.

To put it bluntly, core processors are integrated into the same core.

The photonic computer project team mastered the boss's idea of ​​developing optical brains, which was to create a supercomputer that surpassed modern computers and establish Universal's leading position in computer and scientific research.

Relying on the original idea, it is obviously impossible to meet Xiao Qiang's requirements.

After thinking about it, they saw the company's supercomputer system, and suddenly had an idea and decided to use multi-core manufacturing method to integrate dozens of processor cores to maximize the performance of the chip under the existing technical conditions.

Speaking of which, with the rapid development of semiconductor technology, the speed of processors is getting faster and faster.

In the entire computer, although the processor can calculate tens of millions or hundreds of millions of times per second, the processor does not exist alone. It needs to exchange data with memory and temporarily store some of the calculation results in memory, so that the temporary calculation results can be continuously called to complete the calculations.

Other devices such as processors and memory need to exchange data through the system bus.

But it is not easy to increase the bus frequency.

From the early 33mhz, 50    .6    .     00                                                                                                         �

To solve this bottleneck, one is to integrate memory in the processor, connect the two directly, and use a direct connection to bypass the limitations of the bus. The other is to completely change the bus transmission method and replace slow circuit interconnection with faster optical interconnection.

Among these two methods, the direct memory of the processor has long been adopted, and this high-speed memory is called cache. However, this increases the difficulty of processor manufacturing and increases the cost; secondly, it still needs to exchange data through the bus and motherboard memory, transfer it to related hard disk storage, or transfer it to various devices such as graphics cards, sound cards, etc. The bus still stubbornly limits the speed of the computer.

This is a method that treats symptoms but not root causes.

Therefore, the most thorough solution is to replace the bus.

The optical brain uses all-optical computing, all-optical signal transmission and storage, which can be said to have completely updated the original computer components.

The photonic computer project team led by Xiao Jin used to try to be in place at one time, using all-optical signals to manufacture all-optical signals, but failed repeatedly. This time they changed their course. They admitted that under modern technology, it is not realistic to realize all-optical technology. Therefore, they turned around and used the achievements that have been achieved in recent years to replace existing computers step by step, restricting the speed of relevant parts.

Forty-core processors are an attempt to improve existing chip technology.

Integrating forty processor cores into one chipset can effectively avoid bottlenecks in data transmission and maximize the performance of the chip.

The photon computer project team discussed for a year before deciding the final plan. The researchers gathered the hard work of everyone and worked hard to design the most efficient data processing assembly line structure, and designed the instruction set one by one. Finally, they drew the relevant circuit diagram.

The only trouble is that the forty cores are spread out on a silicon wafer base. The area is too large.

They use a layered method to make ten core processors per layer. Then they make a data transmission layer above it, use the laser transmission technology accumulated over the years to create a laser data transmission layer through integrated circuit manufacturing methods. On the laser data transmission layer, a cache layer corresponding to each processor core can be created, and each core can build a single g

In this way, each group will be composed of a core layer, a laser transit layer, and a cache layer. Four sets of stacks form a complete processor chip.

Optical interconnect speeds up data throughput, and the data transmission speed of a single processor core reaches an astonishing 10t per second. When all chips perform their maximum performance, their peak data processing speed will reach a record 20 trillion times per second!

If it is truly manufactured successfully, a processor chip will be twice as good as the supercomputer system currently in use, and the usage efficiency will be much higher!

This idea is very clever, but it is difficult to implement it.

The production of a single core and a photoelectric conversion laser data transmitter is complete, and there is no problem with the corresponding memory stack, but it is more than three times more difficult to complete at the same time.

What's more, ten cores are needed to be made in a layer of silicon crystal base, with two corresponding layers of laser data transmission, memory and other components on it.

Each core and the corresponding laser dataset and memory layer are composed of hundreds of millions of components. As long as there is an operation error in the middle, all efforts will be wasted.

When Xiao Qiang was busy building a special laboratory, Xiao Jin and others tried to make forty core processors. However, they were scrapped continuously and failed in one success.

Austin, who often came to the laboratory to watch the progress, saw that they were wasting their heartache, and his muscles on his face were pounding.

