Technological Explosion

Chapter 83 Technological Explosion (5k)

After Yu Dazui left, Yang Yi sat in his office and began to think about the future of the bee colony.

I traveled abroad to Norway and Iceland. The scenery was beautiful. I felt the wind on the snow-capped mountains, daydreamed under the aurora borealis, and fed seagulls by the fjords. I saw everything I wanted to see and took all the photos I wanted to take.

But after he finished playing and came back, he found that he didn't really want to go out again. It wasn't that the scenery was bad, but that he felt uneasy.

When you're abroad, everything feels off – you can't speak the language, the food is different, and the people on the street have something in their eyes that you don't understand.

He prefers the peaceful environment in China to the lawlessness abroad.

At worst, we can work hard to develop virtual reality technology and strive to allow people to travel the world without leaving home.

Moreover—he felt that his future self could do it. Because the system's advanced technology had already shifted in that direction, and it was becoming increasingly outrageous.

Facial capture technology (research and development)

[Body tracking technology (research and development)]

These are the two latest cutting-edge technologies that Yang Yi has drawn. They look ordinary, but you only realize how powerful they are when you read the instruction manual provided by the system.

Facial capture can achieve micro-expression level, capturing even the slightest movement of an eyebrow, the curve of the corner of the mouth, and the movement of the eyeballs, all in real time with extremely low latency.

The body capture is even more outrageous; the movements of the limbs, torso, and fingers—all ten fingers—can be independently identified.

These two technologies perfectly compensated for the shortcomings of his previous vision of a virtual girlfriend.

Previously, there was tactile feedback—gloves, shoe covers, tongue covers, and that coded item.

But that's just one-sided acceptance. The device gives you feedback; you feel touch, temperature, and vibration, but you can't respond.

For example, if a virtual girlfriend sticks out her tongue, you can only feel it by making the same movement.

All your tactile sensations are pre-arranged by the system; it's simply an incredibly immersive movie experience.

It's different now.

When you stick out your tongue, facial and body capture will record your movements and transmit them to the AI ​​model.

AI understands your intentions based on your actions—if you want to kiss, it will make your virtual girlfriend kiss you.

If you reach out your hand, the AI ​​will bring your virtual girlfriend closer.

If you step back, the AI ​​will make your virtual girlfriend stop, tilt her head, and look at you with a puzzled expression.

From passive to active.

From "You watch" to "You participate".

This isn't a movie; it's a free-interactive virtual reality game.

With these two cutting-edge technologies, Yang Yi even felt that he was capable of creating Sword Art Online.

Of course, having the ability to make it doesn't mean you can actually make it; there's still a lot of advanced technology involved.

For example, data transmission, local computing, world-class servers —

These are all hardware gaps.

A long way to go.

However, by fully understanding current technology, they developed works such as "Mita," "Summer Rhapsody," and "Father."

"Yan's Mother" — still light and easy.

It could also develop horror games like Outlast and Silent Hill.

Is "Mita" a horror game or a dating sim?

Anyway, this corner is definitely going to be occupied by Yang Yi, the King of Wood.

We could even go a bit further and turn real people into AI, then who wouldn't want to replace Trump and enjoy "Sanae causing trouble"!

They could even have someone collect the body data of a "dark version" of Lin Chi-ling and turn it into a game.

Yang Yikun's thoughts were hard.

Of course, Yang Yi cannot spread color or unethical black technology in China; he will impose fundamental restrictions.

Create a special key and keep it in your own hands.

Use all of this for good.

We've gone a bit off-topic, but Yang Yi has already lost her composure.

He pulled Shen Xiao, who was lying on the sofa reading a novel, up.

I left work early.

He needs to calm down.

Two hours later, Yang Yi, lying in bed, finally found the peace she longed for.

My thinking has also become much more agile.

The most important thing now is the press conference in May.

Yang Yi agreed with Yu Dazui's suggestion. He had worked in the telecommunications industry for thirty years, rising from a junior engineer to a managing director. He had experienced countless storms, encountered numerous pitfalls, and talked to countless clients.

Yang Yi's advantage lies only in his pseudo-prophet status and his system.

In terms of individual ability, they are completely inferior.

Take the press conference for example; Yang Yi imitated the mobile phone press conferences and game showcases from her previous life.

Although the answer is correct, Yang Yi's idea of ​​merging the two is wrong.

Technology is technology, and games are games.

As Yu Dazui said, the best solution now is to hold two press conferences.

A game showcase and a product launch event by Beehive Technology.

