The Supercomputer Was Taken Care Of (Part Two)
"Oak Technology?"
Inside the ward.
Upon hearing the name mentioned by Academician Pan, Xu Yun was immediately taken aback.
It's this company?
After receiving the reward of the Supercomputer module, Xu Yun has been eager to realize it as soon as possible.
Furthermore, Huadun Biotech's development does indeed require a Supercomputer, especially now that Lu Xiao has arrived, this need has become even more urgent.
Therefore, Xu Yun, or rather the company's Board of Directors, had conducted in-depth research on the Supercomputer industry over a period of time.
As mentioned before.
To date, 80% of the Supercomputer in China on the Top 500 list are from the Liu family.
To be honest, this is a very alarming number, bordering on monopolistic.
However, on the other hand, although the Liu family accounts for more than 80% of the Supercomputer production, 20% of the Supercomputer still come from other companies.
This 20% of Supercomputer capacity is distributed across the country, and regionally it can be categorized into three provinces or municipalities:
Beijing, Guangdong Province, Sichuan Province.
Oak Technology is a Supercomputer manufacturing company located in Sichuan.
It's quite a coincidence.
Its owner, Guo Hao, is also an alumnus of the University of Science and Technology of China Gifted Class, and one of the earliest batches of graduates. Although University of Science and Technology of China established the Gifted Class in 1978, the computer software major was not officially launched until 1984.
From a certain perspective...
Guo Hao was one of the first professional computer scientists in Huaxia, and he was even more senior than Xiao Rong.
However, after graduating, Guo Hao did not choose to stay and become a programmer or hacker. Instead, he went into business and started his own company, tinkering with communication equipment.
Then came two or three decades of ups and downs. In 2019, Guo Hao and several partners founded Oak Technology.
Oak Technology primarily provides customized commercial computer services, and its target customers are clearly defined:
We don't work on top-tier Supercomputer; we only focus on the mid-tier and lower-tier markets, specifically Supercomputer ranked 300-1000 globally.
Meanwhile, Guo Hao has made significant achievements in storage and access models and parallel programs, and can even achieve higher vector processing results than Liu family at lower precision.
Therefore, although Oak Technology has not reached the top position in the industry in recent years, it has actually been Little Japan a pretty comfortable life.
Take last year as an example.
They produced two Supercomputer last year, both of which were mid-to-low-end machines ranked between 800 and 1000 meters. One was supplied to China, and the other to A-san.
The total price of these two Supercomputer is approximately 200 million yuan, with a profit of 50 to 60 million yuan.
After deducting company operating expenses and employee bonuses, Guo Hao's annual income is still over eight figures.
Thinking about this...
Xu Yun couldn't help but glance at Academician Pan beside him.
He had mentioned the Supercomputer issue a few days earlier when he welcomed Lu Xiao, and Supercomputer was indeed on the company's agenda.
Because even though a Supercomputer costs hundreds of millions of Huaxia currency, at first glance it seems that Huadun Biotech simply cannot come up with that amount of money in a short period of time.
But in reality, this thing is just like home renovation.
The renovation costs don't need to be paid in full upfront, and the renovation team won't start work today and finish tomorrow; it takes a long, phased process.
Therefore, as long as the company puts up a deposit of 20 to 30 million, the manufacturer can start production.
The entire production and assembly process will take anywhere from six months to one or two years – even at the fastest possible time, the company will have the ability to pay the full amount by then.
Therefore, what stood in Xu Yun's way during this period was not the issue of money, but rather the choice of partners.
In fact.
As a domestic Supercomputer manufacturer with a fairly good reputation, Oak Technology was originally on Xu Yun's or rather the Board of Directors' planned contact list.
Unexpectedly, before the inquiry letter was even sent out, Academician Pan took the initiative to make connections.
This action was clearly telling Xu Yun something:
We're on the same team, so there's a discount!
Just as Xu Yun was slightly lost in thought...
Guo Hao, who was on the bed, came to his senses first. He let out a rather exaggerated "Ah!" and quickly sat up straight.
"Oh dear, what's wrong with me? I was so focused on greeting Old Pan. Come, come, Dr. Xu, please have a seat, please have a seat."
Judging from Guo Hao's posture, if it weren't for the IV drip in his hand, he probably would have gotten out of bed himself to move the stool for Xu Yun.
Then Academician Pan gave Xu Yun a look, signaling him to sit down.
Xu Yun obediently did as instructed.
