Chapter 496: Games have frame rates, but does reality have them? (5737)

Mirage 1.0! ?

Self-developed search engine!

Everyone at the scene was in an uproar.

Wait, Fang Yu didn’t deny that this was a real machine demonstration just now. Instead, he said that it was based on Yuzu Technology’s self-developed engine - Mirage 1.0. Does this mean that this was really achieved on a mobile phone?

Not to mention the processing power of a mobile phone, even a PC with a GTX1080 dual-firepowered processor, or the PS4 Pro just released by Sony, cannot achieve this effect even with peripherals!

And this was achieved on a mobile phone?!

Or a self-developed engine?

"Of course, the underlying architecture for the video after the world introduction wasn't ready yet, so it was generated using AI. However, the game graphics for the 'Shadow' section were indeed developed using the 'Mirage Engine.'"

As soon as Fang Yu finished speaking, everyone present couldn't help but change their expression.

It’s really a real machine demonstration!

Fuck! This is outrageous!

Even if it wasn't AR and wasn't combined with the real scene, a considerable portion of the image would have been difficult for the mobile phone GPU to render.

The A10 released by Apple last month is not good, and Qualcomm's Snapdragon 821 is even worse!

If combined with real-time computing of AR real scenes, desktop-level products will have no chance!

Fang Yu knew what other people's expressions were without even looking, and smiled slightly.

"Some may question how the Mirage Engine can achieve real-time modeling and rendering on a lightweight computing platform like a mobile phone, and achieve almost zero latency in integrating it with the real scene."

Fang Yu took a stylus and wrote and drew on the keynote displayed on the big screen.

"I won't go into detail about the specific technical issues today, I'll just talk about the principles."

"First, the cloud architecture of the Mirage Engine actually handles the majority of the rendering calculations. What's returned to the phone is only the rendered 'status' data, which is actually very small, only about 3MB per second."

Cloud computing? Only 3MB of data transfer per second? Ten years ago, Half-Life wouldn't even be enough to render a wall, right?

“That’s not what everyone understands.”

Hearing some puzzled discussion below, Fang Yu shook his head: "We don't push the frame stream to the phone. Instead, we transmit the scene status, material mapping information, lighting solution, etc. of each frame back using a custom protocol. This part of the protocol data only takes up 3MB/s."

Is this true? Zhu Yifei's lips moved slightly, but before he could ask, he heard Fang Yu continue:

"Secondly, the reason these 'status' data can be so concise is because of Mirage's unique 'environmental probe array' technology."

Environmental probe array? Hearing this new term, everyone stared at Fang Yu intently.

"The 'Environmental Probe Array' is not a physical probe, but a purely software-based virtual concept developed by us based on an AI vision system. The calculations are performed on the mobile side."

"By invoking GPU resource rebinding, we can embed thousands of virtual probes into the visual environment within 2.5 seconds. Once these probes are embedded, the cloud-rendered material is solidified on these probes, ensuring that no matter how the camera moves, there will be no delay or delamination. This alone saves over 90% of AR re-rendering resources."

Everyone seemed to understand what was said. Although they were game developers, only a handful of them could understand how this underlying technology was achieved.

"Third, it's the most important feature of the Mirage Engine—'Simulated Reality Refresh.' It's this feature that enables the Mirage Engine to achieve the 3D rendering effects you just saw, using the computing power of mobile GPUs."

Having said that, Fang Yu paused, looked around, and saw that everyone had a confused look on their faces.

Simulated reality refresh? What does that mean? Floating refresh rate? VVR isn't exactly new technology, but mobile GPUs don't have it yet, right?

Fang Yu turned a page of the keynote.

"Simulated display refresh is not about adjusting the refresh rate, but an algorithm that can save 95% of rendering resources compared to Unreal Engine 4 in most environments. This allows Mirage to devote more computing power to better rendering effects."

Save 95% of rendering resources in daily scenes! ?

No algorithm can do this!

Zhu Yifei's forehead and cheeks were red, and he raised his hand with a frown on his face:

"Mr. Fang, even if the environment probe array you just mentioned can reduce the repeated rendering of materials in each frame, as long as it is a moving picture, there will still be a large amount of materials that need to be rendered in real time in each frame."

"In your first demo, the frame rate on screen was 38-60 fps. If this effect were rendered in real time, even with the theoretical performance of a GTX1080, it would be terribly slow, with a maximum of only around a dozen fps."

"Even the worst game engines effectively utilize over 30% of the GPU's core execution units. Unreal Engine 4 is somewhere between 45% and 60%, so there's not much room for improvement. It's not that I don't believe it, but saving 95% of rendering resources is a breakthrough in theoretical limits."