It is used to attract different audiences.

However, as mentioned before, the problem is that a battery alone cannot support a technology launch event.

Yang Yi needs to produce more results within two months, more tangible and presentable achievements.

Looks like I'll have to go to the semiconductor branch and try my luck.

AI is also a top priority in Yang Yi's business empire.

Especially with the gradual implementation of cutting-edge technologies such as facial capture, body capture, and haptic feedback, AI becomes the thread that connects everything.

Without AI, haptic feedback is just vibration, facial capture is just animation, and body capture is just a motion library.

These technologies are highly dependent on AI, making the development of AI an urgent matter.

For this reason, Yang Yi contacted several headhunting firms.

Currently, there is practically no sign of AI model development in China.

Acquisition was not an option, so we had to build our own team and start from scratch.

Moreover, Yang Yi, the head of the AI ​​large model company, personally holds it in his hands.

The next morning, Yang Yi didn't go to the company; instead, she drove directly to Beehive Semiconductor.

The semiconductor branch is located in the Shenzhen National Integrated Circuit Design Industrialization Base, about half a district away from the headquarters of Hive.

When Yang Yi arrived, it was not yet nine o'clock. The young girl at the front desk was wiping the table. She was stunned when she saw him. The boss's arrival was too sudden; no one had notified her in advance.

Is Wang Teng there?

"Yes, Mr. Wang is here early today."

Yang Yi nodded and walked inside.

The corridor wasn't long, with closed laboratories and offices on both sides. At the end, Wang Teng's office door was ajar, and a voice on the phone could be heard inside: "The photomask has been delivered—yes, production there won't start until next month—okay, you can get back to work."

Yang Yi knocked on the door.

Wang Teng hung up the phone, looked up and saw him, paused for a moment, then stood up: "President Yang? What brings you here?" "What a coincidence! The chip samples just arrived, and I was just about to come and report to you later."

"Come and take a look." Yang Yi sat down. "There's been a major technological breakthrough in swarm batteries. I'm planning a tech launch event in May. What can you offer?"

Wang Teng tidied up the documents on the table, his tone carrying a sense of relief that "finally someone has come to ask": "There are several, some are still in the laboratory stage, and some have already been tested and approved."

"Let's get it running first."

Wang Teng stood up: "Please come with me."

The two people entered a laboratory.

Several engineers tried to stand up when they saw their boss come in, but Wang Teng waved them off and walked straight to a testing instrument. He took out a small box from the electrostatic box, opened the lid, and found a chip the size of a fingernail inside.

"The first-generation GPU engineering sample, codenamed FG100, is a 12nm chip manufactured by TSMC. It just arrived last month and has already been powered on. The officially released model is called FG190. The first digit '1' represents the first generation, the second digit '9' represents flagship performance, and the third digit '0' represents no further subdivision."

"So fast!" Yang Yi was a little surprised and picked up the chip sample to examine it under the light.

This was the first time he had ever seen a bare GPU chip.

"I'm not quite sure either. The research and development speed is like it's been cheated. Even the researchers themselves find it incredible."

This was the first time Wang Teng had ever witnessed such rapid scientific research, and his tone carried a strange feeling—the unpredictability of life, like a large intestine enclosing a small intestine.

"TSMC's process is like this," Wang Teng explained from the side. "Once the design is sent over, they first make a photomask. After the photomask is made, it goes on the production line for photolithography, deposition, etching, doping, and so on, layer by layer."

After the process is complete, a whole wafer emerges, densely packed with chips. During dicing, a diamond blade is used to cut along the scribe lines, separating each chip individually. The resulting wafers are called bare dies, and these are then packaged to become the sample wafer we have in our hands.

"How is the performance?" Yang Yi asked.

"Traditional raster performance is 53 percent stronger than GTX1080Ti." He wrote a comparison on the whiteboard.

"The upgrade from 1080Ti to 2080Ti, according to industry practice, is about 30% to 35%. Ours is 53% stronger than the 1080Ti, which means it is about 20% stronger than the 2080Ti."

Yang Yi nodded upon hearing this. Twenty percent may not seem like much, but in terms of performance alone, it's already close to that of the 3080.

The.

Wang Teng turned to the next page of the test report.

"We won't just release one model. A single wafer will have hundreds of cores, and not every single one can run at full frequency."

They select the best-performing chips from the center of the wafer to make the flagship FG190. Then they cut down—removing a set of computing units and lowering the frequency slightly—to create the FG170, which is 15% to 20% less powerful than the flagship. They cut it again to make the FG150, another 15% to 20% less powerful. And then they cut it again to make the FG130.

