Chapter 297 Butterflies will eventually cross the sea (6.4k)
"It's Yuan Guang's bad taste. Don't you think this scene is very similar to the Macintosh commercial during the Super Bowl?" Wang Ruixin watched it As Chen Yuanguang's head occupied the entire screen, he couldn't help but complain.
The tickets were obtained by Li Tianle. Weilai and Guangjia Aerospace have a deep cooperation. A lot of tickets for this conference were reserved for these cooperative companies.
It is extremely easy for Li Tianle to get two tickets for the press conference.
Li Tianle, who was sitting next to him, asked in a low voice: "What advertisement?"
(A classic advertisement in Apple’s history)
“You know “1984” written by George Orwell.” Wang Ruixin said.
“I know.”
“Apple shot an advertisement called 1984. The meaning of that advertisement is that 1984 will not be the 1984 written by George Orwell, but the 1984 written by McGinn. Tower Computer changed 1984.
The ad was basically about a woman wearing red shorts and a white tank top with a Mac logo rushing into a hall full of expressionless, robot-like bald men and hitting them with a hammer. Broken screen showing Big Brother's lecture
The huge avatar on the screen is very similar to this scene. If TVs of various sizes are placed around the conference, and Yuan Guang's avatar appears on the TV at the same time, it will be even more similar. ”
Li Tianle was a little confused and didn’t quite understand what the other person wanted to express.
He looked around, everyone was immersed in Chen Yuanguang’s speech, only the two of them were chatting.< br>
Li Tianle’s voice became even lower this time, “What do you want to express? ”
Li Tianle knows that his friend is not the kind of person who likes to make blind shots, so what he said must have a meaning.
Wang Ruixin said: "If there is a religion of science and technology in China, then Yuan Guang is the true leader.
Don't you think this scene is like a sermon? Everyone is worshiping.
It’s just a matter of doing some religious rituals.
I didn’t see that Yuan Guang had any talent in this area before. ”
Modern history has led to Lao Zhong believing in technology. , if there really is such a thing as the Technogod, Laozhong must be the earthly kingdom of God for this type of sect.
After hearing this, Li Tianle shook his head helplessly and did not answer any more.
Chen Yuanguang was still saying on the screen:
“Everyone should know that Dongda’s consumer electronics, semiconductors, new energy vehicles and other fields have faced problems due to the suspension of supply by foreign suppliers.
Chips, especially advanced process chips, we used to rely heavily on imports. If we relied on domestic supply chains, we could only produce 7nm chips at most. However, topological semi-metallic chips are essentially stacked. The transistor method is equivalent to 1nm.
The transistor size does not really reach 1nm. The equivalent means that it is equivalent to a 1nm transistor.
What is very helpless is that the new photoresist, the core raw material of the topological semi-metal process, has also stopped supply.
Everyone is more or less aware of the impact on our industry. I just want to say one thing this time, that is, the impact will not be too long and will be much shorter than Washington predicts.
At most half a year, the problem can be solved in half a year.
For suppliers led by Xida and Neon, the moment they choose to withdraw from the Dongda market, the door of the Dongda market will be closed to them forever. ”
The audience burst into applause like a roaring tsunami.
After the applause died down, Wang Ruixin turned to look at Li Tianle: "Look, this is Yuan Guang's appeal.
Although Yuan Guang did not explain clearly what process was used and how to solve the problem of chip limitation.< br>
But the five thousand spectators present have no reason to believe that Yuan Guang can solve it.
I think Yuan Guang’s nickname should not be the God of Light, but the leader.”
< br>Li Tianle said of course: "This is natural.
If we hadn't been roommates with Yuan Guang and just looked at Yuan Guang's achievements, we would still worship him.
Even as a roommate, I admire Yuanguang's achievements.
Many things can be said to be luck, such as room temperature superconductivity, topological semimetals and perovskite batteries. I have seen some people on Zhihu or Quora saying that they are good luck.
I don’t believe this argument. Why is it that Yuan Guang always has good luck and not others?
But even we think these are just good luck.
