…………

As Fang Nian's grandmother grew older, she gradually became more talkative.

This is a common change that most women undergo when they enter their seventies or even their octogenarian years.

To use a not-so-nice word, it’s called: broken mouth.

I experienced it once every year in my previous life, and sometimes I was impatient. But when I experienced it again in this life, it felt more or less cordial.

I am very happy to chat with the old man.

After Grandma Fang Nian came to apply this time, she also brought about a more interesting change -

Mr. Guan Qiuhe, who hated Tangli dialect deeply, inexplicably entered the process of accepting it.

That is acceptance, integration and learning.

After all, it is obviously easier for her, an extremely young girl, to learn Tangli dialect than for an old person to understand and speak Mandarin.

In the living room, an old man and a young man were chatting away about something.

Fang Nian's cell phone rang.

After checking the number, Fang Nian answered the phone and said with a smile: "Academician, have a great weekend. Didn't you spend the holidays with your wife?"

"Hello, Mr. Fang, I am no longer as romantic as you young people as I am getting older." Chen Jianye said with a smile.

Then he reacted: "It turns out that today is Chinese Valentine's Day, so you didn't bother Mr. Fang."

"That's not possible. No matter how busy you are, you still have time to answer the phone." Fang Nian smiled, "If the academician has something to do, it's okay to say it."

Chen Jianye didn't talk nonsense and got down to business.

"Mr. Fang, it's like this. There are two things I need to report to you. First, I will briefly report on the progress of Shenlong 2;"

“Because Shenlong 2 is still based on the 28nm process, according to the current form, it can only be found by TSMC, so there is no need to change the PDK (Process Design Kit);

Secondly, referring to last year’s internal chip design planning and development plan within the laboratory and this year’s independent chip R&D and design development outline, Shenlong 2’s CPU design highlights the design concept of one large core + several small cores;

At present, Shenlong 2 has basically completed the redesign of all units, and is in the process of combined design and simulation verification;

The progress is slower than expected. The internal estimate is that the first actual tape-out will not take place until the beginning of October, mainly because of the factors of following the company's internal coordination strategy and cooperating with the EDA research and development work of Yangcheng Zhuyan Laboratory."

Having said this, Academician Chen Jianye paused briefly.

Then added: "In terms of the large core design of Shenlong 2's CPU, no modifications have been made to the arm instruction set architecture, but many independent modifications have been made to its micro-architecture; ①

Some execution units have been added, such as the ALU (Arithmetic Logical Operation Unit) that can execute multiplication instructions, and the read-write unit (LSU) has been creatively added;

Through simulation verification, it is optimistically estimated that the read and write performance will be improved by more than 50%, the IPC performance will be improved by more than 20%, and the power consumption will be reduced by 10%;

These design changes are a major part of Shenlong 2’s research and development expenditures. For this purpose, we specially invited on-site technical support from Synopsys and Cadence, um..."

"Zhu Yan doesn't have the technical solution capability in this area yet."

"This is also an important reason why it takes two months to tape out after the design has been basically completed."

After hearing what Academician Chen Jianye said, Fang Nian smiled and said: "Academician has worked hard, everyone has worked hard, and they have done a good job."

"I also ask academicians to understand from the perspective of Frontier. The coordination strategy of internal laboratories must be carried out. This is also the reason why Frontier has established so many laboratories with clear division of labor;

The road to independent research and development of integrated circuits cannot be a breakthrough in a single category."

After hearing what Fang Nian said, Chen Jianye quickly said: "Mr. Fang misunderstood, we are not complaining.

Shenlong 1 has just been put into mass production, and it needs more time to achieve a breakthrough in market share. The premature launch of Shenlong 2, which has better performance, will affect the market share of Shenlong 1;

Personally, I have more or less experienced the difficulty of not having an industrial ecosystem."

"Thank you for your understanding." Fang Nian smiled.

Speaking of which...

These two documents, the chip design planning and development plan and the chip independent research and development, design and development outline, were entirely written by Fang Nian.

At most, Wen Ye, Gu Yu and the others helped with the formatting and text proofreading.

Fang Nian really doesn’t understand chip design work at all.

But he knew the subsequent development pattern.

Apple started independent research and development on the A6 chip, and finally tried to adopt a development route of one large core with multiple small cores, and repeatedly led the core processing power of the Android system, and then was widely accepted;

In the end, all mobile processing chip IP solution providers arm had to launch the new ultra-large core architecture of Cortex-X1.