After Xiao Qiang finished the reconstruction of the special laboratory, he turned around and came back to the optical brain project team. Xiao Jin and others were frightened by the failure in the front. They took the initiative to shelve the manufacturing of forty core processors and decided to use a two-step approach to solve this problem.

One is to first make a single-core chip, which is a conventional chip with built-in large-capacity cache and uses photoelectric conversion laser transmission.

Even with a single core, complete photoelectric conversion decoration can still increase the performance of hybrid chips by more than ten times as before. If the connection with other devices is all optical interconnected, the speed of the chip may also be able to allow researchers to accumulate technology and lay a solid foundation for the next step of making more core processors.

Another is to abandon the photoelectric conversion layer and use a multi-core processor based on conventional technology. The photoelectric laser converter is used for data transmission between the processor and other devices such as memory, instead of the previous bus structure.

Xiao Qiang made a decision without thinking about it: first make a single core and integrate it into the laser data transmitter with the peripherals.

Only by reducing the connection between the processor and the photoelectric converter can the ability of optical interconnection be maximized.

This difficulty is much more difficult than the plan proposed by the project team.

But this is necessary again.

With the current technical limit, it is estimated that optical interconnection between chips and other devices can only be achieved, but in the future, it will still develop towards optical processors. Instead of cramming at that time, it is better to accumulate relevant operating experience early and prepare for subsequent research.

The project team gritted the bullet and took more than fifty days to create a processor that roughly meets the requirements for combining laser converters.

For this processor, countless waste products appeared, but the coat was collected through instruments and specially produced data, and a large amount of experimental data was collected.

Due to the use of various data recording software, coupled with the data collected from the electronic sensing device of the tester, the supercomputer converts all experimental steps into 3D experiments in a timely manner.

Xiao Qiang instructed to suspend the chip and only do it, and conduct strict analysis of all data.

Based on these 3D images, the software personnel of the project team decompose the production process into tens of thousands of links based on these 3D images. The environmental data changes in each production process were repeatedly compared and analyzed. The optimal set was summarized, and the alternative method was used to integrate it into a complete operating process system and input it to the robot in the special laboratory.

After more than a year of use, the performance of the test robot has been continuously improved.

The software personnel made countless updates and modifications to the driver and the active judgment program, allowing the robot to have more active intervention capabilities on the basis of strictly following the basic data requirements. It can automatically enter the next step when the instrument and the various main test data collected by the experimental subjects meet a relatively optimal condition, changing the robot's rigid operation, resulting in the failure of the experiment.

In addition, software personnel have added learning mechanisms to the robots, which can master the changing patterns of various data during repeated operations.

After that, according to this rule, and according to the best operating conditions in advance, proactive operations are taken to ensure that each operation step is implemented in a relatively optimal data environment, improving the success rate.

The project team members had fierce debates on each link and put forward their own revisions.

In this special laboratory, the robot operating program solicited everyone's opinions and revised it repeatedly, and finally confirmed it to be completed in two months.

After entering this program into the robot, the first actual operation was made, all members of the project team were given an unexpected surprise: fifteen experimental robots formed a production pipeline, and a total of 100 special processors made on the same chip were 63 pieces, with a success rate of 63%!

Such high yield rate is extremely high quality that cannot be matched by manual control!

Of course, a single-core optical interconnect processor does not have practical effects, which only means that everyone has finally got the basic qualification to enter the next step of experiments.

Immediately, the project team decided to produce dual-core chips to prepare for the future production of eight-core and thirty-two-core processors.

But Xiao Qiang's words made the project team confused: Go directly to the forty core!

When the success rate on a single core has not reached 100%, it is only 40 cores. Even if the distribution production method is adopted, the production process is continuously improved, and the production process is produced by a special laboratory, the success rate is so low that it makes people feel numb. I don’t know that it will take tens of thousands of times to produce a qualified product!

Everyone told their reasons, believing that it should not be rushing to achieve success, and it would be better to proceed step by step, but Xiao Qiang was unmoved and forced to go directly to the forty core.

Failure, failure, failure...

After repeated failures, even the members of the project team felt desperate, but Xiao Qiang was not discouraged and insisted on the same.