Deciphering the laws of virus evolution, inducing the virus to evolve in a specified direction through a broad-spectrum vaccine, and creating a new vaccine design theory is not always good luck.
A virus prediction model was established and successfully predicted the mutation of COVID and the mutation point of the virus in five years.
This is not just luck.
As for the artificial intelligence model for predicting the properties of small molecule metals, this model is still used today by many companies I know. Based on this model, many algorithms that can be used in practice have been developed.
Is this achievement a small breakthrough in the field of computational chemistry in recent years? It has led to the improvement of the status of computational chemistry as a whole.
These latter points are enough to prove that Yuan Guang's basic skills are solid. It's just that too many more dazzling achievements have covered up these things that prove basic skills. "
Wang Ruixin immediately said: "I just think this scene is a bit strange.
Maybe I haven’t been back to China for a long time, and I always want to complain when I see such a scene. ”
Wang Ruixin did not defend anything, and Li Tianle also said nothing.
The video on the stage had finished playing, and after the applause subsided, Lin Jia came out to speak:
"I have many identities. I originally thought that my next press conference would be to hold a new energy vehicle product launch for Space, but I didn't expect that I would be holding a product launch for Biren Technology.
I I also browsed major social media, and there were a lot of interpretations about Biren Technology on the Internet. I believe that everyone who came to the scene also knew what Biren Technology was like.
It is as everyone guessed. , the product we are releasing this time is the graphics card developed by Biren Technology - Longyin I type.
We are divided into two versions: basic version and pro version.
Run the scores on the ladder. Among them, the basic version of Longyin I is the benchmark 3080, while the pro version is the benchmark 3080ti.”
The main companies that make GPUs in China are Moore Threads and Biren Technology. Before that, Biren Technology mainly made AI computing cards, which means they did not develop graphics cards for players.
Therefore, Moore thread is the only one that develops graphics cards for individual users, and Moore thread products can be described in one sentence as being usable.
The most advanced personal products launched by Moore Threads before were only benchmarked against 2060. Even due to driver and adaptability reasons, the running scores were comparable to 2060. However, in many fields, the actual use may only be Reaching the level of 1070Ti.
You must know that 1070ti is a product of 2016.
That was ten years ago.
Biren Technology is equivalent to raising this level from 2016 to 2020 at once.
The applause from the audience became even louder.
You talk about buying an AI computing card, but not many people will buy this thing themselves.
But when you talk about buying a graphics card, everyone is still very interested.
“Do you want to buy it or not?”
"What about you?"
"I don't advise you to buy. You advise me not to buy. I want to wait and see."
"I will definitely buy it. The God of Light himself will not come out to personally This is probably the first time for a user’s product. I will definitely support it.
I don’t think I can afford the car it sells in the future. I can afford it if it supports a graphics card. < br>
Besides, Western University is targeting the domestic semiconductor industry. I will support it with real money.”
When the individual audience members in the audience were discussing, most of them thought they could support it.
“Longyin I adopts an equivalent 6nm process, but we can actually achieve 3nm.” Lin Jia said. "Relying on the national supply chain, we can achieve the equivalent of 3nm.
So next year, here, we will hold another press conference to launch the Longyin second-generation graphics card that benchmarks NVIDIA 50 series graphics cards. . ”
Produced nationwide, this is like an aphrodisiac, especially at this juncture.
The main reason is that everyone used to think that just de-mizing it would be enough. This is also a way to solve the problem. Now it seems that de-mizing is not possible and must be made nationwide.
Without domestic production, we will face crises at any time. After all, you want to expect other countries to listen to you instead of Western University. This is unrealistic.
Even France is a beneficiary of the international order constructed by Western University. They may sway on non-key issues and may side with you for some reasons. On key issues, they At most, maintain a neutral attitude.
To a certain extent, the entire industrial chain is produced nationwide and has gradually turned into a giant industrial Leviathan. This is also forced.
“The Dragon’s Song II just now is just a picture. Next, let’s go back to the Dragon’s Song I graphics card released this time. Let’s take a look at two real-machine demonstration animations, which were released the year before last. The domestic 3D masterpiece "Black Myth: Wukong"."