Um……

Apple's self-research only conducted self-research on the micro-architecture and did not make any changes to the instruction set architecture.

After knowing the development pattern, and in turn understanding it through public information and the summary reports of so many experts before the establishment of Baize Laboratory, Fang Nian can certainly write a very effective forward-looking development outline.

Anyway, the development of companies all over the world is similar, and people who customize company strategies basically don’t understand the core technology.

Fang Nian is not a special case of genius. He just understands the future and then relies on reality and some personal development concepts to formulate a more forward-looking development plan.

After experiencing the missing parts of the last on-site meeting, it is reasonable for Chen Jianye and Fang Nian to explain the breakthroughs of specific modules in more detail.

This does not require technical knowledge, and some proper nouns are not that difficult to understand.

One thing is that the large amount of R&D expenditures of Baize Laboratory on Shenlong 1 and Shenlong 2 were given to some foreign companies.

For example, the companies Chen Jianye mentioned, synopsys and cadence, are both excellent manufacturers of EDA software, and both provide excellent front-end and complete integrated circuit design and verification tools.

Especially the verification simulation tool platform system.

It is an important tool for formal pre-silicon evaluation testing, also called pre-silicon evaluation testing in the industry.

This type of simulation tool platform is relatively excellent and can be so expensive that the entire company's book assets of Loongson can only be used a few times.

The tools of synopsys and cadence are of this type.

The price of technical support for a resident engineer is exorbitantly high, at US$2,000 per person per day, excluding food and accommodation.

In the past, in order to pursue efficiency, Bai Ze would call a team directly from their head office, and it would often take 20 or 30 people to work for half a month.

Ten million yuan was gone in the blink of an eye.

It does not include some special technical requirements.

At present, the contract value of Baize Laboratory to these two companies alone has exceeded 100 million yuan.

In addition, the high-performance computing platform built to complete simulation verification also requires an investment of 100 million yuan.

It takes several hours for the rendering CPU to actually run for a few seconds, which means the performance is not enough, so Baize has been investing in building a high-performance computing platform.

In the near future, there are plans to cooperate with Bi Fang to upgrade to a cloud computing platform to save costs.

Simulation work almost runs through the entire design process.

After all, you can’t just do a little bit of design and then go to tape-out without finalizing anything. You have to conduct various verifications and make sure that there are basically no problems before tape-out.

The investment in simulation is huge, and tape-out is not cheap.

Anyway, these are the reasons why it is so expensive to be excellent in the field of chip design.

Chip design is not about using a computer to draw two pictures. It involves a massive project. Even if you are using the ARM public version architecture, there is a lot of other technical work.

But from a more fair perspective, there is a priority in terms of technical content...

The design progress of Shenlong 2 is quite good.

Fang Nian didn't waste any more nonsense and said with a smile: "Academician, there is some good news, please tell me quickly."

After all, Baize Laboratory currently only has two major projects: research on Shenlong 2 and the confidential laboratory.

So Fang Nian is quite looking forward to it.

Sure enough, Chen Jianye said: "The confidentiality laboratory has made some breakthroughs, but..."

Having said this, Chen Jianye hesitated for a moment.

"It cannot be regarded as a breakthrough in the independent design of mobile phone processing chips. To be precise, it is that the autonomous instruction set architecture that has been initially completed before... has some problems."

Chen Jianye said with some embarrassment: "We have carried out a lot of verification and revision work, and even completed the preliminary design of the microarchitecture based on this instruction set architecture, and even a certain amount of simulation work;

Ultimately it turned out... that it was really not suitable for small mobile devices,"

"Especially since we started cooperating with Loongson last month, after some rationalization and adjustments, we no longer have access to mobile phones and can only be used for computer CPUs..."

Fang Nian: "!!!"

-

①: The difference between instruction set architecture and microarchitecture has been explained before. The simplest thing is that the former is a specification and the latter is a method;

If the former changes, the entire system will change, which is why Wintel is so invincible;

The latter change mainly affects IPC (referred to as architectural performance);

Let’s list another formula: CPU performance = clock frequency (MHz or GHz) × IPC (the number of instructions executed by the CPU in each clock cycle);

The clock frequency mainly depends on the production process, that is, the number of nanometers; therefore, the excellence of IPC performance becomes the difference between the performance of CPUs of different manufacturers with the same process. What affects IPC performance is the CPU core, that is, the micro-architecture.

======
Chapter completed!