94: Chapter 94 War of Attrition and Humanized Management
Shenzhen, Xingchen Academy of Sciences EDA Task Force Laboratory.
Time: August 29th, 1:47 AM.
Only three sounds remained in the lab: the low hum of server racks, the rustling of whiteboard markers across the surface, and an occasional cough—seventy-two-year-old Li Weimin had been working continuously for seventeen hours.
The three whiteboards in front of him were filled with mathematical symbols and circuit diagrams. On the board to his left was the traditional EDA flow: Logic Design -> Logic Verification -> Synthesis -> Placement -> Routing -> Timing Verification -> Physical Verification -> chip Production. This was a linear process, requiring an average of fifteen iterations to converge.
On the board to his right was his new concept: Timing Constraint-Driven Co-design. It introduced physical layout estimations at the conceptual stage, allowing logic design and physical design to proceed in parallel with mutual feedback.
But the problem was stuck at the core of the algorithm—how to build a mathematical model that could accurately estimate timing?
Li Weimin had already tried seven models, all of which failed. Either the estimation error was too large, or the computational complexity was too high to have practical value.
He took two steps back, rubbing his aching eyes. Under the cold white lights of the lab, the old Academician's figure appeared exceptionally frail. His white coat hung loosely on his body, and behind his thick glasses, his eyes were bloodshot yet still sharp.
"Academician Li, you should rest," the young assistant engineer, Xiao Zhou, said cautiously. "It's almost two o'clock."
"You go back first," Li Weimin replied without turning his head. "I'll think a bit more... I feel like I'm just one thin layer away from the answer."
Xiao Zhou didn't dare to persuade him further. He knew Academician Li's habits—he wouldn't leave the lab until the problem was solved. He quietly went to the pantry, mixed a cup of protein powder, and placed a pack of biscuits by Li Weimin's hand.
2:23 AM.
Li Weimin suddenly froze. He stared at a formula on the whiteboard, a differential equation he had scribbled down three hours ago that seemed unrelated to the current problem—it described the boundary layer effect in fluid mechanics.
Boundary layer... timing constraints... physical layout...
A bolt of lightning struck his mind.
"No! We've been wrong all along!" Li Weimin's voice exploded in the empty lab. "Timing constraints aren't boundary conditions; they're driving conditions! We shouldn't check the timing after the design is finished; we should use the timing requirements to drive every step of the design!"
He practically lunged at the whiteboard, frantically erasing the original derivations as new formulas poured out like water:
"Assume the chip is a multi-dimensional state space, where each logic unit is a state point, and timing requirements define the 'gravitational field' between state points... Physical layout isn't about finding the optimal solution, but about finding the stable equilibrium point within this gravitational field... Logic design and physical design are no longer sequential; they are two dimensions of the same optimization problem..."
The pen tip flew across the whiteboard, making rapid rustling sounds. Mathematical symbols, circuit symbols, and algorithm pseudocode intertwined, gradually forming a complete theoretical framework.
Xiao Zhou watched in a daze. Although he was only an assistant engineer, as a graduate of Tsinghua University's microelectronics program, he could understand—this was a revolutionary shift in thinking. If implemented, the entire chip design flow would be completely restructured.
4:05 AM, the final stroke was made.
Li Weimin stepped back, looking at the three filled whiteboards, and let out a long breath. Then, without warning, his body swayed, and he collapsed to the side.
"Academician Li!" Xiao Zhou rushed forward to support him.
Li Weimin waved his hand, his face pale but his eyes bright: "I'm fine... low blood sugar. Give me some water."
He sat in a chair, downed a whole glass of glucose water in one go, rested for five minutes, and then suddenly stood up: "Xiao Zhou, wake up everyone in the algorithm group. Now, immediately! I want to verify this model!"
"Now? At four in the morning?"
"Right now! Inspiration doesn't wait!"
