Luo Sheng did a full-body spa after breakfast, and started work around 10 a.m., but the location was in the Science and Technology Complex Building.

Today is the weekend, and for the rest day, Luo Sheng will no longer deal with the company's affairs at this point in time, but do his own private research and tinker with his own black technology.

Unless it's something that is particularly urgent in the company.

"The host (o?▽?) o, the latest batch of hardware equipment of Sugon Technology will be delivered tomorrow. ”

Luo Sheng had just sat down in his private office when Cortana's voice came to his ears.

Luo Sheng nodded and said casually: "Put it in the warehouse first." ”

Cortana: "Is the master going to upgrade Cortana again?" ”

Luo Sheng sat in the desk, leaning back in the chair with his hands on the back of his head and closing his eyes and thinking: "Yes." ”

Cortana: "(?????) The host is amazing, super looking forward to it, huh~~"

Of course, the upgrade of supercomputers depends very much on the progress of semiconductor technology, but more importantly, it lies in the system architecture and algorithms, which will always be the soul and core of computers or artificial intelligence.

To achieve the same instruction, algorithm A needs 100 steps to implement, but algorithm B only needs 10 steps to implement.

Luo Sheng wants to forcibly upgrade his private supercomputer this time, and the hardware must not be able to do it in a short period of time, which is too dependent on the technological development of the semiconductor industry, and now the semiconductor industry represented by Intel has achieved the 45~32nm process level.

Throughout the semiconductor chip industry, the toothpaste factory is the first chip manufacturer in the industry to launch a 32nm process processor, and now the mainstream process is still 45nm, as for the 7nm, 5nm process processor in this era is simply science fiction.

The current chip process in China is difficult to describe, not to mention.

Obviously, hardware upgrades should not be thought about for a while, and the performance of private supercomputing can only be improved as much as possible from the two ways of system architecture and algorithms on the basis of existing hardware.

It's very difficult for outsiders, it's super difficult, but for Luo Sheng, developing a new algorithm is a very simple thing, and even one person can do it, and now Xiao Na can also assist him to further improve development efficiency.

After this upgrade, the performance of the private supercomputer in the science and technology complex building will be increased by about 10 times, from the petaflop to the petascale.

It can be called black technology.

At present, Luo Sheng's private supercomputer "Nafu", which is named after Xiao Na's official name, has a performance of 230 trillion operations per second.

In other words, Luo Sheng will reach at least 2000 trillion operations per second after this upgrade, and it is estimated that it will take about a month and a half to complete the upgrade plan, and the hardware needs to be replaced, and the total budget for this major upgrade will be about 120 million yuan.

At that time, it will surpass the current U.S. Department of Energy's "Jaguar" at Oak Ridge National Laboratory.

In the 2009 list of the top 500 supercomputers, the "Jaguar" currently topped the list with a speed of 1,800 trillion operations per second, and is a civilian computer mainly used to simulate climate change, energy generation, and other basic science research.

Of course, Luo Sheng's private supercomputer can also be able to perform these tasks without pressure, but the fundamental purpose is to upgrade and optimize Cortana, which is the first priority, and Cortana's upgrade and optimization will bring about an improvement in overall benefits.

It has an unparalleled significance.

Upgrading and optimizing for Cortana is only part of Luo Sheng's personal work, which will take more than a month to complete this round of updates and upgrades, and he can't spend all his time on it, otherwise it can be done in half a month.

Luo Sheng has not officially stepped down as CEO of the two companies, and he has to participate in some major meetings, but most of his time has been spent on scientific and technological research and development.

Since 2009, Luo Sheng's work has basically focused on development.

Since May, he has been separated from the cloud computing development team of Bluestar Technology, and he has left the BlueStack open source cloud platform behind, basically handing over to the following technical team and Lu Qi to operate this business segment.

Luo Sheng has focused on the research and development of the third generation of the Côte d'Azur mobile phone, and this year's Cte d'Azur will release the latest generation of smartphones.

……

Lens Semiconductor, Mobile Processor Division.

"The introduction of a high-K insulating/metal gate configuration in the 45nm process is a milestone achievement and Intel has developed this technology. Both of these technologies are designed to solve the same problem: how to ensure that the gate works efficiently at a very small size. ”

The person who spoke was Xu Zhijun, who is currently serving as the CEO of Lens Semiconductor, but he does not have much say in the mobile processor division, here Luo Sheng has the final right to make the decision, and it is also a more privileged division of Lens Semiconductor, and there is an independent R&D building, which has been agreed when the two sides jointly established this company.

But it's critical that both sides have broken through talent barriers and patent barriers.

Luo Sheng was also present at this meeting, as well as several scientists, research scholars, including some mathematicians, etc., all gathered in this conference room to discuss.

