PS: The third update, ask for a monthly pass.

……

The next day, when the final end was still preparing to receive the Qualcomm team, HiSilicon Semiconductor came to Hangzhou first, which was beyond everyone's expectations.

The wolf has a keen sense of smell, where there is meat, it can smell it from afar, and once it encounters it, it will get it desperately; Wolves, indomitable, brave and hard; Wolves are a group of group creatures, and every time they appear, they are a large group, and they dare to speak in the face of behemoths.

This is the outside world's evaluation of Huawei, and it is also summarized as a 'wolf culture', although Huawei does not admit it.

Twenty years ago, a small communication agent, now the world's top 500, with an annual turnover of hundreds of billions of yuan, the world's largest communication equipment manufacturer, mobile phone business, semiconductor business and various fields have made great achievements.

HiSilicon Semiconductor is a fabless semiconductor company under Huawei, whose main business is the design and sales of chips for mobile phone processors, set-top boxes and monitoring equipment, and the results are good.

"At this time, HiSilicon visited us in front of Qualcomm, apparently for the 16nm process. The first 16nm chip manufactured by TSMC was the HiSilicon Kirin 950, which was at the end of last year. However, at that time, the Kirin 950 was only a less mature chip, and now that more than half a year has passed, all companies have entered the era of fierce competition in the 14/16nm process, and HiSilicon must also launch its own processors, otherwise it will only lag behind. ”

"However, the production capacity of TSMC's 16nm process process is insufficient, and in the event of an accident, Huawei terminals will be out of stock, and it is undoubtedly more secure to find another foundry. And Huawei and Samsung are competitors, I heard that there have been conflicts over the screen panel, and the cost of Samsung's 14nm process foundry is slightly higher, so naturally they won't go to Samsung foundry, so they found us. ”

Liu Dacheng analyzed the reason for HiSilicon's sudden visit to Su Shi.

"Receive them well, if they ask for testing, it's naturally great, but don't rush to agree to the OEM first, and then make plans when Qualcomm comes." Su groaned.

After the project director of HiSilicon came to the end-end semiconductor, he first visited the production line and witnessed the 20nm process of Apple A8 packaging and molding.

There are many finished wafers, and you can directly use a ready-made piece for trial production. A 12-inch wafer can make more than 400 cores, of which it is enough to be damaged, Liu Dacheng really doesn't believe that this 12-inch wafer can't make a chip.

However, the preparation was rushed, and it was the first time to test the 16nm process, and in front of customers, everyone was a little nervous.

After dozens of process steps, the wafer was already visibly damaged. However, this is also normal, even if TSMC's 16nm process has already been mass-produced and tested, the yield rate in the first few months was only 50, and Samsung's 14nm process is even worse, not even 50. Even until now, the yield rate of both companies is only hovering at 60, the more precise the process, the higher the difficulty, not to mention that the 14/16nm process is still in its infancy, and the yield rate is not high.

Not to mention that this is the first time to try the 16nm process at the end of the day, and it is not surprising that there is damage. By the halfway point, there were already dozens of problems with the visual inspection.

"It's still acceptable!" Liu Dacheng's fists were full of sweat.

Photocopying, etching, repeating layering, cutting...... Chip manufacturing is a complex systems engineering.

"30 damaged!"

After a 12-inch wafer is cut, it can be made into more than 470 cores, and one core is a processor. After 30 years of damage, there are more than 300 pieces left.

"Ok! Not bad! Encapsulation! ”

At this time, it was already early in the morning, and although the shift had changed, Liu Dacheng and the supervisors of several factories still stayed up late to stick to their posts. Even HiSilicon's team didn't sleep, they worked at Huawei, and working overtime was commonplace.

"The packaging is complete, and there are 308 healthy ones left!"

"Test!"

After a few tests, dozens of failed ones were eliminated, and the final passing finished product was a total of 272. Nearly half of the loss, but it is already a very good result, after all, the first round of trial production.

"The yield rate is more than 55, and you are really the first time to trial produce a chip with a 16nm process?" Several directors of HiSilicon widened their bloodshot eyes, but the surprise was more incredible. They have been to TSMC, and it is the same chip, but at the beginning of trial production, the yield rate was very bad.

"Uh......" Liu Dacheng was stunned.