There were piles of scrapped silicon wafers, but in a few days, Xiao Qiang would clean up these things and then send them new ones. And each piece was of extremely high quality.

Some of the scientific researchers in the project team are continuing to analyze operational data, modify the steps, and improve the success rate. Other researchers are also making other devices connected to the processor.

All computer-related components can be integrated into a chip to minimize performance degradation caused by circuit connections.

Every chip must be integrated into the photoelectric converter where data is exchanged with the processor. The difficulty is not much easier than manufacturing the first photoelectric hybrid chip.

At this time, Xiao Qiang provided more than 200 special laboratory robots to the automation department, and there were more than a dozen special laboratory rooms that were successfully renovated.

He made the decision and transferred two of them to the photon computer project team for the purpose of the photon computer project team. All the other researchers were shouting that the boss was biased.

Tens of thousands of failures have caused all the project team members to fall into deep despair. Later, even despair is gone, leaving only numbness...

In a blink of an eye, another year is coming.

On this day, Xiao Qiang routinely came to the photonic computer project team to observe the development progress of the chipset.

Over the past year, the success rate of chipsets has increased at the stool speed. From the beginning, only twenty-four cores were successfully produced, and one core and one core slowly increased. On average, every May, the success rate can be increased by about one core.

If the operation was not carried out by special laboratory robots, ordinary scientific and technological personnel would have been nervously broken long ago.

By last month, the entire chipset had already produced twenty-six chips. And the success was still far away and there was no end to see.

Before he could step into the project team's office, he heard bursts of crying from inside.

He was shocked. He pushed open the door and saw that many researchers inside were covered with tears and still cried bitterly. Xiao Jin sat in his seat with no expression on his face. He was like a wax man. Several chief engineers sat together, holding hands. His lips trembled and he cried, but he couldn't say a word.

"What happened?" Xiao Qiang's heart tightened and asked loudly at the door.

Is the experiment completely failed, or... or is the experimental data lost?

Those who were crying looked up and saw his appearance, and suddenly burst into tears. The chief engineers also burst into tears and burst into tears.

Xiao Qiang rushed to Xiao Jin with a quick step, shook his shoulder hard, and asked, "What happened? What's wrong with you?"

Xiao Jin was shaken by him, and there was a little anger in his misty eyes.

When he saw Xiao Qiang, he suddenly burst into tears, grinned and cried like a child: "Mr. Xiao... Mr. Xiao, chip... the laboratory said..., processor chip..."

Xiao Qiang's heart was cold and he sat dejectedly on the chair opposite him, and had not spoken for a long time.

All his efforts are placed on photon computers. With the help of the light and brain, Universal Company can fare away all its opponents, one step ahead of them, two steps, and even widen the gap with their opponents for ten or twenty years, so that they will never catch up.

Now it seems that all the efforts in the front have been wasted.

Everything has to start from scratch again.

He moved his neck, felt the giggling sound of his bones, and his body was extremely stiff.

In the room, some people were crying, and some were crying silently. Xiao Jin, who was opposite him, seemed to be in a state of madness, crying and laughing, not knowing what he was doing.

Xiao Qiang tried hard to stand up, patted Xiao Jin's face hard, and said in a hoarse voice: "Stay calm! Calm down! It's nothing if you fail, let's come again! I will always support you, ask for money, I will get money for you, we are drawing people from other departments, and asking for skills. I will lose and steal and rob them, and I will get them back for you! Now, cheer me up! I won't work today, and all the people in the land will go back to rest. Thank you for your hard work in the past few years, I'll take your vacation!"

Project team general technology

Ou Hongbo suddenly stood up and shouted: "Why should we finally make a forty-core processor on vacation? We should speed up the progress immediately and assemble and test immediately. You can't take a vacation, absolutely not take a vacation!"

After he finished speaking, he paused and ran out.

Xiao Qiang felt a darkness in front of his eyes. Suddenly, all the blood seemed to be on the top of his head, and a sentence echoed in his ears: "...Make a processor with forty cores..."

Success?

He fell back into the chair, just like Xiao Jin did just now, and he was silent and dumbfounded.
To be continued...