core things (public account: AIchip001)
author | Xinyuan
editor | Moying
core things reported on June 3 that on Wednesday morning, NVIDIA founder and CEO O Huang Renxun had a remote video online exchange with global media such as Xintiao, and took everyone to "cloud" to visit NVIDIA's new headquarters building Voyager. During the
exchange, Huang Renxun responded to the latest progress in Arm's acquisition, his understanding of the Metaverse Metaverse, optimistic AI direction, measures to ensure chip supply and price control, whether he will build his own wafer factory, and whether he plans to introduce real-time ray tracing technology into mobile devices.
Huang Renxun and global media remote video
During the interview, Huang Renxun was like a kind big brother, casually complaining that he had just eaten a super spicy burrito.
1. VR/AR meta-universe can realize time travel
Huang Renxun was interviewed by the media and located in the new building Voyager, which will accommodate 3,500 NVIDIA employees. The media such as Xindongxi and other attending the conference were fortunate to become the first batch of people in the world to "travel the cloud" of this new building.
According to Lao Huang , Voyager is designed as a city inside the building, with a very large open space, and its roof angle is just right, and the sunlight will not shine directly in, which makes the eyes feel very comfortable. At the same time, there is almost no need to turn on the air conditioner, and the indoor temperature is quite pleasant.
NVIDIA simulates the entire building using ray tracing on its supercomputer DGX to ensure that all designs maintain a delicate balance between natural ambient light.
Huang Renxun explains that the corridor in the new building of NVIDIA headquarters
is very wide. Huang Renxun said that the reason for this design is that in the future, robots can stroll in corridors and carry things for humans. It can also be used as a "tool person" for remote virtual presentation. As long as you bring a VR/AR headset, you can feel the perspective of walking in the park by connecting to the robot's "eyes".
When answering his opinion on metaverse Metaverse, Huang Renxun also mentioned the example of simulated architecture again. Before constructing a physical building in the real world, you can design a data model similar to a physical building in the virtual world, train a robot, simulate lighting conditions, and simulate how to allocate air conditioners to reduce energy consumption. At the same time, the virtual world will follow the laws of physics. NVIDIA's Omniverse cloud native platform is providing such support, for example, BMW has used this platform to create digital factories.
In Huang Renxun’s view, the metaverse is actually a virtual world shared by many people, connecting to the real world we live in.
This virtual world has a real economy, and everyone has a real incarnation. You can communicate with your friends in the virtual world, and you can almost feel right next to each other. To some extent, it is a way to achieve time travel, allowing people to "reach" faraway places at the speed of light.
2. Confidence in regulatory authorities' approval of the acquisition of Arm
When asked about the progress of the acquisition of Arm, Huang Renxun revealed that they are passing regulatory approval, which is expected to take 18 months, and will probably be approved later this year or early next year.
"I am very confident in this transaction." Huang Renxun said, "They do CPU, we do GPU and DPU... Our company is complementary, so in essence, we will bring innovation, and these are the results of the gathering of companies."
In his opinion, this acquisition will bring more choices to customers. After the two companies merge, Arm's R&D scale will become larger, with more professional R&D knowledge, and more resources to invest in R&D, thereby creating better products and giving the market more choices. For NVIDIA itself, the deal will bring about a larger ecosystem of developers and create greater value for both companies.
Regarding whether Arm will surpass x86 in the future market, Huang Renxun believes that the market will be diversified and there is no only answer.
NVIDIA will continue to support x86 CPUs in the market it serves, and also support CPUs based on Arm architecture based on market applications and computing needs.
3. AI direction, especially optimistic about the recommendation system and natural language understanding
. There are two AI models that make Huang Renxun very interested, one is the recommendation system and the other is natural language understanding.
recommendation system is called "the most valuable software method in the world" by Huang Renxun, driving all Internet services; natural language understanding is smart enough to answer questions and make suggestions with conversational AI.
In his opinion, these two models are the most valuable software in the world today and require a very large machine. This is the original intention of NVIDIA Grace CPU design.
