1991 Starting from the core

Su Yuanshan took Claude towards his office.

Although Yuanxin's desktop CPU processor design department has now all moved to Shanghai, Su Yuanshan still has the most complete information and the latest progress in his office.

"We have thought about it very carefully and believe that the speed of floating point operations must be an important indicator to measure the performance of a CPU in the future, and it is particularly affected by frequency."

Claude sat across from Su Yuanshan, staring at the constantly angry steam from the coffee pot, and said seriously: "Cyrix has never had the opportunity to face Intel directly before. In addition to our insufficient resources, there are also reasons why we have not been able to break through the shackles of the architecture."

His words seemed to be explaining why Lao Shanzhai was there before, but in fact... they were complimenting Su Yuanshan.

Because the current Sol architecture was provided by Su Yuanshan, and when he first got the architecture, he intuitively determined that the architecture provided by Su Yuanshan had surpassed the Intel Pentium architecture.

Su Yuanshan smiled after hearing this and slowly poured a cup of coffee for Claude. Then he whispered a few words softly.

“Instruction Per Clock”

Claude was startled when he heard this.

Su Yuanshan thought for a while and said: "Let's abbreviate it and call it IPC. It means the instruction efficiency of each clock - this is the most fundamental indicator. Frequency is just an appearance and just a means to improve IPC."

"Take moving bricks as an example. If a person wants to improve the efficiency of moving bricks, of course he can run faster - this is to increase the frequency. But he can also move two or three bricks at once. Although the frequency has not changed, the efficiency has. It’s doubled.”

Claude slowly fell into thought.

Su Yuanshan was not in a hurry and sipped the coffee slowly.

There are many ways to improve IPC, whether it is superscalar, super pipeline, dynamic branch prediction, or later generations of very long instruction words, all of which can improve IPC.

"You mean, we will focus on super scalar and super assembly lines in the future?" A few seconds later, Claude immediately looked up at Su Yuanshan and asked.

"If you don't focus on a certain direction, you can't give up any direction. If you focus on frequency, you are really going down the wrong path." Su Yuanshan paused and said with a half-smile: "For example, now Intel and AMD, they have gone down the wrong path. evil ways."

Claude was quite speechless.

This is probably the only thing Su Yuanshan dares to say, and he doesn’t dare to say it anyway…

In the past year or so, he and Su Yuanshan have contacted each other hundreds of times via email. He already knew that Su Yuanshan was a truly talented architect in CPU design. Every time when technical issues were discussed, he would mistakenly think that Su Yuanshan was an infatuated architect. Many years of old world.

"Haha, but we can't give up on frequency. 1GHz is indeed a too tempting commanding height, and the performance improvement brought by the increase in frequency is also too intuitive." Su Yuanshan smiled again and said: "Come on, let me show you Ideas for next-generation architecture.”

…

That afternoon, Li Mingliu and Gao Xiaodi also came to the provincial capital from Shanghai. As the current leaders of the CPU team, they are the ones really responsible for front-line technology.

Together with Claude, the four began to discuss ideas for the next generation architecture in Su Yuanshan's office.

At present, the technical foundation of Yuanxin and Xinghai is not enough. Even with the "genius" Su Yuanshan cheating, it can only be said that they can barely catch up with Intel. To truly achieve transcendence, we have to wait until we have the ability to realize "one generation of mass production, one generation of reserves, and one generation of R\u0026D."

Now that both CPU teams have completed the current design, it is logical to get involved in the research and development of the next-generation architecture in advance.

"32-bit processors will be eliminated by 64-bit processors. This is inevitable." Su Yuanshan turned the pencil in his hand. Although he spoke very quietly, his voice was unusually calm: "It doesn't matter what laymen think, but we must Be clear about this.”

"But if you take too big a step, will you drag your eggs?" Li Mingliu flicked the ashes of his cigarette with a solemn look on his face - he learned this sentence from Su Yuanshan. Seeing Claude looking at him blankly, he immediately said it again in English.

Claude laughed: "We all agree that 64-bit processors are the future, but that still requires the support of the operating system... The current operating system is not even 32-bit."

"It's easy to become cannon fodder if you walk too fast." Gao Xiaodi also nodded and smiled: "Don't be unable to use it because our CPU is too advanced... That would be funny."

Seeing that the three of them were cautious about 64-bit processors, Su Yuanshan couldn't help but smile and was quite relieved.

——What he fears most is that because he is too awesome, these buddies will all listen to him and become a monster who draws pictures and claps.

That would be boring.

"Yes, but we can't wait until the demand comes before rushing to tackle the problem. That will only be done by mediocre companies." Su Yuanshan first confirmed the concerns of several people and continued: "I will boldly predict that even if there is no We, at most, will have 64-bit processors in five years."

The three of them were all stunned.

Li Mingliu, in particular, looked at him very seriously.

Li Mingliu knew that once Su Yuanshan started to predict, he would really stop running away.

——Has this guy heard something?

impossible! Xinghai is a purely commercial enterprise there, not a spy company.

"And our development of 64-bit processors also means another thing - we can completely get rid of Intel's X86 stick." Su Yuanshan said calmly.

Whether it is multi-core or 64-bit. These technologies that belong to the "near future" have actually been proposed. Even Intel is already considering 64-bit processors, and AMD is not far behind.

It's just that they are currently considering and "seeing" the future trend, but they have not taken action.

Yuanxin and Xinghai have not seen it at all, or they have seen it and think it is a fantasy. They have long ago... first realize CPU profitability.

This kind of thinking is fine for small companies like CYrix before. After all, the main purpose is to survive.

But for large companies, this is fatal.

Cyrix sold itself to National Semiconductor and then sold it to VIA. In the end, it only produced an MII and M3, and then stopped making CPUs.

The reason is simple, VIA can't keep up.

When other people's forward-looking technologies have begun to be realized and implemented, you are still chasing the old path that others have taken and still cooperating with their old structures. How can you play with others?

"Therefore, I suggest that the Xinghai side begins to conduct forward-looking exploration of multi-core architecture, while the Shanghai Stock Exchange side begins to explore 64-bit instruction sets."

Su Yuanshan paused and said with a smile: "Of course, I will participate in it."

"How much effort does it take?" Claude asked.