Twenty minutes later, six bleary-eyed algorithm engineers were urgently summoned to the lab. When they finished listening to Li Weimin's explanation, all traces of sleepiness vanished instantly.
"This... this overturns the entire theoretical foundation of EDA!" The group leader, Liu Yang, a twenty-nine-year-old genius engineer, spoke with a trembling voice. "Academician Li, if this 'Timing Gravitational Field' model holds true, all our existing tools will have to be rewritten!"
"That's why I want you to verify it now," Li Weimin's eyes were glowing. "Use the revised design of 'flint no. 2' for testing. The traditional flow currently takes seventy-two hours for one run. With my new algorithm, the theoretical estimate is twenty-four hours, and the timing will be even better."
"But implementing the algorithm takes time..."
"I've already written the pseudocode for the core parts," Li Weimin pulled up a computer document. "Divide into three groups: Group one will implement the Timing Gravitational Field generation algorithm, Group two will implement the logic-physical collaborative optimizer, and Group three will build the verification framework. I want to see preliminary results within eight hours."
"Eight hours?!" someone exclaimed.
"This is a war, not academic research," Li Weimin's voice was resolute. "Spark can't wait, and Huaxia's semiconductor industry can't wait. Let's begin."
The lab lights were as bright as day. Six engineers plus Li Weimin, seven people in total, began a race against time.
The sound of typing was as dense as a rainstorm. On the screens, the lines of code grew rapidly. The coffee machine worked non-stop, and empty cans quickly filled the trash bins.
7:00 AM, Group one reported: "Timing Gravitational Field generation algorithm complete! Preliminary tests show calculation speeds seventeen times faster than traditional methods!"
10:00 AM, Group two reported: "Collaborative optimizer core module complete! But the convergence still needs adjustment..."
Li Weimin took charge personally, checking the code line by line and pointing out issues: "Here, the step size for gradient descent should be dynamically adjusted. Use a smaller step size where timing constraints are tight and a larger one where they are loose."
2:00 PM, Group three reported: "Verification framework set up! We can start the full-process testing!"
"Load the 'flint no. 2' data!" Li Weimin ordered.
The server cluster began running at full capacity. On the screen, the progress bar advanced slowly but steadily.
The waiting time was the hardest to endure. Some fell asleep at their desks, others mechanically stuffed bread into their mouths, but Li Weimin remained standing before the large screen like a statue.
6:00 PM, progress at 70%.
9:00 PM, progress at 90%.
11:47 PM, the progress bar reached the end.
"Done!" Liu Yang shouted.
Everyone gathered around the screen. A report page popped up:
[New EDA Algorithm Test Report · flint no. 2 Revised Design]
Traditional Flow Results:
Total Time: 72 hours 18 minutes
Timing Convergence: 89.7%
Longest Path Delay: 1.23ns
Estimated Total Power Consumption: 2.1W
New Algorithm Results:
Total Time: 23 hours 42 minutes (67.1% reduction)
Timing Convergence: 96.3% (6.6% increase)
Longest Path Delay: 1.11ns (9.8% reduction)
Estimated Total Power Consumption: 1.9W (9.5% reduction)
Conclusion: The timing-driven collaborative design algorithm has passed verification. Efficiency has increased by 3.05 times, and design quality has improved across the board.
A deathly silence filled the lab.
Then, a burst of wild cheering erupted. Some jumped for joy, some hugged, and some simply cried. Liu Yang rushed to Li Weimin, so excited he was incoherent: "Academician Li... you... you've made history! This is absolutely the most significant breakthrough in the EDA field in the last ten years!"
Li Weimin didn't smile. He stared at the data on the screen, checked it three times, and then said softly, "It's still not good enough."
"What?"
"The timing convergence should be over 98%. Power consumption can be reduced by another 5%." Li Weimin turned to look at his exhausted but excited team. "But achieving this result in the first verification proves the approach is correct. Over the next two months, we need to refine this algorithm to an industrial-grade level."