After Lin Jia finished speaking, the screen was divided into left and right sides, both showing the same game animation, and the animation effects were very smooth.
"We can see that without turning on the ray tracing special effects, the basic version of Dragon Song I can perfectly support the 1080P image quality display of Black Myth, while the pro version can support 4k ultra-high image quality.
Of course, the premise is that ray tracing is not enabled.
This graphics card is not yet equipped with ray tracing technology.”
The conference lasted about an hour. After listening to the whole process, Wang Ruixin said Li Tianle and I went to the nearest Hangzhou cuisine restaurant for dinner.
Wang Ruixin asked: "What do you think is the most important message in the whole scene?"
Li Tianle knows this clearly: "In chip technology, the national production can reach 6 nanometers this year and 3 nanometers next year.
To put it bluntly, the time window for Western University to obtain this time is only one year. Foreign chip manufacturers The time to enjoy the premium is only one year.
If they cannot achieve a major technological breakthrough within this year, then Dongda will catch up.”
“Yes, this is definitely their chance. I didn’t expect it before making the decision.
Because Dongda’s domestic lithography machines have only conquered 14nm in public, and there is still a lot of water in this 14nm, it is just de-mized. < br>
From the de-mized 14nm to the nationally produced 6nm at today's press conference, the gap is too big. The decision-making group in Washington made a fatal strategic misjudgment. "It can only be said that Yuanguang is powerful enough and can counterattack in just one month." "Li Tianle said with emotion.
Yuanguang University was the focus of the School of Chemistry, but at that time he became the focus not because of scientific research results, but because of his relationship with Lin Jia. At that time, it was not easy to get girls from the School of Economics and Management.
Not to mention that Lin Jia was one of the top five girls in the School of Economics and Management in terms of appearance. In terms of overall strength, she was the number one girl. >
Everyone was envious at that time
Li Tianle lamented that even today, Yuan Guang is still the focal point among them.
“No, it definitely cannot be completed within a month. It has definitely been prepared a long time ago, but it was only taken out today.
Like Biren Technology and Maolai Optical The acquisition, from R&D to tape-out to production, has definitely been completed.
It was only announced at this point in time, and the changes in industrial and commercial information were only recently exposed.
I always feel that this is a hole dug for Western University in advance. If you don’t lift the table, then we will cooperate well. If you lift the table, then everyone will stop playing together. We can defeat the semiconductor bureau without you. Keep playing.
This time we started with graphics cards, and it was clear that we were targeting Nvidia. We can conquer graphics cards, so Dongda can come up with alternatives for all types of semiconductor chips. "
"This is a good thing. ”
I don’t know if this is a good thing, but it is definitely a good thing for consumers.
The next day, pre-orders for the Longyin series of graphics cards were launched on all platforms, including Taobao, JD.com, Pinduoduo, etc. Pre-sales are launched simultaneously on both channels.
The price of the pro is 1,999 yuan, while the basic version is only 1,599 yuan.
The graphics card performance of the 3080ti is similar to that of the 4070, and for the 50 series, the benchmark is the 5060.
Compared with NVIDIA graphics cards with similar performance, the price of the Longyin series is about 40% lower.
This price is definitely a great thing for consumers who have recently purchased or replaced their phones.
“Do you think this price can be sold?” asked an old employee of Biren Technology.
Basically all the senior executives of Biren Technology have been changed, and the middle-level leaders of the functional departments have also been changed.
Basically, they were replaced by employees from Light Armor Aerospace.
There are no changes in the lower-level employees.
Even those who originally wanted to leave their jobs stopped thinking about it after learning that their company was acquired by Guangjia Aerospace.
This is equivalent to joining Guangjia Aerospace directly. It is no different from transferring from Harbin Buddhist College to Harvard with the mentor.
Some bloggers on Zhihu who got offers from Biren Technology but didn’t go there regretted it to death. They were like I got offers from *** company and Biren Technology, which one should I go to?
Such questions are filled with rhetorical questions from new respondents: "Did the blogger go to Biren Technology?"
Everyone wants to see regretful answers to satisfy their own bad taste.