Ignoring Xu Yun's existence beforehand put Guo Hao in a passive position from the start, and Academician Pan's role as an intermediary also meant that he had to make some concessions on the price.
Therefore, Guo Hao didn't bother with any pretense and asked directly:
"Dr. Xu, judging from what Old Pan said, your company is planning to customize Supercomputer?"
Xu Yun nodded, scratching his head with a simple and honest expression.
"That's right, so when I heard that the teacher was coming to visit you today, I shamelessly followed along."
Upon hearing this, Academician Pan secretly gave a thumbs up.
Although he didn't mention Guo Hao's affairs to Xu Yun from beginning to end in an effort to avoid suspicion, Guo Hao was unaware of this situation.
With Xu Yun saying this, Guo Hao would have to owe Academician Pan a favor.
After thinking for a moment, Guo Hao asked Xu Yun:
"Dr. Xu, what specifications of Supercomputer does your company want to customize?"
"Specifications, huh?"
Xu Yun looked up at Guo Hao, a hint of curiosity on his face, and said:
"Mr. Guo, could you please explain what the specific specifications mean?"
Guo Hao nodded; this is indeed one of the most crucial business modules in custom Supercomputer.
"Supercomputer have many specifications, but when it comes to detailed data, they can be simply encompassed by two attributes."
"The first aspect is the computing power of Supercomputer. Currently, based on their computing power, Supercomputer can be divided into Exascale, PetaFLOPS, and P-scale."
"The 'E' here stands for Exascale, which is 100 quadrillion operations per second."
"This category didn't actually exist in the industry last year, but this year, with Frontier becoming the world's first supercomputer to break through the exascale computing bottleneck, the E-era has officially arrived."
"Furthermore, above Exascale is Zettascale, which is the zettascale level, in the tens of trillions, and belongs to the next stage of goals."
Guo Hao paused, then lowered his voice and added:
"Of course, even with the addition of the Exascale category, our company still can't produce Supercomputer of that level. We keep this business mainly to maintain appearances."
Xu Yun and Academician Pan smiled wryly.
This kind of thing is not uncommon in other industries. For example, the bicycle tire repair shop downstairs from Xu Yun's house claims to be able to completely modify a Bentley.
Guo Hao then coughed lightly, and his complexion returned to normal.
"And PetaFLOPS computer refers to a petaflop, which is ten quadrillion floating-point operations per second."
"Currently, almost all Supercomputer on the Supercomputer ranking list are in the PetaFLOPS. For example, the top 300 is 2.3P, the top 500 is 1.6P, and the top 1000 is around 0.8P. Anything below 1P is what I call the P-level."
"Based on this classification, we can then introduce the second attribute I mentioned, which is power consumption."
After speaking, Guo Hao thought for a moment and took out a laptop from beside him.
After opening the laptop and fiddling with it for a few moments, he pushed the screen towards Xu Yun:
"Dr. Xu, this is the corresponding power consumption for the computing power—this is generally not disclosed to the public, so please keep it confidential."
Thank you.
Xu Yun thanked Guo Hao and took the notebook to read.
"Exascale energy consumption: 50 megawatts per hour"
"A 2.2P unit consumes 6 megawatts per hour, which is equivalent to 6000 kilowatt-hours of electricity."
"0.8P generates 2 megawatts per hour, or 2000 kilowatt-hours of electricity."
The data provided by Guo Hao actually includes not only power consumption, but also the footprint of the water cooling equipment, PUE value, and so on.
This figure is about 10% higher than the data Xu Yun had learned. After all, Oak Technology is not a leading Supercomputer manufacturer; its main focus is on overall cost-effectiveness.
The unit price of electricity for top- Supercomputer may indeed be lower, but this lower price is predicated on the fact that their consumption base is also large.
Equipment that consumes 50,000 to 60,000 kilowatt-hours of electricity per hour will certainly have a much lower unit price than equipment that consumes 2,000 to 3,000 kilowatt-hours—but such a low price is clearly not something Xu Yun needs for the time being.
Soon after, Xu Yun's gaze settled on one of the recommended meal packages.
This package is a bundled deal, and it's labeled with the words "Year-end loss-making promotion".
The package has a computing power of 1.3P, a PUE value of 1.22, and is configured with 5×5 tiles.
Each of these tiles has a bandwidth of 4.5TB/s, and each module also has a heat dissipation cover with a capacity of 15kW, minus the heat dissipation for 40 I/Os.