“…… In fact, we have done simple tests before, but it is not a complete processor. Sue, who had just arrived, interjected. The yield rate is low, and the price is high, which cannot be known to customers.

"I see." They don't doubt it either, it's normal that way. Then he said: "Mr. Su, we need to send some back to the real room for systematic testing, if there is no problem, we will do our best to send someone over to negotiate business with the final end." ”

"This is a must, Manager Liu, hurry up and let people pack the chips." Sue commanded the most.

"Thank you, Mr. Su."

After HiSilicon's people left with dozens of samples, they took the high-speed train back to their Shanghai headquarters overnight, and even Su Shi asked them to have a supper. Huawei people are positive, that is well-known in the industry, although there are many overtime, but they are treated very well, and even have the company's share dividends, so they are naturally hard-working.

"Everyone has worked hard, hurry up and go back to rest, there will be a higher test tomorrow." Su said to everyone the most.

The next day, the Qualcomm team also came to Hangzhou, and they brought the design of the Snapdragon 820. But there are two schemes, one is 14 nanometers and the other is 16 nanometers.

14nm is Samsung's foundry solution, 16nm is TSMC's foundry solution, and Apple's A9 routine is exactly the same, one processor stores zài two sets of processes.

Although 2 nanometers is insignificant, there is not much difference in performance and energy consumption. But it stands to reason that the 14nm process will be better, but whether it is Samsung or TSMC, none of them can meet the needs of Apple or Qualcomm, and the interests of various forces make it almost impossible to eat alone.

Just like the Qualcomm 820, there is a requirement for Samsung OEMs, that is, Samsung's next-generation flagship phones need to use their Snapdragon 820 processors. But TSMC Qualcomm is not easy to offend, the best solution is to share some shares by one person, even if Qualcomm is not happy with its competitor Samsung, and it is not happy with TSMC that kneels and licks Apple.

It is the interests of these large multinational companies that are the most important.

"Let's try to produce a 14nm solution!" Su was the most decided. The smaller the process, the lower the cost, the better the chip capacity, the lower the power consumption, and the higher the foundry capacity. Moreover, Su is most confident in the capabilities of the end-end factory, believing in the level of these equipment, there is really no reason to choose the latter.

It may not be difficult to go from 22 nanometers to 20 nanometers, but from 16 to 14 nanometers, it is very difficult to move forward every nanometer, otherwise TSMC is not willing to lag behind Samsung and choose the 16 nanometer process.

But for these devices in the end-end factory, this 2nm is not too difficult. After a day and a night, not only did the Snapdragon 820 be successfully built, but the yield rate was also improved compared with the previous HiSilicon Kirin 950, mainly due to everyone's skillful operation.

"God!"

"Incredible!"

"Is the end end secretly nurtured by Intel?"

Several American friends shouted while holding the finished chip product, such a process, such a ability, let them have the illusion of crossing into the Intel production workshop.

Intel is the strongest semiconductor in the world, not only designed, but also the corresponding instruments are produced and sold by them, with the world's most cutting-edge technology. Although the 14nm processor has just been mass-produced, Intel's actual room already has an 8nm finished product. It is not impossible to say that Intel's consortium secretly supports some companies to make money, and the purpose of not disclosing it can avoid the US government investigation.

Even if it is not Intel, the final end is definitely supported by Western companies, because China does not have such capabilities at all, and China's equipment is even stuck in the stage of 45nm process. Texas Instruments? Or IBM? Aren't these companies afraid of U.S. government investigations? These thoughts kept flashing through the hearts of several American friends.

The chip was made, but it was impossible to sign a contract right away. These chips can only be considered engineering test samples, and like HiSilicon, they need to be sent back to the company for rigorous testing. Even if there is no problem in the end, there are still prices, the number of foundries, and the time of foundry that need to be negotiated.

However, this does not affect the publicity of the two companies, even if the cooperation fails, the joint external publicity is mutually beneficial. In the end, Qualcomm can also make customers and shareholders more confident in the company.

As for Samsung and TSMC, they will be nervous about this, but it is in the interests of Qualcomm to let you monopolize and let you kneel and lick Apple. Although the production capacity of the end is far from threatening them, it is also good to scare and scare

R1152