Rebirth: Starting from Lighting Up the Tech Tree

Chapter 124 Taking a Different Approach

The cooldown countdown for the technology radar has fourteen hours remaining.

Zuo Cheng stared at the numbers on the system panel for a while, then closed it. He didn't need to keep watching; the system would remind him when the time was up.

That afternoon, Shen Yiming submitted the initial draft of the technical framework. Twelve pages long, the core idea was clear: replace single-chip high-performance solutions with a distributed edge AI architecture. Each satellite would deploy only a lightweight decision agent, collaborating with other agents via inter-satellite links to complete complex AI inference tasks.

After reading it, Fang Ze said, "If the latency issue can be resolved, the cost of a general-purpose chip will only be one-fifth of that of the MLU270."

"I'll handle the delay," Zuo Cheng said.

Shen Yiming glanced at him but didn't ask any further questions. Having worked with Zuo Cheng for so long, he knew that when Brother Cheng said "I'll take care of it," he usually really had a solution.

That night, Zuo Cheng did not return to his dormitory.

He sat on the office sofa, with Shen Yiming's technical framework spread out in front of him, looking at it while mentally comparing it with the list of blades on the system panel.

The two branches of communications engineering and aerospace communications together comprise dozens of blades, with seven or eight directly related to inter-satellite communication. Theoretically, their fusion could generate a dedicated communication protocol.

But theory is one thing; he first needs to confirm the cooling status of the technology radar.

At 4:17 a.m., the system panel flashed.

[Cooling of the technology radar is complete]

Zuo Cheng sat up straight.

What he needed wasn't to scan a person, but a technology. While the scanning function of a technology radar primarily targets personnel, he had previously used a different method: directly integrating technology using existing blades, eliminating the need for scanning.

The last technology integration cost 5 points and obtained an adaptive neural network compiler, which directly increased the accuracy from 89.3% to 97.1%.

This time, he needs to integrate the blades in the communication direction.

He opened the panel and browsed the list of leaves on the communications engineering branch.

【Intelligent Satellite Network Scheduling System】—Sky Dome Satellite Scheduling Optimization.

Inter-satellite link adaptive coding – improving the efficiency of inter-satellite communication.

[Low-Latency Transmission Protocol] - A basic communication protocol with latency optimization.

Three blades. If the fusion is successful, the resulting protocol should be a dedicated communication protocol for inter-satellite distributed computing, with significantly lower latency than existing general-purpose protocols.

However, there is a risk. The success rate of technological integration is not 100%. The previous successful integration of three blades does not guarantee success every time. If it fails, 5 points will be wasted, and another 48 hours of cooling-off time will be required.

He did not hesitate.

Sky Dome Phase IV can't wait, Blue Bay Communications can't wait, and neither can the development team of over forty people.

[Technology Integration Launched]

Please select at least two blades for fusion.

He chose those three leaves.

The system begins its calculations. A circular progress bar appears on the panel, and a blue light slowly moves along the circle.

Zuo Cheng's breathing was even, and his heartbeat didn't quicken. He was used to this kind of waiting. Every system calculation was a gamble, but he could afford to lose 5 points; what he couldn't afford to lose was time.

Three seconds later, the progress bar was full.

【Successful integration】

New blades acquired: **Intelligent Nebula Communication Protocol**

Results: Inter-satellite communication latency is reduced to less than 3 milliseconds, bandwidth utilization is increased by 40%, and real-time collaboration of distributed AI tasks is supported.

3 milliseconds.

Zuo Cheng looked at the number, and the corner of his mouth twitched slightly.

Shen Yiming previously stated that the latency of traditional inter-satellite communication protocols is in the tens of milliseconds. To achieve real-time collaborative decision-making between satellites, the latency must be controlled within 10 milliseconds, and the lower the better.

Three milliseconds means that communication between satellites is almost negligible. For distributed AI, this is sufficient.

He summarized the capabilities of the new blades and wrote a technical description, presenting it as an "algorithm-assisted design tool".

The next morning at eight o'clock, Zuo Cheng called the team to the conference room.

"The delay issue is resolved," he said.

Shen Yiming looked up at him, Fang Ze adjusted his glasses, and Han Lu put down her phone.

"I've designed a communication protocol specifically for satellite distributed computing," Zuo Cheng projected the technical specifications onto the large screen. "Key metrics: inter-satellite communication latency below 3 milliseconds, and bandwidth utilization improved by 40%."

The meeting room was silent for two seconds.

Shen Yiming was the first to react, and walked to the screen to carefully look at the parameters several times.

"3 milliseconds?" He turned to look at Zuo Cheng. "How did you calculate that?"

"I ran a lot of simulations using computer-aided design tools," Zuo Cheng said. Shen Yiming wouldn't press him for an answer.

"If it's 3 milliseconds, then the distributed solution is perfectly feasible." Shen Yiming quickly calculated in his mind, "The communication latency between decision agents is negligible, and the real-time performance of cluster collaboration is not a problem."

Fang Ze also reviewed the specifications. "If the latency can be stabilized below 3 milliseconds, the general-purpose chip solution can completely replace the MLU270 in terms of performance. The cost will be reduced by at least 80%."

"Cost isn't the key issue," Zuo Cheng said. "The key issue is time. General-purpose chips are available now; there's no need to wait for Cambricon's customized version."

This statement awakened everyone.

Not having to wait for chips means that the development of Sky Dome Phase IV will not be delayed due to supply chain disruptions.

"Fang Ze, compile a list of general-purpose chips to be selected by today, and I will send you the performance requirements," Zuo Cheng said.

"Okay." Fang Ze nodded.

"Yiming, refine the technical framework and integrate the new communication protocol. Give me a complete solution document by tomorrow."

"No problem." Shen Yiming was already flipping through his notes.

"Han Lu, keep following up on the supply chain issues; don't stop. Even if the software solution is feasible, problems with hardware suppliers will inevitably resurface sooner or later, so we need to prepare in advance."

"clear."

The meeting lasted only fifteen minutes, and everyone's task was clear.

After the meeting, Shen Yiming stayed behind.

"Brother Cheng," he hesitated for a moment, "what exactly is that design tool you're talking about?"

Zuo Cheng looked at him and smiled. "It's an algorithmic reasoning software, made by a friend."

"Your friend's software can reach this level?" Shen Yiming was clearly skeptical, but he didn't press further. Having worked with Zuo Cheng for so long, he knew that there were always benefits to Zuo Cheng keeping quiet.

"Forget about the tools, getting the plan right is the important thing." Zuo Cheng patted him on the shoulder.

Shen Yiming nodded and left.

Zuo Cheng remained alone in the conference room, staring at the technical parameters that were still displayed on the large screen.

With a latency of 3 milliseconds, bandwidth utilization is improved by 40%.

To others, these numbers represent the design achievement of a genius engineer. Only he knows that behind them lies the 5 points spent out of 252 points, and the fusion of three blades.

He opened the panel and took a look.

Points: 252.

Technology Radar: Cooling down, 48 hours.

He turned off the panel, stood up, and walked out of the conference room.

The hallway was bathed in sunlight. He walked back to his office, sat down, and turned on his computer to handle other tasks.

The supply chain crisis has been temporarily resolved, but he knows this is not the end. Tianxing Capital is still involved in the industry chain, and Lin Jianhua will not give up just because of one adjustment to the plan.

He needs to find a more fundamental solution.