carried out technology and product development according to the application needs of different industries, occupying an obvious dominant position in the application of special market. Meanwhile, the company’s graphic display and control products have been actively extended to other fields in recent years.
Hangjin technology, on the other hand, said in the record sheet of investor relations activities that Changsha Shaoguang’s GPU products can be used not only in military equipment, but also in computer systems and office equipment. From the use effect tracking, the product has been able to meet the needs of autonomous and controllable office system and server.
The reporter noted that the national integrated circuit industry investment fund ranks the second largest shareholder of jingjiawei (holding 275 million shares), but does not hold shares in Hangjin technology. Recently, on the interactive platform, some investors also inquired about the cooperation between Hangjin technology and the national large fund. Hangjin technology replied that the large fund needs to cooperate with the company in the form of fixed increase, and the company has no fixed increase plan for the time being.
GPU: from game to AI, autopilot
“Modern GPU is the engine for the development of design, cloud AI, scientific computing, etc., but it is the gamers and their endless needs that drive GPU forward.” Huang Renxun said at the press conference.
Dongguan securities listed three major directions of GPU application in the future — games, artificial intelligence and deep learning, automatic driving. GPU has taken root in the field of games, and achieved great results in the field of deep learning, thus promoting the development of artificial intelligence and automatic driving.
Game is the origin of GPU, and ray tracing technology pushes the visual experience of game to the extreme. In addition, the future notebook will develop towards the direction of stronger performance and thinner thickness. The trend of thinness and thinness of game books will become the new power point of GPU in the game field. However, the rise of cloud games may divert the demand for high-end game graphics cards to a certain extent.
GPU is also an important part of artificial intelligence. Dongguan Securities pointed out that the three major elements of artificial intelligence are data, computing power and algorithm. The essence of deep learning is that through the processing of big data and the established algorithm model, AI applications in various industries can be realized. The algorithm connects data and computing power together to provide efficiency optimization schemes for different subdivision scenarios. Deep learning algorithm needs a large number of simple operations, and has high requirements for parallel computing. In this respect, GPU has a strong advantage.
The Research Report of Soochow securities also believes that deep learning algorithms usually need massive computing to process data, extract data object features and conduct repeated training. The traditional CPU cluster needs several weeks to calculate the neural network cascade with 100 million nodes, while a GPU cluster can be completed in one day, with obvious speed advantages.
The ability of GPU to process instantaneous mass data is also applied in the field of automatic driving. In December 2019, NVIDIA released an autopilot vehicle platform, DRIVE AGX Orin, which was built with an automatic driving chip Orin. At the same time, Didi has reached a partnership with NVIDIA, which will use NVIDIA GPU and other technologies to develop autonomous driving and cloud computing solutions.