This Supercomputer uses 1023 computer nodes, employing a heterogeneous computing model of CPU + GPU.
The interconnection is via 40G Aether, and the water-cooled unit occupies 500 square meters.
The customized price of the entire Supercomputer, excluding the CPU and GPU, is 80 million Huaxia currency. If a CPU or GPU is included, the price will increase significantly.
For example, after configuring it with a Xeon E5-2673v4 20C 2.3GHz, the customized price of the Supercomputer immediately increased to 150 million yuan.
The highest configuration in this segment is the AMD EPYC 7713 64C 2GHz, priced at 230 million.
Then Xu Yun turned to look at Academician Pan, and under Academician Pan's "you decide for yourself" gaze, turned the display screen towards Guo Hao:
"Mr. Guo, what's the lowest price you can offer for this configuration?"
"This set?"
Guo Hao wasn't surprised by Xu Yun's choice; this was basically their best-selling machine.
"Dr. Xu, I wonder if you're planning to choose a different CPU and GPU?"
Xu Yun gently shook his head:
"We plan to purchase this part ourselves."
Xu Yun and Chen Yunji had already made contact when they were welcoming Lu Xiao.
Chen Yunji's Sinica Cambrian is a chip manufacturer. As one of the companies affiliated with the USTC system of China, it is almost impossible to find a supplier in China with a lower price than Cambrian.
Admittedly, for ordinary companies, Cambrian's chips may have a significant gap or risk compared to top-tier chip suppliers.
But don't forget.
With the computing power template from the halo reward and the AI Mimi at his disposal, Xu Yun has absolutely no reason to worry about this issue.
Seeing that Xu Yun's attitude was quite firm, Guo Hao gave up on continuing to promote the product and focused all his attention on the price.
"Dr. Xu, to be honest, this package is already very cheap; our company doesn't make much profit from it."
"How about this, 7777.7, what do you think of this price?"
Guo Hao's expression seemed sincere, but Xu Yun didn't buy it at all.
Joke.
These days, from street fruit stalls to real estate sales, even young women with large breasts are constantly complaining about not making a profit. Who believes that?
However, regardless of whether he believed it or not, Xu Yun's experience in bargaining was indeed not very extensive.
If you cut too little, Guo Hao might agree in the next second, leaving you no chance to take back your move.
Since that's the case, why not cut a little more?
So he gritted his teeth and said a number:
"Mr. Guo, how about this: 59.99 million? If that's acceptable, I'll sign the contract right away!"
Guo Hao:
"(╯‵□′)╯┻━┻?????"
If Academician Pan hadn't been there, he would have been ready to flip the table right now.
Do you really think this is a computer mall with a profit margin of 100% or more?
There are only about five or six core profit points for Supercomputer. If the CPU is purchased by the company itself, a profit margin of 20% for a Supercomputer is considered good.
Firstly, because Oak Technology is not a purely upstream enterprise, it also needs to purchase many components from suppliers, which incurs some expenses.
Secondly, the companies that purchase commercial Supercomputer are not complete computer novices; they are all high-tech companies, and each of them knows the pricing inside and out.
If you treat customers like guinea pigs to be slaughtered in a computer mall, they'll turn around and leave immediately.
Therefore, the price quoted by Oak Technology —even the initial price—is quite reasonable.
The profit margin they presented to the client was only about 15%, and with Guo Hao's expertise in database access models and parallel programming, they could free up a few more points, and that would be about it.
In other words...
Supercomputer power rating of 80 million would cost approximately 64 million.
Therefore, Guo Hao's target price is roughly around 73 million, with a profit of about 10 million.
7777 is a strategy of taking a chance, even if there's no guarantee of success. If Xu Yun falls for it, he'll make a big profit; if not, he can negotiate the price better.
Unexpectedly, Xu Yun's reaction was no longer a question of whether the trap had been set or not; he snatched the sack, turned around, and was about to put the Supercomputer in its trap and carry it away.
This is pure robbery!
In fact.
Guo Hao wasn't the only one who couldn't hold back; Academician Pan, standing beside him, was the same.
Xu Yun's move wasn't bargaining, it was outright ripping people off!
Then Academician Pan took a deep breath, calmed the smile on his lips, and took the initiative to speak:
"Alright, Old Guo, Little Xu isn't very good at bargaining, so please bear with him."
