1991 Starting from the core

The seventh floor of the science and technology park office building.

In Shanghai, where every inch of land is at a premium, Yuanxin Science and Technology Park and its wafer fab can still enjoy over two thousand acres of ultra-high welfare benefits. This can only be said that Shanghai’s support for the semiconductor industry is implemented in action.

Gao Xiaodi has a deep obsession with the big office - he is the first batch of graduate students to enter Yuanxin - in an era when college students are called the favored ones, graduate students are the favored ones. But he still could only work in the workstation area with many undergraduates and even junior college students - even if Li Mingliu left later, he still stayed in the workstation area and led a group of idiot recruits to be trained in the most basic EDA operations.

After he began to be responsible for the CPU architecture design, although he had his own office, he still couldn't worry about the team and continued to take root in the work area, staying with ordinary employees...

It wasn't until he arrived in Shanghai, put the CPU on the last bus, and actually took charge of the CPU department that he got himself an office large enough to show movies.

But now, this office, which is even more exaggerated than Su Yuanshan's, has been serving as the conference room of the architecture design department.

A group of architecture engineers were puffing away in Gao Xiaodi's office, looking at the architecture concept diagram on the blackboard and discussing softly.

Gao Xiaodi twisted his neck, took a long puff of cigarette and waved his hand: Let's do this for today. I think everyone is confused. How about I give you guys a few days off? You go on a trip. lock up?

Lao Gao, you said this. Ma Jingcheng, who was sitting on the sofa, was overjoyed: I'm afraid Lao Li won't approve it, right?

Ma Jingcheng is considered a turtle master - he previously studied at the National University of Singapore. After graduating with a master's degree, he joined TI's semiconductor factory in Singapore, and then as expected, he found an organization in the factory - he and Lao Zhang worked together A group of people recruited from China took over.

As Zhang Rujin came to China to build a factory this time, he also returned to China. But he did not choose to join the wafer fab, but entered the CPU design department of Yuanxin.

In his words, he still hopes to do something creative.

Facts have also proved that Ma Jingcheng is indeed very good at CPU logic design, so he quickly grew into one of the leaders of the Yuanxin CPU department.

Can I just give you six and a half days? Gao Xiaodi put down the cigarette butt and said with a flat mouth: Be more flexible, brother.

A group of people in the office laughed.

——The rules of the science and technology park require the approval of the person in charge of the science and technology park for a holiday of more than seven days. Within seven days, all that is needed is the nod of the department leader.

Then are you leaving?

Gao Xiaodi shook his head: I won't leave, I'll keep an eye on you. Damn it, the good news from Claude's side is so frequent that we are too passive. If we don't use some black technology, we won't be like dwarfs when we meet in the future. ?”

After he finished speaking, he heard a joking voice coming from outside the door: Who wants the dwarf to wait?

…

Three minutes later, Su Yuanshan waited until the smoke in the room cleared before walking into the office.

Senior brother, are you filming the Peach Blossom Show? Su Yuanshan stood with his back to the window, with a wry smile on his face: I'm afraid the smoke in your room must be evil.

Hehe... I'll get a ventilation fan tomorrow. Seeing Su Yuanshan standing, Gao Xiaodi was too embarrassed to sit down, and instead leaned against the desk with Li Mingliu.

Don't you know you should smoke less? Su Yuanshan said helplessly, and glared at Li Mingliu and a few others: And you too, you smokers... If you really want to continue smoking, then I advise you to change your career. Build the legendary nanorobots and conquer lung cancer.

Hehe... Everyone laughed.

Finally, when the smoke was almost gone, Su Yuanshan sat down again.

He looked at the blackboard in front of him and saw two simply drawn squares representing the core, with various protocols and parameters written in the middle. Of course, it was more about some thoughts.

Are you working on dual-core? Su Yuanshan looked around and asked with a smile.

Yeah. Gao Xiaodi wiped the back of his head and smiled: We simulated it carefully. According to the current situation, even though the highest main frequency is only 100MHz, it is estimated that in a few years, the main frequency of the CPU will be We are heading towards 1G... but the main frequency has an upper limit after all. If you want to improve execution efficiency, dual-core or even multi-core should be the inevitable choice.

And since it is an inevitable choice, we can take the time to do it first so as not to panic later.

After hearing this, Su Yuanshan nodded and said, That's the truth. So what's the question now?

There are many problems now... First of all, we are not sure whether to adopt dual-Die packaging or dual-computing core packaging. Secondly, the connection architecture in the middle is also very complex, and requires a dedicated data processing protocol and arbiter. And according to the current With the current fab process, even if we design it, it will be difficult to tape out...

Well……

Su Yuanshan looked around and saw that everyone's expressions were not relaxed. They must be really entangled.

After pondering for a few seconds, Su Yuanshan decided to point these people in the right direction first - Gao Xiaodi was able to think at this time that dual-core and even multi-core are inevitable in the future and involved in it in advance. This is something Su Yuanshan greatly appreciates and is very happy to do. Accept and even support. But as someone who has witnessed the birth of glue dual-core... I know that in terms of implementation difficulty, dual-Die packaging, that is, glue dual-core, is extremely tempting.

If Gao Xiaodi chooses the glue dual-core, it means that they will waste several years.

Su Yuanshan doesn't have that much American time to waste with them - and he might even destroy them.

Let's discuss it first. Su Yuanshan stood up, walked to the blackboard, and picked up the chalk.

Theoretically, direct packaging of double DIE is very trouble-free in design, but it does not have the potential for sustainable development.

Gao Xiaodi immediately asked: Why?

Su Yuanshan smiled and said: Because after the seemingly simple method, there are a series of problems that need to be solved. The first is that additional arbiters are needed to configure the cache and instructions. Secondly, it will increase the occupancy rate of the wafer, and then it will also Increased power consumption and heat dissipation pressure - we can use simple calculations to calculate the power consumption when the main frequency enters the GHz era.

Now, our transistors are about two million, and the power consumption is not high - but according to the existing technology prospects, even if there is no leap in process and design, when we need to rely on dual-core to improve execution efficiency, Transistors will probably reach more than 30 million. By then, a CPU will not be able to reach hundreds of watts...

The most important thing is that integration is the most test of wisdom and the most beautiful. But before that, I have an idea.

Su Yuanshan looked at everyone: Let's squeeze out the potential of traditional single-core CPUs first.

Wait a minute. Gao Xiaodi suddenly interrupted Su Yuanshan.

Su Yuanshan raised his head and signaled Gao Xiaodi to continue speaking.

Gao Xiaodi pursed his lips, his eyes shining: To squeeze out the potential of traditional single-core CPUs, I also have an idea - one core can be used as two.

Su Yuanshan blinked: You said it.