Rebirth 08: Rise from copycat phones

Chapter 662 A Family of Three
Looking at the APO7000 graphics card in front of him, Xu Shenxue felt even more confident about the future development of artificial intelligence technology.

The Artificial Intelligence Research Institute is continuously conducting research and development, attempting to develop more advanced artificial intelligence technologies.

According to Professor Lin, they hope to innovate the core algorithms of artificial intelligence and try to develop an artificial intelligence algorithm that can handle both low-precision and high-precision calculations!

This artificial intelligence algorithm will use low-precision computing power to simulate the fuzzy decision-making of the human brain, while high-precision computing will maintain the ultra-high precision of the computer. By combining the two, it can take into account the advantages of both the human brain and the computer!
That's incredibly powerful... It's almost like having a computer cheat code applied to a person's brain in a science fiction novel.

Of course, it's not easy to do this!

High-precision calculations require powerful superconducting quantum computers and APO computing cards to provide massive computing power, while low-precision calculations require top-tier neuromorphic computers to provide powerful low-precision computing capabilities.

This involves three core areas: superconducting quantum computers, GPU-based computing cards, and neuromorphic chips!
Then, how to combine these three types of computing power to work together is also a huge project... This is the YANC system that Zhuyun Group has spent many years developing.

Finally, we need one more crucial element, the soul of artificial intelligence technology: the core underlying algorithm!
Breakthroughs in the hardware field relied on an industrial base and a large number of highly skilled top-tier scientific researchers.

The latter, the field of algorithms, is purely the exclusive domain of a very small number of top scientists... Ordinary high-level talents don't even have the qualifications to participate or assist.

Specifically within the Zhiyun Group, this matter is more directly Professor Lin's responsibility... Currently, among all of humanity, only a mathematician with superhuman intelligence like him can accomplish this kind of work. Others... let alone research and development, they can't even understand Professor Lin's research results!

The core underlying algorithm is still based on mathematical theory and the development of mathematical tools... It's pure brainpower, and there's no other alternative.

Therefore, when the next generation of artificial intelligence technology can be developed directly depends on when Professor Lin makes a new breakthrough in the field of mathematics!

This is also why Professor Lin is the only W15-level scientist in the Zhiyun Group, and his security level is the same as Xu Shenxue's... The role of such a top scientist is enormous.

It is currently unknown when Professor Lin will be able to complete it, but Xu Shenxue believes he can do it... If all else fails, Xu Shenxue will arrange another SSS-level research quota for him.

In the past two years, Xu Shenxue has accumulated a lot of achievement points in the system, with about three million achievement points on hand. Now he can add about two million achievement points a year.

If all else fails, he'll just endure it for about four years, saving up all his achievement points without spending a single one, and then accumulate ten million achievement points to exchange for an SSS-level research quota.

With this level of research opportunity, Professor Lin, a top mathematician whose intelligence is already at the pinnacle of human potential, will undoubtedly be able to break through the limits of human capabilities and unleash research and development capabilities beyond human imagination!

In short, no matter what, Xu Shenxue is determined to develop this new generation of artificial intelligence technology!
------

After looking at the APO7000 graphics card, Xu Shenxue also saw several other three-nanometer chips that were in the small-scale trial production stage.

Including the S1303 chip expected to be used in the flagship S22 released this year, the core of this chip is actually very small, only 120 square millimeters, but it does integrate multiple CPU cores and GPU cores, as well as a high-efficiency NPU core - that is, AI core.

In addition, it integrates a 5G full-network compatible baseband and other core management functions.

These days, a mobile phone SoC is a highly integrated chip with extremely powerful comprehensive capabilities; it's not just a simple CPU.

However, despite integrating so many cores and containing over 20 billion transistors, the chip is not very large!

Because it uses Zhiyun Microelectronics' N3 process!

