Genius is just the threshold for me!

The next day was the second day of the lunar new year.

Lin Dong slept until 8:30.

After getting up and having breakfast, Lin Dong pulled the curtains halfway closed, grabbed a bottle of milk, sewed them shut, and then settled into his laptop on the sofa to work intently on his design.

Crack—

At that moment, the door lock turned slightly.

Xu Qingling came in with the key, wearing a beige down jacket. Her cheeks were slightly flushed from the cold wind outside. After changing her shoes, she called out to Zhang Xiali and Lin Weiliang, who were cleaning up the dishes in the kitchen, and then ran to Lin Dong's side.

A faint, familiar fragrance wafted towards me.

Lin Dong snapped out of his reverie, looked up at Xu Qingling who was standing in front of him, and couldn't help but smile: "Why did you come today? Shouldn't I be the one going over there today?"

"What? You're supposed to go over today...?"

Xu Qingling paused for a moment upon hearing this.

But the next second, remembering it was the second day of the Lunar New Year, she immediately realized what Lin Dong was talking about, her ears turning red as she playfully scolded, "You're thinking too early! Don't think about it!"

"Okay, okay, I don't want to."

Lin Dong chuckled and reached for the milk on the coffee table.

Xu Qingling helped take it and handed it over, then sat down on the armrest of the sofa. Looking at the design diagram on the screen, which resembled a five-point harness and included torque formulas, she asked curiously, "What kind of design diagram is this? Were you busy this late yesterday, the first day of the Lunar New Year?"

"Unpowered exoskeleton armor"

Lin Dong took a sip of milk and didn't hide anything, saying, "Yesterday I didn't want to go out to visit relatives and friends. A bunch of uncles, aunts and uncles asked me about my grades, so I stayed home and did some chores."

Since we've already started, let's finish it first... How did you know? Are you so well-informed?

Yesterday he just told Xu Qingling that he was busy and not to look for him.

But he didn't tell Xu Qingling how late he was busy.

"Yesterday, Uncle Wang and Uncle Lin were playing mahjong together at my house. I heard you make a phone call to Uncle Wang..." Xu Qingling explained simply.

Ding-dong-

The two chatted for a while, when the doorbell rang.

Zhang Xiali went to open the door, and soon after, Wang Xingguo and Xu Ziming arrived.

"Lingling?"

Xu Ziming paused for a moment when he saw the shoes by the door.

When he came in and saw Xu Qingling sitting and chatting with Lin Dong, he asked, "Didn't you just say you were going to hang out with your classmates?"

"dad……"

Xu Qingling blushed and called out to Xu Ziming, but still calmly said, "Lin Dong is also my classmate."

"..."

Xu Ziming's eyes widened, and he suddenly fell silent, unsure how to answer.

"Haha, Qingling has grown a year older after the New Year, and she has made progress."

Wang Xingguo laughed a few times, enjoying the spectacle, and gave Xu Qingling a thumbs up.

Done.

Wang Xingguo glanced at the anti-eavesdropping device that had been on since the project started, walked over to Lin Dong and said confidently, "We've contacted the supercomputing resources. They're all on holiday for the Chinese New Year, so many supercomputers can't immediately stop their tasks and free up space."

However, headquarters helped coordinate, and the main staff of Tianhe-1A at the Tianjin Supercomputing Center were still working overtime. At 11 o'clock, the task there was temporarily suspended, and full performance could be freed up for 38 hours. After 38 hours, 45% of the performance could be freed up...

"Is only Tianhe-1A operating at full capacity for 38 hours?"

Lin Dong frowned, recalling the Tianhe-1A supercomputer data he had seen on the intranet.

The Tianhe-1 supercomputer at the Tianjin Supercomputing Center is currently the most powerful supercomputer in China, with a sustained performance of 2.57 pflop and a peak performance of 4.7 pflop.

If described in terms of tflops, which are more familiar to the general public, this means a sustained performance of 2570 tflops and a peak performance of 4700 tflops...

More than a decade later, the high-end computing chips used in China's high-end new energy vehicles and intelligent driving systems will have a computing power of at least 500 tflops per chip.