When Fang Yu saw there was someone to support the actor, his eyes lit up.

Sure enough, Boss Tiger, the road is widening now.

Fang Yu nodded, his lips curled up slightly. "You're right. If we use Unreal's real-time rendering method, it shouldn't run on either mobile or desktop. But, Boss Hu, I have a question."

Zhu Yifei looked sincere and said, "You go ahead."

Fang Yu raised her eyelids slightly and looked at Zhu Yifei: "Games have frame rates, but does reality have them?"

Zhu Yifei was stunned, looking as if he didn't understand: "What?"

Fang Yu pointed out the window. "Games are animated by varying frames per second. Whether it's a change in the environment or the action, it's like the computer drawing dozens or even hundreds of images per second. This is what allows us to see the virtual world in motion."

“But is this how reality works?”

Zhu Yifei's body trembled slightly, as if he understood something, but it seemed that he had not yet fully reacted: "You mean..."

Fang Yu smiled slightly: "I said, reality has no frame rate, only continuous state changes."

"Then why does rendering have to be done in 'frames'?"

Not rendering in frames? Then what?

Seeing the bewildered expressions of everyone present, including Zhu Yifei, Fang Yu shook his head and asked, "Have you all forgotten what Youzi Technology does?"

“We’re in the AI ​​business.”

After hearing these words, some people who were quick to react at the scene had their expressions completely changed.

Fang Yu looked around at everyone and raised an eyebrow. "The real world doesn't have a frame rate, so rendering shouldn't be based on discrete frames. Instead, it should be based on continuous predictions of 'changing trends.'"

"People don't reshape their posture every millisecond, and fabric doesn't recalculate its next position every second."

“All change is continuous.”

"So the Mirage Engine doesn't compute the scene state in the old way of running at 60 frames per second."

"We trained models for mirages based on hundreds of thousands of different materials, limbs, lighting, and environmental reactions, allowing us to directly 'imagine' the next reasonable state based on the existing information."

"We built a set of multimodal state mapping models that integrate materials, dynamics, and lighting, and can predict the next state in real time based on behavioral logic."

"For example, if someone is running, a traditional rendering engine has to calculate the skeleton, skinning, materials, lighting, synthesis, and frame output."

"The mirage engine calculates the model - 'What will his next pose be?' and then renders this series of poses."

"Instead of abandoning frame rate, we make frame rate a display interface rather than rendering decision logic."

“You see 60 frames, but the actual core rendering path only triggers reconstruction at the two-frame level, and the rest are continuous state frames generated by AI.”

"Furthermore, a large portion of the resource consumption for these two frames is also shared by the cloud server. After all, many physical characteristics of mobile GPU chips are not supported."

"It's not just a single graphics module, but an AI engine hub that integrates perception, prediction, content generation, and resource scheduling."

"Yes, that's right. 'Mirage Engine' is essentially a large model, a large model that works together with the cloud and mobile or desktop clients. I just said that it reduces mobile computing resource consumption by 95%, which is actually a very conservative estimate."

"This part of the algorithm gave us a headache for a long time, but we finally got it done. Although we are still not satisfied with the final result, it is usable. There is still a lot of room for improvement."

Fang Yu shook his head, looking dissatisfied.

"Everyone has a pad in front of them. There's no password. When you open it, there's only one app inside: the demo for 'Tainhuang: Mountain and Sea Realm'."

"You can take a look at it. Frankly speaking, besides being dissatisfied with the engine effects, I am even more dissatisfied with the art level and the game mechanism design, numerical values, gameplay, and rhythm planning of this demo."

"There's no way AI can compare to humans in terms of creativity in this area."

“I hope everyone can work together to help me fix the shortcomings of this demo within this week.”

"Thank you all in advance."

With a loud bang, everyone at the scene was shocked.

Fang Yu clasped his hands together and bowed his head modestly. When he looked up again, he saw that everyone had already picked up the Pad impatiently, and soon they were shouting at the screen.

Otakus who love playing games are so impolite, no wonder they lose their gold coins so easily!

Fang Yu curled his lips.

No one in the room paid any attention to his expression. They all stared at the screen with their teeth bared and their faces flushed, but were having heated conversations with their colleagues and friends around them.

Oh my god! Was the demonstration just now real?!

Why didn't you just let us try it out?

Huh? This pet is so interactive!?

Are pets also AI?

Damn it, damn it, damn it!

The gaming industry is really going to change!!!