He looked at the clock on the wall: August 30th, 12:12 AM.
"That's all for today. Everyone, go back and sleep. Tomorrow... no, we'll continue at 2:00 PM today."
"What about you?" Xiao Zhou asked.
"I'll organize the data a bit more." Li Weimin sat back down at the computer.
No one dared to persuade him. Everyone left the lab one by one, and each person couldn't help but look back as they went out—at the seventy-two-year-old man, back held straight, focused on his work in the light of the screen.
At that moment, they understood the meaning of "a scholar of unparalleled caliber."
In Shenzhen on August 30th, the weather was unusually sweltering.
The intensity of the EDA Task Force's work had lasted for twenty-three days. Since Su Chen issued the military order to "conquer EDA in ninety days," this 120-person team had entered a wartime state. At first, it was voluntary overtime, then it became the norm, and later, many people simply lived in the campus dormitories.
10:07 AM, Algorithm Group workstation area.
Twenty-nine-year-old Liu Yang was debugging code. He had been working for thirty-six consecutive hours, with only a three-hour nap at his desk in between. The characters on the screen began to blur and double. He rubbed his eyes and picked up his fifth cup of coffee.
"Brother Liu, you don't look so good," a colleague nearby whispered.
"I'm fine, just a bit dizzy," Liu Yang's voice was faint. "This module is almost tuned..."
Before he could finish, he suddenly slumped and slid off his chair. His head hit the corner of the desk with a dull thud.
"Liu Yang! Liu Yang!"
His colleagues rushed over. Liu Yang had lost consciousness; his face was pale, and his breathing was rapid.
"Call an ambulance! Quick!"
Ten minutes later, the doctor from the campus clinic arrived. Preliminary diagnosis: fainting caused by low blood sugar, but with an abnormal heart rate, he needed to be sent to the hospital. The sound of the ambulance siren pierced the campus's tranquility.
When the news reached the lab, Li Weimin was struggling with a line of formulas on the whiteboard. Xiao Zhou rushed in: "Academician Li, Liu Yang collapsed and has been sent to the hospital!"
The pen dropped from his hand, leaving a long mark on the whiteboard.
"What's the specific situation?"
"The doctor said it was low blood sugar and excessive fatigue, but the EKG showed an irregular heartbeat, suspected to be mild myocarditis."
Myocarditis. Those words made Li Weimin's heart tighten—that could be fatal.
"I'm going to the hospital." He grabbed his coat.
"Academician Li, you haven't closed your eyes for twenty-four hours..."
"I said, I'm going to the hospital."
At 3:00 PM, a scream came from the women's restroom.
Thirty-one-year-old verification engineer Zhang Yue was collapsed in a stall, with a pool of blood beneath her. She was semi-conscious when she was rushed to the hospital. The test results shocked everyone: eight weeks pregnant, threatened miscarriage. And she herself had no idea she was pregnant before she collapsed.
"The patient is severely anemic, with hemoglobin at only 8 grams, far below normal," the doctor said with a grave expression. "The work intensity was too high, she was malnourished, and the early pregnancy symptoms were mistaken for fatigue... Whether the baby can be saved depends on the next forty-eight hours."
When Zhang Yue's husband arrived at the hospital, the thirty-five-year-old man crouched down outside the ward, covered his face, and wept silently. They were college classmates and had been married for six years; this was their first child.
9:23 PM, the third incident.
Thirty-five-year-old Physical Design Group Leader Chen Hao was driving home. He had worked for fourteen consecutive days, averaging less than five hours of sleep per day. At a traffic light on Bao'an Avenue, he felt his eyelids were too heavy to lift.
"Just closing my eyes for three seconds..." he told himself.
Three seconds later, the car hit the guardrail. The airbag deployed, and Chen Hao's head took an impact, knocking him unconscious on the spot.
When the traffic police arrived, they found energy bar wrappers, empty coffee cans, and an unfinished design document in the car.