However, the two systems have different levels of precision. The intelligent driving chip uses a low-precision architecture of fp8 or int8, which has a relatively high fault tolerance and is only suitable for AI inference, not for scientific computing.

Tianhe-1A uses an FP64 double-precision architecture, requiring calculation results to be accurate to more than ten decimal places. It mainly serves various scientific calculations, such as simulating nuclear explosions and weather forecasting.

If the Tianhe-1A supercomputer were replaced with the FP8 low-precision architecture, its performance and computing power could be increased by another 10 to 15 times...

of course.

The cores of current supercomputer chips are all based on the Fermi architecture, which is a traditional precision computing core.

There is no tensor computation core adapted for the VoLTE architecture for AI operations, so AI cannot be directly accelerated.

Using a supercomputer for AI calculations at this point is somewhat like using a CPU to do GPU work; the efficiency will plummet.

The only solution is to use a special intelligent scheduling strategy.

By using a system-level "AI intelligence" strategist, massive traditional computing power can be mobilized, rather than directly using the performance of supercomputers for AI calculations.

In this respect, Lin Dong was already very proficient at using the servers in his own laboratory No. 78.

The field of computational chemistry requires precise calculations, and AI is used only for data processing and intelligent screening, which is also a good fit for the Tianhe-1A supercomputer.

"The problem is that while intelligent scheduling strategies are used to assist AI, current supercomputers are ultimately not suitable for AI computing, and there is a significant bottleneck in the communication bandwidth between chips."

Even if the code has reduced the number of inter-chip communication interactions as much as possible and increased the computing power of a single core, the entire supercomputer can only achieve 58% of its performance at full load.

With this scheduling algorithm that I personally wrote, the Tianhe-1 a2.57 Pflop sustained performance is only barely adequate…”

Lin Dong did a quick mental analysis and shook his head with a hint of regret.

Although he can perceive the flaws, defects, and perfection of things, and has full-fledged system and code optimization capabilities, he cannot break through the limitations at the core hardware level.

Historically, NVIDIA will launch the world's first V100 GPU in 2017, which uses the Tensor Core Volta architecture and is adapted to AI acceleration computing at the hardware level, thus ushering in the era of large AI inference models.

In this life, he intervened in the AI ​​intelligent model, allowing his uncle to make advance arrangements for computing power GPU chips, while also providing certain technical support.

However, to truly achieve a technological breakthrough and develop the first hardware-adapted GPU for AI inference using the Tensor Core Volta architecture, it is conservatively estimated that we will have to wait until 2014.

"So, is there anything wrong with this supercomputer?"

At this moment, Wang Xingguo noticed that Lin Dong hadn't replied and couldn't help but ask.

"Tianhe-1A's 38-hour full-performance test should be barely sufficient..."

Lin Dong nodded and said to Wang Xingguo, "Once all their missions are finished, just open an interface for me to connect to. If that's not enough, we'll talk about it later."

"38 hours is barely enough? This is the most advanced supercomputer in China right now."

Wang Xingguo's eyes widened, and he couldn't help but ask, "Xiao Dong, what are you going to use this supercomputer for? Even a nuclear explosion simulation wouldn't take this long, right?"

China's supercomputing resources are never idle; except for maintenance time, they are always in a queue working 365 days a year.

He has now asked headquarters to coordinate 38 hours of full-performance time for the Tianhe-1A supercomputer, which he originally thought would be enough for Lin Dong to squander, but in the end he could only get a barely sufficient explanation.

"Using computational chemistry to study batteries."

Lin Dong briefly explained the matter that his superiors had asked him to discuss, then took out several documents and handed them to Wang Xingguo, Xu Ziming, and Lin Weiliang, who had come over.

"This is a plan I came up with last night. As long as we solve this battery problem, we can revitalize the entire electronic equipment and new energy vehicle industry."

Add to that Xingdong Company's drone industry and its ongoing development of computing chip industry, and it's highly likely that this will be enough to meet the economic drive for the entire industrial upgrade.

"Based on infrastructure development for people's livelihood, with new industries as the core?" Wang Xingguo took the document and glanced at it, his brows furrowing immediately.

"Building a foundation for major infrastructure projects that benefit people's livelihoods, such as roads, water conservancy, highways, railways, and logistics, while leveraging breakthroughs in batteries to drive the entire new energy vehicle industry chain and the power industry chain, and simultaneously enhancing the competitiveness of consumer electronics and improving the entire consumer electronics industry chain..."