"We decided to design something for those applications that need large AI. At the same time, there are many different markets and advantages and enterprises that support CPUs. So I believe in the diversity and personalization of the future world," said Huang Renxun.
4. Talking about the chip shortage problem, using two strategies to ensure the supply of gaming GPUs
In response to the chip shortage, Huang Renxun believes that cloud computing data center is booming, and smartphones have been using the most advanced technology, and the technology is developing faster and faster, because of these dynamics, semiconductor demand exceeds supply, which is inevitable.
From NVIDIA's performance in the previous quarter, we can see that NVIDIA has sufficient supply to support growth in the second quarter and the second half of the year. At the same time, Huang Renxun also hopes that NVIDIA can have more supply.
NVIDIA did not raise prices. Faced with the problem of the feverish mining tide raising the market price of GPUs, NVIDIA's strategy is to directly create a special product for miners CMP. Because CMP cannot be used to play games, it also has the advantage that it cannot be resold to GeForce customers.
Huang Renxun also revealed that NVIDIA's CMP revenue reached US$150 million in the last quarter, and is expected to reach about US$400 million this quarter.
Meanwhile, NVIDIA reduces the hash rate of GeForce GPU, thereby reducing its attractiveness to cryptocurrency practitioners.
That is to say, if the user insists on using the game GPU GeForce to mine, the performance will not be as good as the CMP product. In this way, NVIDIA hopes to retain more gaming GPU supplies for gamers, while slowly reducing supply prices.
5. Will not build a fab by yourself
A reporter asked, considering the long-term chip shortage, is NVIDIA planning to build a fab by yourself?
In response to this, Huang Renxun replied quite decisively: "No!"
"There is a difference between a kitchen and a restaurant, and there is a difference between fab and foundry. There is no doubt that I can open a factory, just like I can run a kitchen, but it will not be a good restaurant," he made an example, "You can build a wafer factory, but this is not a good wafer factory."
wafer factory integrates service, agility, technology, and capabilities. It requires courage and intuition about the future. In Huang Renxun's view, the work that TSMC relies on for a living is not easy. This won't get easier, but it will get more and more difficult.
"There are a lot of people who are very good at their jobs. We have no reason to repeat it again. We should encourage them to develop the necessary capabilities for the benefit of our company." Huang Renxun said, "At the same time, they now realize that they are now aware that consumption of cutting-edge wafers has been changed because of the way the computer industry is developing. They see opportunities right now, so they are increasing capacity as quickly as possible. I don't think I can do anything about it."
6. It's still time to introduce RTX technology to consumer devices
Huang Renxun said that NVIDIA is very good at involving Arm SoCs, covering products ranging from autonomous robots, autonomous vehicles, etc. As for real-time ray tracing (RTX) technology, when the time is right, NVIDIA can consider introducing it into consumer devices.
He mentioned another way to enter the mobile field, namely NVIDIA's GeForce Now cloud gaming service, which was launched in early 2020. Now there are more than 10 million players from about 70 countries and 1,000 games from about 300 publishers.
"This is the Android device, Chrome device, iOS device, MacOS device, Linux device, and all kinds of devices, whether it's a TV, a mobile device or a PC."Huang Renxun said, "I think this is the best strategy we need to do now. ”
Conclusion: In the face of fierce competition, NVIDIA will accelerate the run
Judging from a series of games and enterprise-level new products launched by NVIDIA this week, as well as Huang Renxun’s interview content, the momentum of NVIDIA’s accelerated computing is increasing. Huang Renxun is quite confident in the growth in the next quarter and the second half of the year.
accelerated computing is contributing to the information world, and one of the most important applications is artificial intelligence. Huang Renxun believes that data centers are the largest market opportunity in the IT industry so far, and GPUs are an ideal accelerator for general-purpose applications.
Therefore, regarding the phenomenon that many Chinese GPU startups have obtained high financing in the past two years, Huang Renxun said that he understands why this market can inspire so many competitors. He also talked about NVIDIA facing, respecting and taking all competition seriously, and will continue to do its best to ensure that it runs very fast.
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-