"I think you two should stop arguing about 7777 or 5999. I'll suggest a number: 6999, how about that?"
"6999?"
Guo Hao frowned, clearly displeased.
Although he had a close relationship with Academician Pan and was prepared to make concessions, the discount of several million yuan was still beyond his expectations.
Cutting the price from the bottom line of 7300 to 6999 is no longer just about giving face to a friend.
Moreover, although he is the largest shareholder of the company, the other partners also hold significant shares, and he must consider their opinions as well.
Seeing that Guo Hao's attitude was somewhat conflicted, Academician Pan thought for a moment and suddenly said:
"Old Guo, how about this, let me ask you a question first."
"If you agree to this price, I promise that within two weeks, your company can attend the launch event of a Nobel Prize-level achievement as a sponsor. Are you willing?"
"?"
A clear question mark appeared on Guo Hao's face as he looked at Academician Pan as if he were a fool:
"Old Pan, are you alright? The bed next to me is empty. Why don't you get admitted and we can be fellow patients?"
To everyone's surprise, Academician Pan's expression remained unchanged, still very serious.
"Old Guo, I'm not joking."
Guo Hao immediately frowned.
He then stared at Academician Pan for a long while, still harboring a series of questions in his mind, but still honestly answered:
"Of course I'd be willing. Only a fool wouldn't do something like that, but is it even possible? It's Nobel Prize-worthy."
Before Guo Hao could finish speaking, Academician Pan swiftly and decisively interrupted him:
"Alright, since you're willing, let's sign the contract."
".?"
Guo Hao's last sentence caught in his throat, and a look of existential doubt crossed his round face.
He was just like many anime fans who flood the screen with comments like "[Did I miss an episode?]" because a new episode was told from a different angle, looking completely bewildered and confused.
A full half minute passed.
Gurgle.
Guo Hao swallowed hard and rolled his eyes twice:
"Old Pan, are you telling the truth?"
Academician Pan looked away, glanced at the sky outside the window, and said casually:
"Old Guo, there are some things I can't say too much about. Whether you believe me or think I'm just fooling you, this is your only chance."
"Also, I want to remind you that the above agreement definitely can't be written into the contract. In other words, if I go back on my word after signing the contract, you still have to provide Little Xu with the machine at that price."
"So, you can also think I'm lying to you and refuse this request."
"Then let's pretend this never happened. Today is just a simple reunion of old friends. I'll ask Little Xu to go downstairs and buy some lunch later."
"Little Xu and I ate spicy hot pot with Beef tallow, while you ate steamed Steamed Bun with Want Want Milk."
"."
Guo Hao's expression shifted inexplicably as he watched Academician Pan speak eloquently.
To be honest, Academician Pan's appearance today was too sudden. As a result, he was on the defensive the whole time and had no chance to verify whether what Academician Pan said was true or false.
If there really is an opportunity to attend a Nobel Prize-level results announcement, then this three million discount is definitely worthwhile.
For example, when Andrew Wiles was giving a lecture tour on Fermat's Last Theorem, the title sponsorship of the Princeton lecture cost as much as $400,000, which was won by Coca-Cola.
That was $400,000 in the last century, and that was before the era of internet live streaming.
The record for naming rights in the internet age was set by Takaaki Kajita in 2016.
The naming rights for his lecture at the National University of Singapore were acquired by Singtel SGX for Singapore $2 million, with 60% going to Takaaki Kajita.
It's worth noting that neither of these lectures was the first time the research findings were presented.
Therefore, if a Nobel Prize-level discovery is indeed published, then spending those millions would definitely be worthwhile.
But the problem is
Was Academician Pan telling the truth?
The direction in which Academician Pan was most likely to win the Nobel Prize has already been explored by his teacher, Anton Zeilinger. Theoretically, even if Academician Pan achieves 18-photon entanglement, it is almost impossible for him to win the Nobel Prize.
From a logical standpoint, the probability is indeed very low.
However, as a longtime friend of Academician Pan, Guo Hao did not believe that Academician Pan would try to fool him with this matter.
Meanwhile, Academician Pan, who should have been in Hefei, suddenly came to Rongcheng today.
A multitude of pros and cons flashed rapidly through Guo Hao's mind, the two sides of a scale that did not exist in reality constantly rising and falling.
That's all.
A full minute passed.
Guo Hao finally made up his mind.
He took a deep breath and looked up at Academician Pan:
"Old Pan, I'll trust you this once. Let's sign the contract."