11:00 PM, in the emergency department of Shenzhen People's Hospital, three hospital beds were separated by curtains.
[part:gemini-3.1-flash-lite]
Liu Yang had already woken up, hooked up to an IV drip, his face still pale.
Zhang Yue was undergoing treatment to protect her pregnancy, and her husband was holding her hand.
Chen Hao had just finished his CT scan; the diagnosis was a mild concussion, requiring three days of hospitalization for observation.
Li Weimin stood in the corridor, looking through the glass at his three young subordinates.
His back was hunched, as if he had aged ten years all at once.
Xiao Zhou said softly: "Academician Li, this is not your fault. Everyone wanted to complete the mission early..."
"How is it not my fault?" Li Weimin's voice was hoarse. "I am the technical lead. I pushed them, I led the way in staying up late, and I set a wrong example for them. I thought... I thought that as long as there was a technological breakthrough, any price was worth it."
He remembered Liu Yang excitedly saying yesterday, "Academician Li, your new algorithm is amazing."
He remembered Zhang Yue discussing validation cases with him just last week; she must have already started experiencing morning sickness then, yet she said nothing.
He remembered Chen Hao patting his chest and guaranteeing, "Don't worry about the physical design, I can handle it."
"The chip can be remade, but what about the people?" Li Weimin muttered to himself. "If something serious happened to any of these three, I would never forgive myself for the rest of my life."
Rapid footsteps came from the end of the corridor.
Su Chen arrived, looking travel-worn; he had just flown back from Singapore overnight. His suit was wrinkled, and his eyes were bloodshot.
"How is the situation?"
Li Weimin turned around. This old Academician, usually so resilient, had red eyes: "Su Chen, I... I might have been wrong. The ninety-day military-style pledge might really have been too rushed."
Su Chen did not answer immediately.
He walked to the hospital room window, looked at the three people lying inside, and watched for a long time.
Then he said: "Academician Li, the one who was wrong is me. It was I who set the unrealistic deadline; it was I who passed the pressure on to you. You and the team have already created a miracle—the timing-driven algorithm. Do you know what this means? It means Chinese EDA might be able to overtake on a curve. But the price should not be like this."
He turned around, his eyes firm: "From now on, I will take over command of the R&D strike team. You continue to be responsible for the technology, but personnel management will be up to me."
Su Chen's "Mandatory Rest Order"
August 31st, 8:00 AM, EDA strike team plenary meeting.
The conference room, which could hold two hundred people, was packed, and some people were even standing in the back.
The atmosphere was oppressive; no one spoke. The three incidents of collapse were like a giant boulder pressing on everyone's heart.
Su Chen walked onto the podium. Without an opening statement, he went straight to the topic.
"In the past 15 days, we have completed a workload that would have taken at least three months under the traditional workflow. Academician Li Weimin led everyone to achieve a major breakthrough in EDA theory, increasing efficiency threefold. These achievements are enough to be written into the history of Chinese semiconductors."
His voice was steady and powerful.
"But today, three colleagues are lying in the hospital. One may have permanent heart damage, one may lose her unborn child, and one has a concussion. If this is the price of victory, then we would rather not have this victory."
Someone in the audience lowered their head.
"Therefore, I am announcing three ironclad rules, effective immediately. This is not a suggestion; it is an order. Violators, regardless of their position, will receive a warning for the first offense and be removed from the strike team for the second."
The regulations popped up on the big screen:
[Mandatory Health Management Regulations for the EDA Strike Team]
1. Mandatory Working Hours System
Daily working hours shall not exceed 10 hours (including a 1-hour lunch break).
Weekly working hours shall not exceed 60 hours.
The system will automatically monitor hours and lock devices upon overtime (including workstations and server accounts).
Power in the laboratory area will be turned off from 10 PM to 7 AM (applications are required for emergencies).
