1991 Starting from the core

Watching Zhang Ruiping go out to get the information, Su Yuanshan first glanced at Zhou Qiang, then at Jiang Wenxi, and raised his eyebrows slightly.

Are you holding back your ultimate move?

...No. Zhou Qiang shook his head immediately.

Jiang Wenxi was also a little surprised. He came from a chipset background and was not involved in the architecture of 3D accelerator cards. In addition, he was busy integrating recently, so he didn't know what Zhang Ruiping was doing.

Of course, in front of Su Yuanshan, no matter what, he still wanted to defend Zhang Ruiping. So he smiled and said: Maybe Ruiping and the others have new ideas that haven't been finalized yet, so let's take this opportunity to let Mr. Shan check them.

Don't take it seriously. Su Yuanshan waved his hand and said modestly: The main reason is that the opponent is too strong. 3DFX, including me, did not expect that TNT2 would usher in a word-of-mouth explosion.

Speaking of this, Su Yuanshan felt a little emotional, even sighing.

Lao Huang, indeed Lao Huang.

After forcibly intervening in the development direction of 3DFX and choosing to embrace Microsoft's DirectX from the beginning, VOODOO has indeed been in a dominant position in the past few years.

However... NVIDIA was like the grass buried under the boulder after all. Although it stumbled, it still survived stubbornly - not to mention this, just like history, TNT was still born as promised.

If TNT released last year was just a little splash, then TNT2 released in April this year suddenly detonated the market, gaining recognition and reputation from players, and overwhelming publicity followed. , the civilian title of King of Cards was put on his head.

Su Yuanshan's memory of the history of graphics cards is not very deep, but he knows that in history, Lao Huang relied on the great success of the TNT2 series to launch the GeForce series, thereby defeating 3DFX, S3 and other opponents in one fell swoop, and establishing his status in the world. .

This aroused Su Yuanshan's high vigilance and prompted him to leave for HK.

TNT2's reputation is booming? Zhou Qiang frowned in disdain after hearing this: Mr. Shan, don't keep staring at the game! You also believe that in the future...

Zhou Qiang saw Su Yuanshan's brow suddenly frown as he spoke, and he immediately realized that even at Yuanxin, artificial intelligence based on heterogeneous computing was an unofficial project and was a future-oriented technology.

So Zhou Qiang immediately stopped and laughed twice: Anyway, the application scenarios are not limited to games.

Of course it's not just games. Su Yuanshan was a little annoyed that this guy was a big talker, but he came here just to determine the future path and tone of GPU development. Heterogeneous computing will even become the main slogan of silicon core in the future, so he glared Zhou Qiang glanced at him and said: Mr. Xi asked you to come here just to develop into professional graphics cards. But whether it is the personal PC market or the game market, these have just begun to rise...

While talking, Zhang Ruiping returned to the office holding his notebook.

Su Yuanshan also looked at Zhang Ruiping again.

Mr. Shan, this is actually the structure that our team has been conceiving half a year ago.

Zhang Ruiping opened his notebook, opened a document and handed it to Su Yuanshan: Our idea is actually the same as that of the current mainstream graphics card core manufacturers. In addition to introducing a stream processor, we also need to provide a higher transmission bit width between the memory and the memory. , at the same time, our architecture will focus on the two aspects of geometric conversion and lighting processing capabilities to reduce the burden on the CPU - frankly speaking, we believe that the only way to hand over all the work that was previously done by the CPU to the graphics core , can it be called the GPU that friends and businessmen are clamoring for now!

This is our immature thinking about the future architecture. The biggest problem at the moment is that we are still demonstrating the difficulty of implementing the design required for the architecture as well as the instruction set and related protocols.

Having said this, Zhang Ruiping glanced at Zhou Qiang again: Also, we must also consider the expansion plan for heterogeneous parallel computing proposed by Brother Zhou. So... the progress is a little slow now.

Su Yuanshan listened and looked at the design draft. Gradually, a smile appeared on his lips.

I didn't realize that this Zhang Ruiping really had some ideas.

As he said, although the GPU has been shouting for so long, the real sign of the GPU is that the GPU has indeed taken over the two most typical functions in 3D application scenarios, geometric conversion and lighting processing, which were previously assisted or even handled by the CPU.

Since it is called a graphics processing core, then in the resource-hungry graphics processing scenario, the graphics card should be the absolute and only main force.

This can be called the core.

And Zhang Ruiping is able to realize this, obviously he has two brushes.

But when he heard Zhang Ruiping talking about Zhou Qiang's team again, he caught a glimpse of Zhang Ruiping looking at Zhou Qiang from the corner of his eye...

Su Yuanshan immediately felt that Zhou Qiang's coming here seemed to have aroused resistance among the former Silicon System employees.

After realizing this, Su Yuanshan quickly played on the alarm clock what he had seen, heard and had conversations with from the time he entered the door until now, and then he became more and more sure of his judgment...

Hearing what Zhang Ruiping meant, it was obvious that he was telling himself that Zhou Qiang's presence here was interfering with the normal research and development progress.

…

After pondering for a moment, Su Yuanshan raised his chin to Wen Xiaoqian, who was sitting at the edge: Xiaoqian, close the door and let's have a small meeting.

After he finished speaking, the three people in the office looked at him at the same time.

Su Yuanshan put down his notebook, looked at the three of them, and laughed: We all have a technical background, including Mr. Jiang, and we have not left the frontline of research and development, so I will speak directly.

Three points. First, brother Ruiping, your architectural ideas are correct. Your understanding of GPU is also correct. I take back my previous suggestion that you adopt the 3DFX architecture. I will instead support your current architectural ideas.

Hearing Su Yuanshan firmly support him at the beginning, Zhang Ruiping was startled for a moment, then clenched his fists, and his face suddenly turned red with excitement.

He opened his mouth, but didn't know what to say. Finally, he held it in for a long time before he finally said: Thank you, Mr. Shan, we promise to complete the task!

Well, I was narrow-minded before, so I hope you won't be offended.

After smiling and waving his hands, he said seriously: The second point is why I suggest using the same architecture as 3DFX - because in the next two years, the competition in the graphics card market will enter a long-term, continuous, and cruel competition. The process is not brought about by the concept and technology upgrade of the graphics card, but is determined by many factors such as PC, Internet, games, and even the level of world economic development.

If we can't enter the market as quickly as possible and get a share of the cake, then when the brand effect appears - and the brand effect has actually already appeared - it will be difficult for us to enter the market again.

At that time, we may have to give up the mass market and switch to professional application scenarios - which will lead to the third point.

Su Yuanshan pursed his lips, glanced at Zhang Ruiping, and then his eyes stopped on Zhou Qiang's face: It is a heterogeneous parallel computing solution.

Heterogeneous parallel computing is Pandora's laboratory's vision for the future era of big data and artificial intelligence. It will definitely come!