Xu Ziming also read it carefully, then shook his head and said, "This plan is too difficult; the local area simply can't handle it."

"Xiao Dong, this is really too difficult."

Lin Weiliang glanced at it and nodded: "Compared to the regional applicability of real estate, these new industrial plans can only be applied to a small number of regions, and are not suitable for most regions."

Although they weren't involved in this planning, their vision and level of expertise allowed them to see the shortcomings of this plan.

That is, its adaptability is too poor, its difficulty is too high, and it is not as simple and effective as the real estate model.

After all, real estate is built and then sold. The deed tax, land transfer fee, local employment, local tax revenue, and other expenses incurred during this process can directly stimulate the local economy.

While this infrastructure project for people's livelihood can revitalize some jobs, the more important taxes such as land transfer fees and local tax revenues are gone.

As for the new industrial plan, it is only suitable for a small number of regions, not for all regions.

"For engineers Lin and Xu, issues related to industry compatibility and local conditions can be resolved through regional policies and transfer payments. It's a bit more difficult, but it can still be accomplished."

At this moment, Wang Xingguo put down the document, looked at Lin Dong seriously and said, "The biggest problem with this plan is not that it is difficult, but that it requires external participation. Our own internal economic driving force is not enough."

"My godfather has a keen eye."

Lin Dong gave a thumbs up and nodded, saying, "Indeed, the biggest problem with using industry as a driving force is insufficient domestic demand. We need external participation; relying solely on our internal resources, we cannot achieve the same results as the real estate model."

The biggest advantage of the real estate model is its wide applicability; it can be implemented as long as there are people and land.

Secondly, the industrial chain is large in scale, with upstream and downstream, as well as indirect and direct supporting industries, covering almost the entire industry.

The last advantage is that it doesn't rely on external sources; it can spin on its own internally, and it can spin as fast as a spinning top can.

New industrial models depend on technology, policies, talent, and external participation; internal transformation alone cannot reach its limit.

Because of a very simple concept, people can accept taking out a 30-year loan to buy a house and then spending three years' income on renovations, furniture, and appliances, but they would never take out a 30-year loan to buy a car and spend three years' income on modifications.

This means that the economic effect that the industrial model can generate internally is only half or even less than that of the real estate model.

The solution is either to go abroad, reap the benefits from overseas, and fill the economic gap.

Either establish industries with higher added value, increase the profit margin and income ratio, and then there will naturally be enough economic surplus to drive industrial upgrading.

The former means that external factors will not give you such an opportunity.

Low-end industries with low profits are fine, but as soon as you have the idea of ​​getting involved in high-end industries to earn more profits, all sorts of methods will immediately come your way.

The latter brings us back to the chicken-or-egg question.

Without industrial upgrading, the industry is still focused on low-end industries, which prevents the increase of industrial added value and allows foreign countries to take the lion's share of the profits.

For industry to upgrade, someone needs to get on the front lines, provide the economic support, and fill the gaps in the industrial upgrading process...

"I read the think tank's report yesterday and studied the domestic industry analysis model."

Lin Dong said seriously, "We are currently upgrading our entire industrial system, and there is still a huge gap in the research and development and production of military equipment."

However, based on the 2016 timeline of the industrial upgrading plan, the economic effect of increasing infrastructure investment in people's livelihood across the country could cover approximately 25% to 30%.

The real estate model is feasible, but it shouldn't be pursued on a large scale; based on the existing benchmark, it can cover 10% to 15%.

Breakthroughs in basic batteries have directly transformed the new energy vehicle industry chain, the energy storage industry chain, and elevated the consumer electronics industry chain to a new level… This can cover 20% internally.

The normal development and improvement of other industries such as photovoltaics and biomedicine can cover 10%, leaving a 30% economic gap.

If Xingdong Company achieves a breakthrough in AI computing chips and AI hardware develops, coupled with the large-scale artificial intelligence model technology that my father currently possesses, this ultra-high value-added industry could cover 10% of the industrial upgrading plan by 2016...

This leaves us with approximately a 20% external gap.

