Science Technology

The true face of “roll king” NVIDIA

After the GTC 2022, the masses of the people rushed to tell each other, and Huang Renxun came to “blast the street” with his new “nuclear bomb”.
Many articles have introduced the new products and technologies that have been released in detail. In a word, cow! Among them, the H100 GPU is the most explosive. It uses TSMC’s 4-nanometer process technology and integrates 180 billion transistors. Its floating point computing power is three times faster than that of the previous generation A100. It is regarded as NVIDIA’s next-generation “nuclear bomb”.
For a time, drums and firecrackers were blazing, and the industry was expecting the “computing monster”, and consumers were also ushered in the day of video card price reduction.
But calm down and think about it. Is Nvidia a great philanthropist who “keeps calculating power for AI and gives warmth to everyone”? Both fans and spectators must admit that Nvidia is a commercial wonder. As the most legendary digital economy stock, its revenue is far lower than that of Intel or Meta, but its market value is far ahead, which is clearly not explained by “technology belief”.
In Huang Renxun’s own words, NVIDIA has launched several landmark key technologies in its history. In fact, it has “generalized” the development achievements of its GPU technology and found that it can do more different things.
The true face of “roll king” NVIDIA
Whether it is the new demand that drives the new products or the new products that activate the new demand is a question of “chicken begets egg begets chicken”. However, we can summarize a template that has never failed from the correlation between Nvidia’s actions and results:
When the mainstream demand has not reached the peak, expand and enrich the product line, even if “wearing waistcoat”, also need to fully layout, and exploit the niche market to the maximum value; Once the mainstream demand declines and ebbs, the nuclear-bomb products that exceed the current market demand will be thrown out (of course, the power consumption also explodes), which will ignite the imagination of the public and Wall Street on GPU again, and further push the valuation of NVIDIA to a new high.
Compared with the “King of Roll King”, Nvidia is more like Jiang Taigong, who sits firmly on the fishing platform, holding the global consumers and industry in a clear position.
Four “rebounds” of supply and demand, Nvidia’s path to computing supremacy
How did Nvidia achieve its supremacy in computing power? Counting several milestones in history, we will find four key “rebounds”.
The first rebound: the growth storm of personal computers.
During the period from its establishment in 1993 to 1999, NVIDIA did not occupy a leading position in the crowded graphics card market.
At that time, there were many manufacturers developing display chips. In addition to IBM, Sony, Toshiba and other semiconductor giants, vertical circuits such as Matrox, 3dfx, Trident, and S3 Graphics had led the way. NV1 and NV2 released by Avida are not competitive and almost bankrupt. Although the TNT2 (also known as “NV5”) released in 1999 won the performance crown, its speed was only 10% to 17% higher than that of NV4, and there was no gap between it and its main competitor 3dfx Vodoo3.
So the first “nuclear bomb” came, and Nvidia launched NV10, namely GeForce 256, the first professional graphics processing core, which directly opened the accelerated market of personal computer games.
The true face of “roll king” NVIDIA
Before that, GPU display chips were all chips with fixed functions, and the emergence of GeForce 256 became the first “single chip processor integrated with conversion, lighting, triangle setting/clipping and rendering engines”, capable of processing at least 10 million polygons per second, enabling GPU to take over a large number of geometric operations from the CPU, solving problems that cannot be solved by general computing, and greatly promoting PC games Creative design and other requirements for GPU.
In order to give full play to the computing potential of GeForce 256, NVIDIA also launched the Quadro framework based on the chip, serving professional drawing workstations, to help creative and technical personnel work more efficiently. Later, programmable shaders were introduced to allow developers to play more creative roles on GPU, such as 3D rendering, game development, special effects production, etc
In the words of Huang Renxun, it is to encourage or mobilize the passion of the global people, let them understand what 3D graphics processors are, and provide them with many tools for innovation.
The true face of “roll king” NVIDIA
The second year after the release of GeForce 256, Nvidia received an order from Microsoft to develop graphics cards for Xbox video game consoles. After that, it relied on the market operation of “half a year updating and one year replacing”. The GeForce series product line has been continuously enriched and comprehensively distributed, covering all kinds of high, middle and low end markets, and has also learned to wear “vest”. It has been slightly improved and improved on the basis of the original chip, and it has been rapidly launched into the market as a new series. As a result, NVIDIA has occupied more than 70% of the GPU market.
By 2007, Nvidia’s market value had risen by more than 500%, and was named the company of the year by Forbes magazine.
The second rebound: the strong thrust of parallel computing.
As early as 2006, Nvidia launched the revolutionary universal computing architecture CUDA and the universal computing hardware Tesla GPU. At that time, however, deep learning was not as popular as it is now. Only some large enterprises and research institutions needed GPU to carry out high-performance computing tasks such as drug invention, weather modeling, financial analysis, etc.
When did Nvidia begin to increase its efforts to activate the demand for GPU parallel computing capability? The answer is 2009.
In this year, Nvidia held the first “GPU Technical Conference”, which preached to “developers, engineers and researchers who use GPU to solve important computing work”.
The true face of “roll king” NVIDIA
What changes have taken place in the market between 2006 and 2009? Under the rule of Moore’s Law, individual consumers have become tired of the performance requirements of computer graphics cards.
During this period, NVIDIA also released some good products, such as the heavyweight Tegra mobile processor, which integrates ARM architecture processor and Geforce GPU, and its power consumption is 30 times lower than that of ordinary PC laptops. Although the product is good, it is difficult to arouse the enthusiasm of consumers. After all, there are so many video cards on the market. As long as you are willing to wait, you can start with a more fragrant price.
At the same time, some GPU defects integrated into Apple, Dell and HP notebooks by OEMs have led to “abnormal failure rate” and become the object of class action. In the first quarter of 2008 alone, Nvidia’s revenue decreased by about $200 million. The share price also fell from $37 to about $6.
The true face of “roll king” NVIDIA
So Nvidia began to expand the layout of high-performance computing. At the first GPU technical conference, Nvidia launched the next-generation CUDA GPU architecture codenamed “Fermi” Fermi, and vigorously publicized the advantages of GPU in large-scale parallel computing tasks.
Fermi architecture is competent as a “nuclear bomb”. On the one hand, its performance is very high. Geforce 4 series products based on this architecture have successfully suppressed competitors in performance, but the power consumption and heat generation of this architecture are also very frightening.
The true face of “roll king” NVIDIA
In any case, NVIDIA has been widely popular in the field of computing since then. In 2010, the world’s fastest supercomputer Tianhe-1A adopted 7168 NVIDIA’s Tesla M2050 GPUs, combining large-scale parallel GPUs with multi-core CPUs, and became the representative of heterogeneous computing at that time.
In 2012, Geoffrey Hinton, one of the three giants of deep learning, and Alex, his student, used GPU to speed up the training of deep neural networks. They made a great success in the ImageNet competition, opened the curtain of the third wave of AI, and further boosted the sales of Nvidia GPU.
The true face of “roll king” NVIDIA
(Geoffrey Hinton and Alex Krizhevsky, IIya Sutskever)
The growth of AI demand also helped Nvidia expand the automotive market. The Geforce GTX Titan Titan released in 2013 represented the top level of Kepler architecture, became the computing foundation of autonomous vehicle and advanced driving assistance systems, and supported key computer vision functions.
The strong demand of AI from academia to industry has driven the price of GPU and the share price of NVIDIA to rise, and completed a stunning “rebound” in the world.