If this game is really made, it will be incomparable to the 3A console games!

Older, middle-aged and younger otakus and fujoshis continuously issued exclamation marks that represent low IQ.

The game content of the demo is still very simple. It is just a novice guide with a short plot. It starts with collecting pets, followed by a teaching battle, and then nothing more.

After opening the pad and concentrating on experiencing the demo twice, Wei Ao raised his head and let out a long sigh, only to see Zhu Yifei also looking up in a trance.

Epic is finished.

When the two looked at each other, they both understood the meaning from each other's eyes, but they didn't know how they understood it.

No one expected that, originally they came here just to do some hard work for a mysterious client for Yu Dan's sake, but here they witnessed with their own eyes the news that would change... no, it should be said that it would reshape the entire industry!

The birth of AI games does not surprise them too much. With the current development speed of AI, everyone predicts that within three to five years, there should be real AI games.

But who would have thought that the first thing to be born was not an AI game, but an AI game engine! ?

The first to die is not the game company, but the engine company?

Wei Ao looked at Sui Yi with pity as he was communicating with the big rat he had just caught with great interest, and thought to himself, just be content. In the art industry, at least high-end artists will not be unemployed for the time being.

Look at Epic, it could die without any warning.

What really determines the quality of a game's graphics is not the graphics card, but the game engine and the development team's ability to apply the engine.

The engine determines whether the game has real-time global illumination, automatic occlusion culling, whether texture grading supports virtual streaming loading, and even whether multiple rendering domains can be dynamically merged within a frame...

These are the real dividing lines of picture quality.

When we say a game’s graphics are good or bad, it is essentially a matter of difference in the project development team’s ability to master, develop, and tune the game engine.

Both use Unreal Engine 3, but Chinese Paladin 6 is a mess. They had to change the engine midway through production, and the graphics are not as good as Gears of War from ten years ago.

Of course, there are also differences in financial resources.

A wealthy team can buy higher quality materials, and the pictures will naturally be better.

Looking at the trailer, it’s clear that this game is being run on mobile phones.

At present, it is good enough to use Unity in the mobile game field, but let alone Unity, even if the Unreal Engine 4, which is still the most powerful in the industry, is updated as often as possible, it is impossible to achieve such a visual effect!

No, it's not just Unreal Engine 4. Last year, a set of Unreal Engine 5 development parameters were leaked online. According to those leaked parameters, Unreal Engine 5 cannot compare to the game engine being demonstrated on the screen now!

Moreover, there are very few engines on the market that can support AR. Unity can only rely on the external SDK to compete with Vuforia's image recognition. Unreal is even worse. Although it has strong rendering, there are not many plug-ins that support AR.

In fact, it wasn't just Zhu Yifei. Some main programmers who were good at engine architecture and artists who were more sensitive to rendering had already realized this problem when the promotional video was played.

——Let’s not talk about the physical collision calculation methods and spatial modeling technologies that we just don’t understand. Just talking about the picture, there is currently no graphics engine that can support this level of AR perception, real-time spatial modeling and high-quality rendering.

The graphics engine is only part of the game engine. A usable game engine must at least have a physics engine, audio engine, input system, scene management, script system, UI system, animation system, network module, editor toolbox...

The bosses of the independent game development teams who came today are almost all in their thirties. They entered the industry because they loved games since they were young.

They are very familiar with the development history of game engines, and they understand how difficult it is to redevelop a game engine in today's era of "high-end Unreal and mobile Unity".

Not to mention the game engine that Fang Yu claimed "has performance far superior to Unreal 4 in all aspects", even if we find a game engine that can match Unreal's performance, no company can do it at present.

Because Unreal Engine is not only a 3D rendering framework, but also an entire ecosystem built around content production. Its development tool chain and material content far exceed any other engine.

Just as CUDA's ecological advantages ensure Nvidia's dominant position in computing cards, Unreal Engine 4 is the most powerful game engine of this era, without a doubt.

Unity has seized the mobile and 2D ecological niches with its lightweight features, but it is still unlikely to pose a substantial threat to Unreal in the foreseeable future.

It is generally recognized in the industry that in this era, the industry ecosystem has solidified and dominance has been formed. It is almost impossible for a game engine company stronger than Epic to emerge. The only one that can surpass Unreal Engine 4 may be Unreal Engine 5.

The game engine developed with the birth of 3D games and was originally called a 3D engine.

In the 2D game era, there was no concept of game engine. At that time, making a game basically meant writing code from scratch. If the code from the previous game could be used, it would be used. If it couldn’t be used, it didn’t matter. It didn’t take much time to write a new one.

