Guo Huaisheng's excitement, Wang Anran understands it very well, but when it comes to the future development of Huaxin Technology's graphics card division, Wang Anran cannot make a compromise.

Unexpectedly, Lao Guo was angry and submitted his resignation to the Human Resources Department in the afternoon after the second-generation graphics card was canceled and a resolution was formed, which made Wang Anran a little unexpected.

"Director Guo, do you want to think about it again, I can guarantee that this resolution does not involve your personal intentions. ”

Guo Huaisheng shook his head and said, "I'm tired, I just want to take a break!"

Wang Anran said: "The company can give you a long vacation, when will you adjust your state and come back to work." ”

"Forget it, Mr. Wang, I don't want the salary of a white-collar company. ”

After Guo Huaisheng finished speaking, he nodded with Wang Anran.

Wang Anran stood up, walked in front of Guo Huaisheng, shook his hand, and said: "In this case, everyone has their own ambitions, and I am no longer reluctant, when I want to come back, Huaxin Technology welcomes it."

Of course, if there is any difficulty, you can also call me directly. ”

"Thank you!"

Wang Anran is sincere in retaining, and Wang Anran expresses the greatest respect for every scientific research talent cultivated by Huaxin Technology.

Some people are cultivated into seeds that fall outside to sprout into a tree.

As for Guo Huaisheng, Wang Anran plans to cultivate a technical master of Huaxin Technology.

Besides, what happened just now, Wang Anran more or less felt a little sorry for him.

Seeing Guo Huaisheng say thank you, he walked out of his office.

Wang sat down on the sofa with a butt in disgrace.

After reviewing the whole thing and calming down, I realized that if I had handled it in a different way, it might have a different result.

In the past decades of technical career, Wang Anran has not made the slightest progress in handling human feelings, and every time he finds out afterwards that he has unconsciously caused harm to others.

Wang Anran shook his head and cleaned up his mood, at this time, there was no time for him to reflect.

In the evening of the same day, Wang Anran held a meeting of the graphics card division.

At the meeting, Wang Anran proposed the concept of ALU logical operation unit for the first time.

“...... In the experimental simulation and the daily processing of the display chip, it is easy to find that the vertex rendering architecture has the shortcomings of low efficiency.

And we spent a lot of algorithm optimization, circuit optimization, to solve how to improve the efficiency of the processing unit of the render pipeline.

However, in the face of complex application scenarios, our graphics card architecture has always had inherent deficiencies......"

All of them were the technical core personnel of the graphics card division, and Wang Anran's words immediately resonated with everyone.

Yes, the processing of graphics involves modeling, rendering, rendering, etc., and rendering includes coloring and mapping, etc., each of which has a corresponding processing unit that is responsible for it.

The advantage of this architecture is that it is a full pipeline from modeling to rendering.

However, the problem is also obvious, when the modeling and rendering tasks of the graphics are not input according to the proportion of the actual processing unit processing power, this will inevitably cause the processing unit to complete the work at hand, and continue to wait for the information from the previous process.

As a direct consequence, there are processing units that are idle, resulting in inefficiencies.

This problem is almost unsolvable in the design of vertex rendering architecture, for the simple reason that the chip is dead, and the proportion of various types of processing units is also dead, and it is impossible to achieve seamless processing every time in the face of flexible and diverse processing environments.

Most of you have a question, what to do?

This is also what Wang Anran wants to ask them, he doesn't want to say the answer every time, which will make the technicians of the graphics card division a lot less fun.

"Three days, you have three days, and I need each of you, or your team, to escalate a vision for a solution. ”

Wang Anran stopped the meeting after speaking, he had already assigned his homework, and now it will be up to him to see how far the people in the audience can go.

A group of high-achieving students from key universities such as the University of Science and Technology, the Academy of Sciences, and Qingdao University, which is the intellectual responsibility of the country.

In addition, Wang Anran has already hinted at the answer, that is, the concept of "ALU logical operation unit", Wang Anran is looking forward to how far everyone's brain hole can be opened.

Three days can do a lot of things, Wang Anran naturally will not wait stupidly, at this time, every second is very important to him, while closely paying attention to the progress of the BISM project, Wang Anran put his eyes on the chip processing technology.

There is good news from IBM's chip production line No. 1, where experimenters have stacked two layers of transistors on a single wafer for the first time.

After Wang Anran received the news, he also came to the processing workshop as soon as possible.

"Mr. Wang......"

"Mr. Wang!"

The technicians on the scene could not hide the excitement in their hearts, after all, they have been ahead of other manufacturers in 3D transistors and have made very meaningful attempts.

Wang Anran was also infected by the anger at the scene, and under the leadership of Lu Renjie, the technical director, Wang Anran saw this lens in the sealing and testing workshop.

Lu Renjie introduced: "This NAHD FLASH memory chip, in the same area, using double-layer stacking, can achieve a 70% increase in capacity, and our test data shows that other technical indicators have not decreased significantly. ”

Wang Anran nodded, in his opinion, in the 1.5 micron manufacturing process, the difficulty of stacking 3D transistors should be much easier than the subsequent 7 nanometer manufacturing process for 3D transistor stacking.

This is similar to the difficulty of repairing a table clock and repairing a mechanical watch, the more precise or the higher the accuracy requirements, the greater the difficulty.

And this structural principle can be clearly seen in the 3D transistor model in the laboratory, which is like a child building blocks, layer upon layer of net barriers.

Of course, there are two difficulties.

The first is the design, the 3D stacked chip, its internal layers of transistors have a logical connection, and the logic circuits of different layers also communicate locally or globally through specific buses.

This puts forward new design requirements in the insulating layer, semiconductor layer, metal layer and corresponding wires, which cannot be solved by tiling logic circuits as in the previous 2D plane.

And the second one, naturally, is chip processing, you must know that each process of chip production has a corresponding yield rate.

Suddenly, there are twice as many processes, and the yield rate will naturally drop significantly, which will lead to an increase in costs.

The price is one of the key factors for the smooth promotion of the product.

3D stacking of chips in this structural form is most suitable for memory chips.

In 2019, Samsung, TSMC, and Toshiba have basically achieved more than 90 layers.

It can be said that there is a lot to be done in this technology.

And Wang Anran still wants to rely on memory chips to make a little money.

()