“Don’t worry, we only need to worry about production capacity. This price is too low.
Even if our price is a little more expensive than NVIDIA’s equivalent chips, consumers will still Pay the bill.
Do you know what domestic premium and patriotic pricing are? If Huawei can enjoy it, we can also enjoy it." Employees from Guangjia Aerospace understand this.
"Then why is the price only set at 2,000 yuan this time?"
"It should be out of caution. After all, this is the first time to make a graphics card."
"I don't think so. The audience for graphics cards are young people, and these young people are price-sensitive consumers. They may pay a little premium for their patriotism, but if the premium is high, only a few of these consumers will buy it.
There are only a few young people who accept Huawei’s pricing, and most of them are middle-aged men who are not short of money,” said another employee who came from Guangjia Aerospace.
After the official sale at ten o'clock in the morning, the 100,000 units prepared were sold out in fifteen minutes.
The revenue in fifteen minutes is equal to the revenue of Biren Technology in the past year.
“You are right, I still underestimated Yuanguang’s appeal. I thought the sales of graphics cards would be very limited, and there are still no market-proven graphics cards.
There are so many consumers blindly placing orders directly.”
“The domestic demand itself is huge enough, and combined with everyone’s frustration over the past month, this order can be digested in a matter of minutes.”
At about the same time that day, OPPO officially posted on Weibo:
"After three years, the butterfly returns"
The accompanying picture is a front-side photo of a chip.
Many people in the digital community know what this means at a glance.
In the first quarter of 2023, OPPO's Zheku team successfully taped out TSMC's 3nm chip. Everyone thought that there would be a second self-developed chip manufacturer in China. The first one was Huawei's Hainan. Thinking of Qilin.
3nm will be a high-end chip in 2023, a chip that can compete with the flagships of Apple and Qualcomm.
Before domestic consumers could wait for more information to be exposed, what everyone was waiting for was the news that Zheku had been disbanded, with only the core architecture team retained.
Three years and return, the meaning couldn’t be clearer.
“In the past three years, there have been frequent news of Zheku’s reorganization. It was not until the end of 24, after Huawei launched the Ark chip, that Zheku officially restarted.
However, this time Zheku is also different from Zheku in 2023. It makes different chips and requires different talents. It is also impossible to directly use the architecture of the tape-out in 2023. Everything has to be overturned.
Zheku's team previously had more than 3,000 people. Before its dissolution, it had launched two self-developed chips, Mariana MariSiliconX and Mariana MariSiliconY. The former was a 6nm process NPU chip, which was very advanced at the time.
In terms of technical strength, it has been fully proved that Zheku has no problem at all, and the probability of making a chip that is on the same level as or even better than Kirin is very high.
Dissolution is very common, but direct dissolution of business departments is uncommon, and direct dissolution of a company with 3,000 people has never been done.
Generally speaking, if chip companies really can’t survive, they can lay off employees bit by bit.
Foreign-funded chip companies like Shenhai basically cut off 50% of their employees this time and another 50% next time. At that time, Zheku laid off 3,000 employees at once, and the compensation provided was also very sincere. It was not at all due to poor management that the employees were laid off. To put it bluntly, it was because of external pressure.
Once you do it, you will face the government in Washington, either TSMC will not OEM for you, or even Qualcomm will not sell chips to you.
The restart of Zheku two years ago was also based on self-research. Who would have thought that Washington's butcher knife would still cut it off. Not only OPPO would not be spared this time, but all mobile phone manufacturers would not be spared.
You are not allowed to do it in foreign markets, you can only do it in the Dongda market.
Originally, these domestic mobile phone manufacturers were very cautious. Although they had established a self-developed chip team, they all wanted to gradually replace it and first launch self-developed topological semi-metal chips on flagship phones.
After all, this thing does have better performance. It has better power consumption, heat dissipation and performance than traditional silicon-based chips. Huawei has this thing, but other mobile phone manufacturers cannot, and it cannot compete in the domestic market. Huawei.
Even so, they only think about Qualcomm chips and we will continue to buy them, so as not to irritate Amerikan.