Zhiyun Microelectronics' N3 process achieves a transistor density of 280 million transistors per square millimeter... According to information, TSMC is currently developing an equivalent 3-nanometer process, which is expected to be mass-produced in a few months, with a transistor density of only about 250 million transistors per square millimeter... Their next-generation improved 3-nanometer process is planned to have a transistor density of about 270 million. This process is still in the technology verification stage, and it is estimated that it will take at least two years before it can be mass-produced on a large scale.

And that doesn't even include yield rate.

Zhiyun Microelectronics' N3 process yield has reached 72%, and it is expected to reach 75% in June.

However, the yield rate of TSMC's N3 process is reportedly only around 50%... This difference in yield means that TSMC's 3-nanometer chips will be significantly more expensive than those of Zhiyun Microelectronics.

Considering the performance gap, the fact that their 3-nanometer process will take several more months to reach mass production, and the yield rate issue.

Overall, according to the internal analysis and estimates of Zhiyun Microelectronics, the company is at least three years ahead of TSMC in logic chip process technology. It will take TSMC at least two years to improve the yield of the 3-nanometer process to more than 75%, and then it will probably take another year to catch up with the performance difference in the 3-nanometer process.

Don't underestimate the short timeframe of three years; it's quite remarkable in the fiercely competitive advanced semiconductor industry. TSMC is backed by a large amount of capital from Europe, America, Japan, and South Korea, and it's the official semiconductor foundry for three major companies: Apple, Qualcomm, and AMD.

They have no shortage of money or manpower for research and development, and they also have no shortage of top-notch EUV equipment and other types of semiconductor equipment... and they have no shortage of the market.

At the same time, such technological progress involving the entire industrial system is difficult to achieve with just a few top scientists. So, to widen the gap, we can only do it step by step... widen the gap here, widen the gap there, and then combine them to form a huge technological difference.

Meanwhile, while others were catching up, Zhiyun Microelectronics did not stop. It continued to move forward, and at an even faster pace. Three years later, barring any unforeseen circumstances, Zhiyun Microelectronics' main processes had fully transitioned to two-nanometer processes, and it had begun to work on 1.5-nanometer processes on a small scale.

Because the large numerical aperture EUV lithography machine from Gulf Technology is already undergoing prototype testing. After a year and a half of testing, it is estimated that it will be mass-produced and used in the construction of the new factory in the second half of next year.

Once this new factory is completed, Zhiyun Microelectronics will begin working on the equivalent 1.5-nanometer process, and even the future 1-nanometer process.

It will be extremely difficult for TSMC to catch up... Even though they have the support of the industrial systems of Europe, America, Japan, and South Korea, as well as the semiconductor industry, technology is not something that can be easily caught up with.

------

Thanks to the high efficiency of Zhiyun Microelectronics' equivalent 3-nanometer process, with 280 million transistors per square millimeter, the area of ​​the S1303 chip is actually very small, only 110 square millimeters.

It's roughly the size of a fingernail... Of course, if you include the area of ​​the outer packaging, it looks a bit larger.

The chips that people see every day are relatively large because they include an outer packaging layer. The actual chip inside can only be seen after the outer packaging layer is polished away.

S1303, Xu Shenxue attaches great importance to the release of the S22 Max/S22 Max Pro models this year... with revenue of at least hundreds of billions of dollars, it's impossible not to pay attention.

Besides the S1303, Xu Shenxue also saw the ZY50 chip undergoing testing and tape-out at Factory 32... This type of large-scale data processing chip is also very important. This chip is a data processing chip specifically developed by Zhiyun Semiconductor to support artificial intelligence technology; it has no other purpose than processing data collected from a large number of sensors.

Currently, it is mainly used in intelligent robots and virtual devices. In addition, it will also be used in some high-end models of Hailan cars. Then, the X14 series low-altitude drones and other large-scale civilian artificial intelligence terminals will also be used.