2. Mandatory Leave System
Must rest for 1 day after 6 consecutive days of work.
At least 2 full days of rest per month (24 consecutive hours without working).
Mandatory 3-day leave every quarter (Spark will cover the costs of short trips).
3. Mandatory Health Monitoring
Core members must undergo weekly ECG, blood pressure, and blood routine tests.
The campus infirmary will be upgraded to a 24-hour health center, equipped with 3 full-time doctors and 6 nurses.
A mental health counseling room will be established, and professional counselors will be hired.
There was a commotion in the audience.
"General Manager Su, the ninety-day deadline..." someone whispered.
"I will re-evaluate the deadline. If it must be extended, then it will be extended," Su Chen said decisively. "But one thing is certain: there will not be another person collapsing from exhaustion. If the progress is delayed because of rest, I will take the responsibility. But if there are casualties due to overwork, none of us can bear that responsibility."
He switched to the next PPT slide, which listed supporting measures:
1. Introduction of Scientific Working Methods
All staff will be equipped with smart wristbands to monitor fatigue.
Implementation of the "Pomodoro Technique": 10 minutes of mandatory rest for every 50 minutes of work; staff must leave their workstations during the rest period.
Establishment of a "Relaxation Zone": equipped with sofas, greenery, music, and books for use during breaks.
2. Nutrition and Logistics Support
A five-star hotel chef team has been hired to be responsible for three meals a day, prepared according to a nutritionist's formula.
Establishment of an "Energy Supply Station": providing fruit, nuts, and yogurt every hour.
Introduction of professional masseurs to provide free shoulder and neck massages every day from 3:00 PM to 5:00 PM.
3. Family Care Program
Establishment of a "Family Care Fund": medical expenses for employees who collapse from exhaustion will be fully reimbursed, plus an additional subsidy of 50,000 yuan.
Monthly "Family Open Day" events will be held, inviting family members to visit the campus and have dinner together.
After-school childcare services will be provided for employees with children (hiring retired teachers).
4. Adjustment of Incentive Mechanisms
Cancellation of "working hours" as an assessment metric, replaced by "effective output" assessment.
Establishment of the "Healthy Work Model Award" to reward employees who are efficient and have regular work-rest schedules.
During the recovery period for employees who collapsed, salaries will continue to be paid, and equity incentives will be increased.
Su Chen scanned the room: "I know many of you might think: chip development is just like this; it requires risking one's life. Yes, it requires giving it your all—but giving it your all does not mean throwing your life away. What we need to compete with is intelligence, efficiency, and methodology, not seeing who can stay up later or who can tough it out more."
He called up the last PPT slide, which displayed a simple formula:
Sustainable Combat Capability = (Ability × Efficiency) / Fatigue Level
"The higher the fatigue level, the lower the actual combat capability. For someone working twelve hours a day, their efficiency in the last four hours may only be thirty percent of the normal level, and they are prone to making mistakes. And the cost of errors in chip design can be weeks of lost time. Therefore, mandatory rest does not slow down progress; on the contrary, it speeds it up."
The conference room was quiet for a few seconds, and then applause broke out. It was sporadic at first, then gradually grew like a tide.
Li Weimin sat in the first row, looking at Su Chen on the stage with a complex gaze. This young man was more far-sighted than he had imagined.
After the meeting, Su Chen found Li Weimin alone.
"Academician Li, you also have to follow the regulations. Moreover, you are a key person to be protected."
Li Weimin smiled bitterly: "I also have to work no more than ten hours a day? This is the critical period for algorithm refinement..."
"That's why I've assigned you an assistant." Su Chen waved toward the door.
A tough-looking young man walked in, with a buzz cut, standing straight, and having sharp eyes.
"This is my cousin, Li Jianfeng, a veteran and a Sanda champion. From today on, he is your 'Health Supervisor.' I have already sent him your schedule: no more than nine hours of work per day, you must leave the laboratory by 10 PM, and you must lie down for a one-hour lunch break."