Now the focus of the discussion is the processor of the third-generation Teal mobile phone, and the chip processor of the new product uses a 45nm process technology, which has been determined, and many technologies have been done.

A senior research expert looked around and said: "Our R&D department came up with the idea of using metal as a gate, because metal has an effect called mirror charge, which can neutralize the influence of the dipole in the insulation layer of high-K materials on the channel and Fermi level. In this way, the best of both worlds, but what exactly these metals are, no one except the Western high-tech companies that have mastered the technology, the outside world knows, it is their trade secret. ”

The insulation layer is the most critical of all the components of a transistor, and its function is to isolate the gate and channel. Because the gate switching channel is carried out by an electric field, and the generation of the electric field is achieved by adding a certain voltage to the gate.

Ohm's law tells humans that where there is voltage, there is current.

If current flows from the gate into the channel, what switch is there to talk about?

It's been leaking for a long time.

Another researcher at the conference said: "Silica is good, but it also has problems when the size is reduced to a certain limit. In this process, the electric field strength is constant, so from the perspective of energy bands, because of the fluctuation of electrons, if the insulation layer is very narrow and narrow, there is a certain chance that the electrons will undergo a tunneling effect and cross the energy band barrier of the insulation layer, resulting in leakage current. ”

As he spoke, the scholar gestured: "It can be imagined that passing through a wall higher than oneself, the magnitude of this current is negatively correlated with the thickness of the insulation layer, as well as the 'barrier height' of the insulation layer. Therefore, the smaller the thickness, the lower the barrier, and the greater the leakage current, the more detrimental it is to the transistor. On the other hand, the switching performance of the transistor, the operating current, etc., all require a large insulating layer capacitance. ”

"It can be seen that there have been a bunch of contradictions in the design goals here, whether the thickness of the insulation layer should continue to shrink, in fact, before this node, the silica has been reduced to a thickness of less than two nanometers, that is, the thickness of more than a dozen atomic layers, and the problem of leakage current has replaced the performance problem as the number one enemy."

At this point, all the scholars in the conference room fell silent.

Sitting here are the smartest humans, and of course they start to figure out how to solve problems.

Humans are greedy and are not willing to give up the performance enhancement of large capacitors, nor are they willing to risk leakage.

In the current situation, a material is needed that has a high dielectric constant and a high band barrier, so that the capacitance can continue to be increased without shrinking the thickness.

In other words, it is a material that protects leakage current and improves switching performance.

In the academic world, in recent years, various brain-opening new designs have been proposed, such as tunneling transistors, negative capacitance effect transistors, carbon nanotubes, and so on.

But in fact, all these designs are basically in four directions, materials, mechanisms, processes and structures.

At this time, Xu Zhijun glanced at Luo Sheng, then looked around at everyone and said, "What about graphene transistors?" The idea of graphene as a channel is the second item, which is transportation, because the electron mobility of graphene is much higher than that of silicon. ”

As soon as these words came out, a scholar shook his head and said: "This magical material is indeed very imaginative, but the problem is that graphene transistors have not made any progress, most of them are conceptual and theoretical stages, and graphene has a hard flaw, that is, it cannot saturate the current, but I noticed that some people in the Western academic community said that in the future, it may be possible to regulate the band gap of graphene to open and close, no longer just zero band gap, in any case, graphene is a very future material, but it is too early to talk about this." ”

Xu Zhijun immediately looked at Luo Sheng and asked, "Mr. Luo, what do you think?" ”

At this time, Luo Sheng was looking at the data information and a theoretical model on the big screen of the conference, and everyone also looked at him, who stared at the data and function model on the screen and said:

"Listening to your discussion, I've been thinking about the question, since electrons move in the channel, why do I have to leave such a large depletion layer under the channel? Of course, I know why, because the physical model needs this area to balance the charge. However, in a short-channel device, there is no need to put the depletion layer and the channel together and wait for the leakage current to flow through in vain. ”

A group of scholars looked at each other, a question they had never thought about.

Luo Sheng snapped his fingers, motioned to a scholar, and then stared at the data model on the big screen of the conference and said:

"Modify it according to what I said, take this part of the silicon directly out of me, replace it with an insulating layer, and below the insulating layer is the remaining silicon, so that the channel is separated from the depletion layer, because the electrons come from the two poles, but the two poles and the depletion layer are separated by the insulating layer, so that in addition to the channel, there will be no leakage."

Everyone's eyes lit up, as if they had seen a cloud and fog.

Everyone is the top group of people with the highest IQ among human beings, and Luo Sheng understood it instantly when he said this.

Xu Zhijun said excitedly: "Brilliant idea, Mr. Luo is worthy of being a design genius, in a word, it is deeply important!" How come so many of us didn't think of it? ”

……

(End of chapter)