The smaller the number, the lower the performance; even numbers were skipped to allow for future sub-models.

This touches on a blind spot in Yang Yi's knowledge.

So that's how low-end models in the chip industry come about.

Yang Yi then asked, "What about ray tracing chips?"

Wang Teng replied with a smile: "Ray tracing is divided into two types: software and hardware."

"Hardware ray tracing chips are tied to GPU chips, and the ray tracing performance of each model decreases with the core specifications."

The FG190 has the strongest ray tracing performance. The FG170's ray tracing performance is 15% to 20% lower than the FG190, and the FG150 is even lower, followed by the FG130.

The first generation of ray tracing was uniformly called "Ray Tracing Generation 1," without separate naming for different model types. However, the performance differences in ray tracing between different GPUs were directly reflected in the GPU model. Users who bought an FG190 and an FG130 would experience different ray tracing experiences—the more expensive one naturally offered stronger ray tracing.

"The software is bound to the Bee Engine, and the software engineers have already added a ray tracing development module to the engine."

""

Yang Yi was surprised that Wang Teng had arranged everything so perfectly; he was now truly relieved.

Has the patent application for ray tracing been approved?

Wang Teng nodded upon hearing this: "I used the method you provided, applied in advance, and then studied it."

"Although the patent has not yet been granted, we have firmly secured the priority rights."

"The patent is not published during this period. If you want, you can mess with Nvidia once and directly cripple them."

"Moreover, our patents for ray tracing and supersampling cover a wide area, which they cannot bypass in a short period of time, and even if they do, the performance will be greatly reduced."

Although Wang Teng's plan was excellent, it was not without its costs.

This means that Yang Yi cannot announce his company's GPU before Nvidia releases its 20-series graphics cards.

This would disrupt the schedule of Beehive Technology's press conference.

Since the patent coverage is extensive, there's no need for such underhanded tactics; we can just confront them head-on.

With this in mind, Yang Yi rejected Wang Teng's plan and explained his reasons.

When Wang Teng saw that his boss had rejected the idea, he stopped mentioning it and continued to explain neural network super-resolution rendering to Yang Yi.

"BeeSU1.0, short for Bee Colony Neural Network Super-Resolution Rendering Technology".

Real-time image super-resolution reconstruction using deep learning models—game rendering of low-resolution images, real-time neural network inference to fill in details, and output of high-resolution images.

The computational load is less than a quarter of that of native high-resolution, and the image quality loss is almost imperceptible to the naked eye.

Yang Yi picked up the test report, flipped through it, and put it back. "How do we integrate BeeSU into the game?"

"Packaged as a built-in plugin for the BeeSU engine. Developers using the BeeSU engine to create games simply need to enable Bee swarm neural network super-resolution in the project settings," and then select the target graphics quality level, balance, and performance. The engine will automatically insert BeeSU calls at appropriate locations in the rendering pipeline. Developers don't need to write a single line of code.

"

Yang Yi nodded. "Don't go into too much detail at the press conference. Just say that there's no loss of image quality, the frame rate is doubled, and developers can use it with a simple click."

Wang Teng led him to another instrument, took out a chip from the electrostatic box, and connected it to the oscilloscope and logic analyzer.

"The first-generation RISC-VCPU, 12nm, TSMC. Officially released model FC190, eight cores and sixteen threads, base frequency 3.8GHz, all-core turbo frequency 4.6GHz, thermal design power 105 watts."

Yang Yi looked at the row of numbers. "What's the strongest product on the market right now?"

"The Intel i7-8700K has six cores and twelve threads, a base frequency of 3.7GHz, and an all-core turbo frequency of 4.3GHz. Our FC190 has two more cores, a higher frequency, and supports hyper-threading."

Wang Teng pulled out a comparison chart: "The multi-core performance is about 60% stronger than the 8700K. The single-core performance is about 15% stronger."

To be precise, the 8700K is less than 5% stronger than the FC150.

Yang Yi was surprised to find that both the CPU and GPU were already designed.

Moreover, its performance is extremely powerful.

"Should we introduce the motherboard next?"

"Haha, indeed!" Wang Teng led Yang Yi to another laboratory.

Several different models of motherboards were placed on the experimental table.

Wang Teng pointed to the largest one and said, "Supreme Invincible Super Pro motherboard, sixteen-phase power supply, eight memory slots, three M.2 slots, four PCIe slots..."

