I have finished reading "The Way of Nvidia" and "The Core of Nvidia" in the past two weeks. Big model artificial intelligence is so popular today. In addition to the persistence of neural network scientists such as Hinton, what is more important is the development of computing power chips represented by Nvidia, a technology company established in 1993, from video games. How did it start and how did it continue to innovate, and what is the history of GPU development? These are my original intentions for reading these two books. Focus on writing the reading notes of these two books.
1. Why start a business
Nvidia has three founders, Prim, Malakovsky and Huang Renxun. The first two originally designed graphics accelerators at Sun, while Huang Renxun was responsible for assisting in the manufacturing of chips at LSI. In 1993, due to political struggles within Sun's company, Prim and Malakowski began to consider starting a business. Then they called Huang Renxun, who had worked with before. Prim was a technical geek, and Huang Renxun could handle the technology. Other things outside.
At that time, Huang Renxun had a stable management position in LSI, and the founder of LSI valued him very much, so he let him manage the department with an annual income of US$250 million. The details revealed in "The Way of Nvidia" are that Prim has no family pressure and is very technical geek, and he thought about it earlier, while the other two founders are a little hesitant, especially Huang Renxun. If Huang hadn't had management frictions in LSI, he would probably have continued to stay in LSI. The book does not give a detailed introduction to what friction he encountered when managing his subordinates, and the word "friction" is even a bit understatement. Some clues can be seen from Huang Renxun's style after becoming CEO. Huang Renxun is simple and direct, leading by example, requires his subordinates to work hard and work hard, and be very persistent in doing things. He is not the boss, so he cannot shape the company as a working person He could not allow his subordinates to earn more income from his struggle culture, and these subordinates may have been following their boss for longer than Huang Renxun, and they may not have established a sense of trust with Huang Renxun. The subordinates would be resistant to Huang Renxun's work culture. In order to resolve the friction, the boss poached a senior manager from outside and Huang Renxun to jointly manage the department. Huang Renxun was very dissatisfied and felt that it was a political operation that prompted him to leave.
In general, there were opportunities for graphics chips at that time, and because they encountered difficulties in their own upward development, they chose to start a business. If Prim and Malakovsky had not started their business, if it weren't for the hindrance of the LSI's rise, Huang Renxun would probably have chosen to co-found Nvidia.
Starting a business requires meeting three conditions: things, people, and money. Before starting a business, Prim already had a graphics chip design plan, and several founders were in place, but they were still short of money. Their entrepreneurship was supported by LSI boss Corrigan, who introduced Sequoia Capital to them, but investors were not optimistic about their projects at that time. Plum was too out of touch with the market, and Huang Renxun had not yet grown up at that time, so it was impossible to say that Know where the market is, but investors believe that there is a market here and believe in Corrigan's vision, so they invested in them. According to this, Huang Renxun gained an experience: even if your business plan writing skills are not good, your reputation will be Take one step ahead and be known to others. This is also reflected in attracting engineers. Several senior engineers from Sun Company joined in the early stages of their business, which shows that Prim also has a good reputation and that Sun Company cannot retain talents.
2. Tech geek Prim leaves the company
Generally speaking, conflicts among founders can easily lead to civil strife in the team, but this problem does not exist at Nvidia. When Prim left, he only managed a small team of several people. Prim is a very skilled person, but he is not. A person who is good at teamwork and leads the team to success.
The book tells the story of Prim's rough view of code as his personal belongings at a certain stage, regardless of team collaboration and R&D processes, which led to a gradual decline in his management rank, which triggered his resignation (1998). However, other contents of this book have long implicitly expressed that Prim and Huang Renxun are not the same person, and their styles of doing things are completely different. It is difficult for such people to survive in Huang Renxun's company. There are several examples: the first example is the NV1 they made (launched in 1995). Prim made technological innovations despite market demand, which led to the failure of the project. Of course, Huang Renxun was probably not mature enough at this time, and he failed to see it in advance. There is a problem here. When NV3 (launched in 1997), Prim still wanted to insist on using innovative technologies, which led to a public quarrel between Huang Renxun and Prim. Huang Renxun won and the company overcome the first difficulty. The second example is when creating NV1, deadline is approaching, and Prim plays games during working hours, playing well, which makes onlookers exclaim that this kind of work is not 100% committed, which makes Huang Renxun very angry.
There are still some positive comments about Prim in "The Way of NVIDIA", such as the general framework he developed makes chip upgrades more efficient, but "The Way of NVIDIA" has fewer positive comments. His method of doing things is very self-centered, which is difficult to be the management of the company. He is not even a suitable entrepreneurial partner. Another entrepreneurial partner, Malakowski, introduced less, and obviously it is not the success of the company in the future. The key role of Huang Renxun is very important.
3. Huang Renxun’s personal characteristics
He himself said: "I am willing and able to put in more effort than ordinary people, and at the same time I can endure more pain than ordinary people", "Greatness has nothing to do with intelligence, but from character", "Success requires resilience", "Education enough Pain and suffering”, “very low expectations”
After reading this book, I felt:Very hard work, continuous learning, and resilience are his most outstanding qualities. Decades of work have been working for a long time, constantly learning new knowledge, and being able to persist optimistically when encountering difficulties. These examples show his qualities. In addition, he has some characteristics, strong social skills, strong desire to win, strong managers, and strong goal orientation.