If we have enough time, we can wait another three years until 2019. The AI ​​and data industries will further improve, as will other industries. We can accomplish this internally without needing external support.

The real estate economic model was actually naturally driven by the four trillion yuan stimulus package as early as 09. However, it was initially considered as an alternative rather than a primary option, so the impact was minimal.

After doing this for several years, it became clear that new industries such as photovoltaics, new energy, and energy storage were completely lagging behind, and the country's economic size simply could not support industrial and military upgrades.

Finally, the shantytown renovation project was finalized in 15 and officially launched in 16, marking the beginning of a comprehensive acceleration of the real estate economy.

This booming real estate economy has also supported a brand-new upgrade in military equipment research and development.

The Fujian aircraft carrier, various new fighter jets, and hypersonic aircraft were all accelerated in terms of project initiation, advancement, and construction in 15 and 16.

"Our industrial scale will either have a 20% external shortfall, or it will be delayed for another three years..."

Wang Xingguo took a deep breath and said solemnly, "The outside world won't let us develop so peacefully."

"Then we'll have to fight it out."

Lin Dong smiled and said, "Military strength is the foundation of everything, and military equipment is the source of all confidence."

Although we are not yet very confident, and not many aircraft carriers have been launched yet.

But the WZ-8 can now target aircraft carriers, the DF-21D has been developed, and soon the J-20 equipped with the Emei engine will also be able to take to the skies.

I think it should be able to support this 20% shortfall, or even support development for another three years.

Lin Dong knew, of course, that the outside world wouldn't give him a chance.

Don't say that the Americans are still tough guys right now. With the US as their big brother leading the way, they are at the height of their arrogance.

Even if ten years later, when the US has recovered its strength and starts relying on its allies for resources, they still don't want to see the rise of high-end industries in China.

On several occasions, they have disregarded world trade regulations and used various pretexts to prohibit the global expansion of domestic photovoltaic, new energy, and other industries.

It wasn't until they flexed their muscles a few times that these guys reluctantly accepted it.

"That's true."

"You only know something after you've done it; no one will be convinced unless you try."

Wang Xingguo nodded seriously, looking at the planning documents in his hand: "Let me give the report. Xiao Dong, you put together a detailed report, and I will personally go up and give it to you in person at the beginning of the year."

In this real estate market, it's best to avoid making heavy bets if possible; there are already many precedents abroad…

The effects and drawbacks of the real estate economic model have been the subject of numerous studies and analyses worldwide over the past fifty years.

If there's no other choice regarding this economic model, then there's nothing that can be done; but if there were a choice, no one would choose it.

Even seasoned civil engineers in the industry wouldn't be willing, since not pursuing a real estate-based economic model doesn't mean not building houses, it just means not using it as the main economic engine.

Everyone can develop new industries, new technologies, and new productivity, and civil engineers won't have to build a bunch of high-rise buildings they can't afford. This is obviously more attractive and appealing.

"No problem, I'll work on this when I have some free time after I finish this design." Lin Dong nodded.

Such a report definitely needs to be detailed and presented in person. This matter can only be handled by Wang Xingguo, who is still a middle school student.

"Xiao Dong, what kind of design are you working on?" Xu Ziming asked curiously when he heard Lin Dong designing something.

"I'll make an exoskeleton for fun. Once battery technology breaks through, this thing will be usable," Lin Dong answered without hiding anything.

Exoskeleton armor?

Army?

Lin Weiliang glared at him from the side, and the document in his hand almost slipped from his grasp.

Seeing that Lin Dong didn't seem to be joking, she immediately put down the documents and called into the room, "Xia Li, it's the second day of the Lunar New Year today. Let's go to Mom and Dad's place and stay for a while before coming back..."

"Didn't you say you wouldn't go this year?"

Zhang Xiali came out of the room: "Isn't this a critical time for the project? Do we have to stay at the base? Can we go back?"

"We're not even core members of the project team..."

Lin Weiliang hurriedly said, "Let's go, let's leave now, or we'll miss lunch. Xiao Dong said he's busy with things at home and won't be going this year..."

???

Lin Dong stared wide-eyed at Lin Weiliang in disbelief. When did he ever say he wouldn't be going to his grandparents' house this year?