16. Supports dual graphics card CrossFire.

Yang Yi smiled helplessly at the bizarre name: "I know you wanted to emphasize how awesome this motherboard is, but this name is just too—"

Wang Teng smiled and explained, "It's okay. The motherboards from now on will all be made by downstream manufacturers, and they'll come up with better names for them. Mine is just to highlight how awesome the motherboard is."

"There are also a few regular motherboards, but they're all stripped-down versions—"

Yang Yi put the motherboard down: "You're very confident today. It seems your research results are amazing. Bring up the rest."

Wang Teng smiled and picked up a thick technical document from the table.

"BeeCore1 is our own Linux distribution. It natively supports the RISC-V architecture, boots directly to the desktop, and includes a browser, file management, and video playback. We've implemented an installation logic similar to Windows, and even the installation steps are the same."

"How does the 86 software run?"

"Bridge1 is a binary translation layer. It translates 86 instructions into RISC-V instructions in real time, with a loss of 6% to 8%, and has a hardware-accelerated coprocessor. Users can download the 86 version of the software, double-click to install, and Bridge1 will automatically translate and run."

"At the same time, we added the architecture to the Queen Bee Engine, and selected native export of the swarm during compilation," directly outputting native binary data without the need for swarm bridge translation.

Performance is 6% to 8% faster than traditional translation. Existing games use BeeBridge translation, and a repackaging tool will be released later to directly convert the 86 version to the native BeeBridge version with zero performance loss.

"We've also developed a Bee Core version of the Bee Engine, and the software development tools are under intensive development, aiming to be completed before the launch event."

Everything's ready! The computer's all set!

Translate what "TM's surprise" means!

At the press conference, they unleashed a series of cutting-edge technologies, leaving their competitors reeling.

Furthermore, it even laid the foundation for AI using neural network super-resolution rendering.

What else could Yang Yi say?

That's all there is to it!

"Great, a 50% raise for everyone, plus an extra two months' salary as a year-end bonus."

"There's still stock." Wang Teng took two more small boxes out of the drawer.

"besides!"

"Link1 is a short-range wireless communication protocol. It was developed incidentally while researching high-speed interconnects between chips. Its bandwidth is dozens of times higher than Bluetooth, its latency is less than one-tenth of Bluetooth, and its power consumption is about the same. It is already integrated into the motherboard, alongside the Bluetooth module. For users connecting gamepads, keyboards, mice, and headphones, the experience with Link1 is significantly better than with Bluetooth."

"The Sense1 is a six-axis inertial sensor that integrates an accelerometer and a gyroscope."

It was done incidentally while developing the power management chip—its power consumption is only one-third of similar products on the market, and its size is half that of other products.

A single chip inside the controller doubles the accuracy of arm movement tracking compared to current solutions, while consuming half the power. It works with drones, VR devices, and smartphones. The material cost is less than ten yuan.

This is absolutely outrageous! Opening the door is outrageous to the extreme!

These two scraps, taken individually, could even be considered miniature cutting-edge technology.

It is definitely worth continuing to study.

Yang Yi is no longer worried about running out of things to say; instead, she's worried about having too much technology and the press conference being too long.

GPU, CPU, Bee Core, Bee Compatibility Layer, Bee Connect, Bee Sensing.

There are also detailed ray tracing and neural network super-resolution rendering.

I should also mention the ray tracing and super-resolution rendering capabilities of the Bee Engine 2.0.

The official version of the engine will definitely be shown at the game's Direct presentation.

"Don't forget to provide peripherals for the bee colony with BeeLink and BeeSense."

Wang Teng replied, "I've already given it to them. The two companies are so close, I can't forget it."

Yang Yi smiled mysteriously: "There's one you've definitely forgotten."

Wang Teng scratched his head, unable to recall anything he had overlooked.

"Your software team is now in charge of Queen Bee Engine 2.0, right?"

Wang Teng nodded. He was in charge of the largest number of people in the entire Beehive Technology organization.

This resulted in many odd and ends being registered under his name.

"Have you considered adding controller drivers for swarm peripherals to it?"

Wang Teng slapped his forehead: "Oh dear, I really forgot."

"You must always remember that Bee Queen Engine actually has overlapping business with many departments."

"You should separate the team and keep the data synchronized across the subsidiaries at all times."

"The version you updated will affect the drivers for both headquarters and the peripheral subsidiary."

Wang Teng quickly nodded in agreement and promised to standardize the team's procedures.

>