4. The key genes to success
(1) Work very hard.Huang Renxun himself is a workaholic. The entire company led by him has this atmosphere, and internal employees roll quickly, and only those who work hard and have the ability can stay. Of course, to make the entire company's work fanatic, in addition to screening those who are willing to fight, the company also needs to be able to continue to succeed and allow employees to share the benefits of the company's success. Nvidia has done a good job.
(2) Long-termism and timely capture trends.Continuously doing graphic rendering, doing it for decades, launching CUDA and subsequent cudnn, and then moving from RNN to Transformer, are all manifestations of long-termism and capture trends. Be prepared before the industry starts, and turn in time when the industry changes.
(3) Suitable management methods for large teams.The opportunity for Nvidia to establish political struggles within a large company. These experiences will probably make Huang Renxun intentionally avoid internal struggles in the company regardless of the interests of the company. The "flat", "information transparency", "project system", "whiteboard culture", "mission is the boss", and "five major matters" he adopts can have several aspects of value: First, make those who are capable People jump out to take charge of projects without sticking to organizational structure; second, they force managers to go deep into the front line, and they cannot just report to avoid being in a position; third, the CEO goes deep into the front line to ensure that the methods of doing things are implemented, while maintaining the CEO's judgment on the market force.
5. The way to grow a small company
(1) Make products that big companies don’t want to make.
Sun Microsystems and Silicon Graphics were large companies when Nvidia was founded. They were more willing to build high-profit workstations than low-profit PC markets, which gave Nvidia an opportunity. The direction of not doing big companies has always been the way for small companies to survive. For example, more than 10 years ago, investors often asked, "Does Tencent do it?" As long as Tencent does, other small companies will have no chances. It is difficult for a company to compete with a large company.
The situation has changed a little in recent years. When capital is no longer a decisive factor, small companies can do better in organizational efficiency and talent density. For example, DeepSeek, which has been very popular these days, has surpassed a large number of domestic companies. In terms of competition between large companies and small companies, on the surface, it is a small company against a large company, but in fact it is a small company against a team in a large company, a group of entrepreneurs against a group of workers, In terms of organizational efficiency and talent density, small companies have a full chance to win. The growth process of Nvidia is also a model in this regard. It has higher organizational efficiency and faster work. It has also attracted a large number of industry experts in the middle and late stages, and has formed a Matthew effect on talent scale.
(2) Avoid falling into political struggles and making mistakes.
When Prim proposed the graphic accelerator plan at Sun, he was not supported by the founder and was eliminated under the political struggle, leading to the founding of Nvidia. Founders of large companies rarely continue to work on the front line like "Huang Renxun". Founders who leave the front line are transformed into professional managers, which is easy to be unpragmatic. For example, Sun is trapped in the disease of a big company: competing with PPTs that impress executives, rather than competing with innovative ideas.
In order to solve the "big company disease", in recent years, I have seen some big companies' unique solutions: ban PPT and use WORD documents instead, as if PPT is the source of all evil. It is emphasized that the team must be younger, and young people must be promoted, as if a team that is managed by young people can win battles. I think the root cause of the disease is not the presentation tool, nor the team age. Although using WORD means more detailed details (similar to Huang Renxun’s whiteboard culture), it means that managers should go deeper into the front line; although the rapid promotion of young people means that the team is not rigid and recognizes credit rather than placing seniority, it is even more The essential point is to require management to maintain a mentality of being able to fight on the front line and to make combat-capable people stand out. I think Nvidia's flattening and project system are the better solutions, avoiding information poor and avoiding talents being bound by organizational structure. .
(3) Flattening and transparent information.
Even if Nvidia becomes a big company, they are still relatively flat inside, so decision makers can capture changes in front-line trends in a timely manner. Transparency allows all levels of the company to maintain a consistent mission. When doing project management, use the project system rather than hierarchical management, so that the right people can be on the right projects without being bound by the organizational level.
(4) Continue to reflect on and improve.
Technology and market disconnection and poor internal communication have happened at least twice in Nvidia, once NV1 and once NV30, but they all came through reflection. Huang Renxun advocated criticism and reflection in public, which formed the company's reflection style.
(5) Crisis, trends and extreme execution
The culture of a company is an extension of the CEO's style. The CEO works very hard and brings the ultimate culture of doing things. For example: The "speed of light" that Nvidia often calls is essentially the "first principle" or the "ideal state" we often call when making technical solutions, which is the best state we can achieve.
Nvidia's two sentences "The $0 billion market" and "Our company is only 30 days away from bankruptcy" can show crisis awareness and trend thinking. But there are very few companies that can do well. I heard the sentence "It's only 30 days before bankruptcy" for the first time when I first graduated more than ten years ago. It was part of Baidu's culture, but Baidu has been doing this for more than ten years. Staying in place and missing many opportunities, it is really getting closer and closer to bankruptcy. Also, the book "Only Paranoids Can Survive" mentions that Intel has been looking for the "ten times faster factor", but Intel has missed the mobile era and the AI era in the past decade, which shows that there is culture but no good executives. It’s also not good. From Nvidia’s story, we will find that the top leader is the best executor.
6. Timeline
Started a business in 1993
The first product, NV1, was launched in 1995 (failed)
In 1997, the third product, NV3, was launched, and it was resurrected (NV2 failed and was not launched on the market)
Name the graphics accelerator GPU in 1999
2001 NV20 (GeForce3) supports programmable shader technology, and GPU expands to non-game fields
CUDA was launched in 2006
2012 Hinton students’ GPU-trained AlexNet model won the ImageNet competition
ChatGPT appears in 2022
Note: This article is also stored on your personal official account.