In its early days, this chip was used in large-scale artificial intelligence terminals in the defense field, such as the X-series artificial intelligence fighter jets. However, it was later replaced by a dedicated data processing chip for defense. Similarly, the EYQ chip was replaced by a high-performance general-purpose computing chip for defense.

These high-performance chips, designed specifically for defense, can withstand more extreme environments without considering cost. Therefore, they are made with a relatively large area and have good fault tolerance. In addition to being used in the defense field, this series of high-performance chips for defense is also used in a series of spacecraft of Nanmen Aerospace, especially space robots and interstellar spacecraft.

In fact, since last year, the yield rate of the 3-nanometer process has been only over 50%. Before Zhiyun Microelectronics had even started large-scale mass production, it had already begun small-scale production of a batch of new-generation computing power chips and data processing chips for defense purposes... Defense chips don't care about the cost; they only care about performance!

Let alone a yield rate of over 50%, even if it's only a few percent, they'd still dare to do it!
Because these defense-specific chips are produced in small quantities, most of the cost is actually research and development costs and tape-out costs... This figure can easily reach billions of dollars.

The actual manufacturing cost is very low; the price of a single 12-inch silicon wafer using the 3-nanometer process is only around 200,000... Compared to billions, 200,000 is negligible.

Furthermore, defense chips costing a few thousand or tens of thousands of yuan to manufacture are insignificant compared to the hundreds of millions or even billions of yuan worth of X1 and X3 fighter jets.

The cost requirements for defense-specific chips and mass-produced commercial chips are vastly different and cannot be generalized.

Therefore, the APO7000 graphics card at Zhiyun Group has not yet been used, while the X1 and X3 fighters next door have started using the new generation of 3-nanometer process computing chips last year.

When Zhiyun Group started using APO7000 graphics cards on a large scale, they had already completed the upgrade of all their old models that were still in service... Incidentally, this upgrade work was also done by a secret subsidiary of Zhiyun Group. After all, this kind of upgrade is not just about replacing a chip; it also requires replacing a whole set of supporting equipment, updating the data protocol, and upgrading to a newer version of the X system.

So even though the earliest batch of X1 fighter jets have been in service for many years, their core computing system is actually upgraded every one or two years, always maintaining the latest technological level.

When Zhiyun Group was designing the system, it took upgrades into account... so upgrading the computing power system is very convenient.

However, these defense-specific chips are classified as highly confidential, and it was not appropriate for Xu Shenxue to take a group of people to see them, so he did not go to see them.

Sigrún has taught at the Iceland University of the Arts as a part-time lecturer since and was Dean of the Department of Fine Art from -. In – she held a research position at Reykjavík Art Museum focusing on the role of women in Icelandic art. She studied fine art at the Icelandic College of Arts and Crafts and at Pratt Institute, New York, and holds BA and MA degrees in art history and philosophy from the University of Iceland. Sigrún lives and works in Iceland.

Xu Shenxue was very satisfied with this inspection tour of Factory No. 32, as the results in all aspects were quite good.

The 3nm process has a good yield and has already entered the mass production stage. Before long, the newly produced APO7000/240GB self-use version will be able to be used in the second-generation YANC system.

The mature N4 and N5 processes are also performing well, with yields continuously increasing, costs continuously decreasing, and production capacity continuously increasing. This is also very important for Zhuyun Group's strategic plan to apply the equivalent five-nanometer process node on a large scale in various fields.

Before leaving Factory No. 32, Xu Shenxue met with the factory's management and key scientists as usual, and even spotted Yu Taihe's younger brother in the crowd!

Yu Taihe is the inventor of the room-temperature and room-pressure superconductor MMK34. Although it was ultimately proven that his invention of MMK34 involved a large element of luck, and his actual skill level was slightly inferior, the scientific community only values ​​results…

Since he has created such a top-tier material as MMK34, then Yu Taihe is a top-tier scientist.

Therefore, Yu Taihe is now a W13 level scientist at Zhiyun Group.