For example, the Xuan-Yuan Sword 2 released 22 years ago had a total capacity of 33MB, which was only 12MB after compression. During installation, 11 3.5-inch floppy disks with a capacity of 1.44MB were required.

After removing art, MIDI music and sound effect resources, the actual code part of Xuan-Yuan Sword 2 is only about 1-2MB, and the total code volume is no more than 30,000 lines.

Only large companies like Ren Tianguo and Sierra, which were already leaders in the industry at the time, had some functional module packaging tools in their development departments.

It was not until 24 years ago that ID developed the world's first 3D FPS game - "Return to Headquarters 3D", and the engine + content development paradigm gradually took shape.

A year later, ID developed a groundbreaking FPS game in the history of games - DOOM (Doom) based on John Carmack's 3D engine, and licensed the DOOM engine to other companies through paid licensing. The complete concept of the game engine was officially established.

Moreover, in the following two or three years, ID developed Quake and Quake II based on this.

The 3D engine of the Quake series was once the most popular 3D engine before the birth of Unreal Engine, and it dominated an era.

Many 3D games used to have the 3Dfx Voodoo and Quake II or III logos printed on their disc covers, proving that they supported 3Dfx's Voodoo graphics cards and used the Quake game engine as a gimmick to attract players. (Note 1)

Until Epic's Unreal Engine came out.

In fact, Unreal was originally a game. The overall performance of the first generation of Unreal Engine was not better than Quake III of the same period. Although it had better color and rendering effects, its polygon capabilities were far inferior to the Quake engine.

In addition, Unreal's optimization was not as good as Quake III at the time, bugs emerged one after another, the development tools were not mature, and the physics engine's clipping problem was extremely serious.

However, Unreal has gained favor with more game developers after Epic announced open licensing.

Because it looks good and is cheap.

Although Unreal's technology is not more powerful than Quake III, it has more saturated colors, epoch-making lighting and translucent rendering effects, and a more varied scene system... In a word, when resources are limited, Unreal tends to look good, while Quake tends to be practical.

In addition, the licensing cost of Unreal is much lower than that of Quake. Although the direct licensing price of Quake is lower than that of Unreal, the price does not include the editor and also requires sales rebates.

The most important thing is that Quake's developer ID is very arrogant. Epic was able to assign dedicated technicians to provide customer service for Unreal Engine. The customer submitted a reproduction case request for support. When he received a reply, not only did he get a detailed answer to his question, but he also helped him fix the bugs he found.

In comparison, ID's Quake has no support at all. If you finally get in touch with it, it might be John Carmack who answers the call, and then you'll be scolded by him.

John Carmack of ID has a typical programmer's mindset, believing that technology comes first and everything else is bullshit.

If someone asks him for authorization and pays money, he will send the source code to you. Whether you can understand it or not is your own business. If you can't use it, it only shows that you are not a good user.

He was right, saying that I supported open source and didn't want to charge money, but the company insisted on charging me, and it had nothing to do with me.

In the end, good looks defeated practicality, business defeated technology, and service defeated arrogance.

In the Unreal 1 era, the two sides were still fighting each other, but in the Unreal 2 era, ID was basically defeated. By the time Unreal 3 was released, Epic had basically dominated the high-end 3D engine field.

After the release of Unreal Engine 4 two years ago, Epic became even more prosperous. Unless there is no money in the entire industry, as long as there is money, Epic's Unreal Engine 4 will be the first choice for the engine when making 3D games.

Not only games, in the two years since the release of Unreal Engine 4, it has begun to make efforts in many fields such as film and television production, industrial R&D simulation, education and training, digital interaction, etc., and even the construction industry has not been spared. A considerable number of BIM visualization projects and smart city projects use Unreal Engine.

But now, every developer who is trying out the "Taihuang: Mountain and Sea Realm" demo will not believe that Epic has any chance of survival in the face of such an absolute generation gap.

The seemingly indestructible line of defense that Epic used to build with development tools and material libraries now looks so ridiculous.

Exactly the same way Ysou destroyed Chihiro!

Wei Ao glanced at Fang Yu, who was surrounded by other developers, and suddenly remembered the giant pillar rising from the bottom of the sea at the end of the second promotional video.

Improve the gameplay?

Wei Ao's eyes flashed, and he opened the Demo development document attached to the Pad and studied it carefully.

The technology is indeed shocking, and the times have indeed been changed, but as I said, whether a game is fun and whether people can continue to play it is not enough!

Only improve one demo?

Are you underestimating me?

(End of this chapter)