In the end, the butcher's knife was cut, and it was even more targeted than they thought. You have to buy Qualcomm chips, and you have to let go of foreign markets.
Just at this time, domestic chip manufacturers gave an olive branch, OEM, right? We can OEM, but we can only OEM 6nm, do you want to do it?
OPPO has Zheku before and has tape-out experience. Anyway, Dongda does not allow us to enter foreign markets. I might as well cooperate with a company in the optical armor series.
As for another level of worry, OPPO insists on developing self-developed chips, and Qualcomm will completely stop supplying them. I don’t think there is any need to worry about this. You cannot only target OPPO.
If you target all Dongda mobile phone companies, doesn’t that mean you are giving up the market to Huawei? In addition, Apple and Samsung cannot even think about entering the Dongda market.
So I think OPPO was forced to have no choice this time and made up its mind to go against Xida. ”
Analysis by a certain OPPO digital blogger.
But one thing in his analysis is wrong, and that is the manufacturing process.
Frontier Technology OEM for OPPO Not 6nm, but 3nm.
The version of the chip design that was tape-out in 2026 will be used directly. It was already tape-out successfully that year. This time, nothing needs to be changed, only debugging is required. Craftsmanship.
OPPO Senior Vice President and former OnePlus founder Liu Zuohu posted a long Weibo post, which can be regarded as OPPO’s semi-official channel to express his voice:
“The chip that OPPO will release this time, we named it Butterfly. Butterfly-9400S, manufactured using a 3nm process, is the same chip that Zheku had difficulty delivering three years ago.
When developing this chip, our Zheku team worked 7x16 hours a week. After carefully studying a large amount of competitor data and referring to a large number of demand use cases, the entire architecture design team completed a highly competitive architecture design in just 4 months and completed performance functions within 6 months.
This is not an easy task. The SoC that did not come out has functions that are outstanding even today.
It uses the latest ArmCortex- X-series CPU cores and latest GPU cores, rich CPUL1/2 cache and huge L3/SLC cache, sub-100ns DDR latency, asynchronous MTE, AIDVFS/FI/SR), provided by a series of efficient in-house custom IP Power.
On the night when OPPO decided to disband Zheku three years ago, I kept playing Faye Wong's song on repeat in my office in the OPPO office building. That song was called "Butterfly".
The reason why I played this song is because there is a lyric in this song, which is: Just like a butterfly that cannot fly across the sea, no one has the heart to blame.
We gave up Zheku and self-developed chips because of the pressure from Amerikan. This pressure forced us to choose to bury our five years of hard work with our own hands.
This kind of pain cannot be described in words. My thought at the time was that maybe the butterfly would never be able to cross the sea.
Coincidentally, the chip that Zheku only developed and successfully mass-produced is called the Mariana Trench. The Mariana Trench is the deepest trench on the earth. Our idea at the time of naming it that way was that we were going to The most difficult thing in the world is to explore the limits like going deep into the deepest part of the Mariana Trench.
The result is that butterflies cannot fly across the sea, Zheku cannot fly across Mariana, and even if the tape-out is successful, there is no way to launch our self-developed 3nm mobile phone SoC.
Later, when we were preparing to develop our own chips last year, this time we used a new process, but faced the same dilemma. I was almost desperate this time. As you all know, we encountered such force majeure continuously. Isn’t it so painful that a chip that can obviously be launched cannot be released?
I once thought that OPPO was inherently at odds with its self-developed SoC. Fortunately, before we disbanded Zheku this time, Guangjia Aerospace contacted us and asked us to wait. The result is our own domestic chip. The industry chain can achieve the 3nm process.
Today, the chip that failed to come out that year will be launched within this year, and this time it will come back through a nationally produced process.
The butterfly will eventually cross the sea! ”
OPPO has a natural advantage because they have succeeded before and can just use that design.
This was almost the last life-saving straw for the OPPO chip R&D team, so after receiving the invitation, they almost lost no time and the entire team moved to Mianyang to start work.
After the successful tape-out verification, OPPO officials and senior executives began to make public statements.
(End of this chapter)