He first served as the director of the Guangcheng Artificial Intelligence Research Institute under the Zhiyun Group, and later rose to become the director of the Guangcheng Materials R&D Center under the Zhiyun Group... He has begun to transform into a management scientist.

He's a bit less skilled at R&D these days, but he's still very good at leading teams and managing affairs...

Xu Shenxue was quite familiar with Yu Taihe, but he didn't expect his younger brother to be so outstanding. He had become a W10-level R&D staff member at Zhiyun Microelectronics and was a core engineer at Factory No. 32.

This level is quite remarkable for a regular R&D staff member. In the 32nd factory under Zhiyun Microelectronics, it is enough to lead the R&D of an advanced process project... Outside of that, that would be considered a top-tier semiconductor process scientist.

If they could poach talent, companies like Four Star Semiconductors would be willing to offer at least three to four million US dollars in annual salary to lure them away... Four Star Semiconductors is currently technologically lagging behind and has been trying every means to poach talent from all over the world. They are also willing to spend money to recruit highly skilled personnel.

Unfortunately, they can't poach them... All of Zhiyun Group's core technical personnel are bound by non-compete clauses, which not only incur hefty penalties but also easily lead to the leakage of their original company's secrets, making them easy targets in lawsuits... In that case, the companies that poach them will face exorbitant compensation amounts in the hundreds of millions of dollars.

Poaching talent is too costly!
Of course, Yu Taihe's younger brother's annual salary is also not low. His basic salary starts at seven or eight hundred yuan, and then there are stock incentives. His total annual income will not be less than 12 million yuan.

This level of income is already top-tier in China; most senior executives of large listed companies don't earn as much as him!
Core engineers who can lead teams to tackle advanced technologies are very expensive!
When Xu Shenxue saw Yu Taihe's younger brother, he spoke to him for a while, saying that both brothers were quite capable... It was no easy feat for two brothers to become senior scientists.

Then, when Xu Shenxue was chatting with Yu Taihe's younger brother, he learned that not only the two brothers, but their sister was also quite accomplished. She had graduated with a PhD from a top-tier 985 university and was currently working at Yihai Technology Group as a virtual simulation algorithm engineer, earning an annual salary of tens of millions...

Hearing this, even Xu Shenxue, who was well-informed, couldn't help but sincerely admire the parents of these three siblings... Yu Taihe's younger brother said that his parents were genuine rural people, his father had only attended junior high school and his mother had only graduated from elementary school, so it wouldn't be an exaggeration to say that they were semi-illiterate!

The couple supported their three siblings' schooling by farming and doing odd jobs.

But it was from such a family, under such difficult circumstances, that Yu Taihe and his two siblings, three top talents, were raised...

This is very unscientific!
At the end of the meeting, Xu Shenxue was still moved and specially instructed the general manager's office to arrange for someone to visit the Yu family, meet the Yu parents, and present them with some small gifts to thank them for raising two or three such outstanding talents for him!

These were just minor incidents; Xu Shenxue only had a fleeting interest and forgot about them afterward.

But the secretaries in the general manager's office will not forget!
While Xu Shenxue was still on his inspection tour in Tongcheng, that very evening, a secretary from the general office of Liuhe Investment had already rushed from Shenzhen to Guangzhou with gifts.

The next morning, accompanied by several senior executives from the Guangcheng branch of Liuhe Investment Company, the secretary went to Yu Taihe's home to visit the elderly couple and presented them with gifts on behalf of Xu Shenxue...

This excited and moved the two elderly people... In their early years, they worked themselves to the bone, gritting their teeth as they farmed and did odd jobs, saving every penny to support their three siblings' schooling.

All of this effort paid off when the three siblings started working after graduation. They moved from their hometown to Guangcheng and moved into a villa worth hundreds of millions of yuan that Xu Shenxue had given to Yu Taihe, living a life that outsiders considered superior.

However, the fact that Xu Shenxue, the big boss, specially sent someone to visit them today made them even more excited!
(End of this chapter)