Li Weimin's eyes widened: "Are you... sending a bodyguard to monitor me?"
"You can understand it that way." Su Chen did not back down. "If you work overtime, he has the authority to 'escort' you out of the laboratory—physically. If you don't eat lunch, he will bring the meal to you and watch you finish it. If you stay up late, he will cut off the power to the laboratory."
"Nonsense! I..."
"Academician Li." Su Chen interrupted him, his voice suddenly becoming very soft. "Do you know what will happen if Zhang Yue's baby cannot be saved? She may never be able to forgive herself, never be able to forgive her work, or even forgive you and me. And you, you are seventy-two years old; if one day you also collapse, do you think this team can continue?"
Li Weimin fell silent.
"You are a national treasure." Su Chen said every word clearly. "Your brain, your seventy years of experience, are more precious than an EUV lithography machine. These precious things must be used for the most critical technological breakthroughs, not consumed in endless overtime. One year of healthy work from you contributes more than three years of working while ill."
After a long time, Li Weimin sighed: "Fine, I will listen to you. But I have one condition—those three children's salaries must be doubled during their treatment, and their equity incentives must be increased. Also, after they recover, they are not allowed to return to high-intensity positions immediately; they must adapt by starting with light work."
"It has already been arranged." Su Chen smiled. "Not only that, we are helping Zhang Yue's husband with a job transfer—from Guancheng to Shenzhen, so he can take better care of his wife. Chen Hao's parents have been brought from their hometown and are living in the campus family apartments, with expenses covered by us. Liu Yang's girlfriend is in the capital, and we have provided her with an interview opportunity at our Shenzhen branch."
Li Weimin was stunned: "You... you thought of all this?"
"Academician Li, the essence of technological competition is talent competition. And protecting talent is protecting the most core competitiveness." Su Chen said seriously. "What Spark wants to build is not a sweatshop that squeezes employees dry and then discards them, but a place where talent can maximize their value while also living a happy life. This is sustainable development."
Li Weimin looked at the young man in front of him and suddenly understood why so many people were willing to work for him.
It wasn't because of the high salary, not because of the equity, and not even entirely because of the national mission.
It was because, working under this man, you feel that you are a person first, and then an engineer.
"By the way, there is more good news to tell you," Su Chen said. "Data received this morning: ten days since the launch of the overseas talent return program, we have received a total of 893 resumes, including 167 A-level or higher talents. 42 people have officially signed contracts, including Lin Haoran—he chose the Singapore indirect return-to-country plan."
Li Weimin's eyes lit up: "Lin Haoran? That genius architect from Nvidia?"
"Yes. He has already arrived in Singapore and has started building the GPU team. Also Wang Weiwei, Chen Rui... a large group of top talents are on their way back." Su Chen looked out the window, where the morning sun was rising from the east. "So, Academician Li, we must take care of our health. Because the most exciting battle is about to begin. We need you to be healthy and clear-headed to lead this increasingly powerful team to fight a battle that will truly change history."
Li Weimin straightened his back.
At that moment, this seventy-two-year-old old Academician had a light in his eyes that was rekindled, like that of a young person.
"Good," he said, "I will take care of myself. But you must also promise me one thing—when this batch of overseas talent arrives, you must let me meet them. Especially Lin Haoran; I want to have a good talk with him about heterogeneous computing."
"It's a deal."
Two hands were clasped tightly together.
The intersection of one era and another, the relay of one mission and another.
Outside the window, the sky over Shenzhen was cloudless. This young city was waking up; traffic began to flow, and the glass curtain walls of the high-rise buildings reflected the morning light.
And on the other side of the Pacific Ocean, more people were making choices.
The road home might be winding, but the direction was already clear.
The second half of the chip war, the battle for talent, had just begun. And this time, China would no longer be just passively defending.
🔊 Text To Speech
Listen while reading