The hard way of the Otis chip reflects the industrial dilemma Moore's Law is invalid?

In May, Shanghai was sunny, and Andreas Gerstenmayer, global CEO of European printed circuit board (PCB) manufacturer AT&S, was a bit heavy.

Gesima flies to China in the May earnings season, and China is one of the most important markets for Otis. However, due to the failure of the new plant in Chongqing to achieve its target capacity and the huge price pressure on the semiconductor package carrier board, Otis lost a total of 22.9 million euros last year. Compared with the profit of 56 million euros in the previous fiscal year, the profit was sharp. Sliding down. It is reported that the total investment of Otis in the Chongqing plant is as high as 480 million euros, which is the largest single investment of the company so far.

The dilemma faced by Oster in the Chinese factory is also a true portrayal of the slowdown in Moore's Law. Ge Simai said: "As the market slows down and demand decreases, the price pressure on semiconductor package carrier will continue to grow, and the performance in the next fiscal year will continue to be affected." However, he is still optimistic, with the gradual advancement of technology Appeared, the company will eventually move towards profitability step by step.

Otis's difficult steps in the downstream chip packaging market reflect the plight of the entire industry chain. In fact, the past year has been a challenging year for the entire semiconductor industry.

Slow down

Intel co-founder Gordon Moore proposed "Moore's Law" in 1965, predicting that the number of transistors and resistors integrated on semiconductor chips will double every year. At its core, the performance of the chip will gradually increase and the cost will gradually decrease. However, in the 50 years of the development of Moore's Law, the challenge of physically producing smaller chips is gradually increasing.

Ge Simai told the First Financial Reporter that the upgrading of semiconductor industry technology is slowing down, especially in the chip processor industry. The 10 nanometer processor enters the market much later than Otis expects, so the corresponding The plan for the carrier to enter the market has also been postponed. "At present, the 14-nanometer processor on the market stays much longer than we think. According to the announcement of many microprocessor manufacturers, it is reasonable to say that it should be updated." Ge Simai told the first financial reporter “In general, there will be a new generation of CPUs coming out of the market every two years. But until now, we have seen that the pace of upgrading seems to have slowed down a lot, a year and a half later than originally planned.”

Indeed, as the "big boss" of the chip industry, Intel's 10 nanometer chip production has experienced several "jumping tickets."

Intel is constantly adjusting its process technology metrics to accommodate the slow pace of Moore's Law in terms of economics and chip size. Initially, Intel's product replacement cycle was a year and a half, and the cycle was extended to two years, while the recent 14-nanometer chip has been in the market for nearly three years.

Related statistics show that the company's process technology changed from 45 nanometers to 32 nanometers for about 27 months, from 32 nanometers to 22 nanometers for 28 months, and from 22 nanometers to the current 14 nanometers for 30 months. Since September 2014, Intel has made no progress in terms of process technology.

Intel told the First Financial reporter that the new 10 nanometer chip is expected to eventually come out in the second half of 2017. The industry expects that the 10 nanometer chip will stay in the market for about 3 years. Although more advanced than 10 nanometers, the 7-nanometer chip called Technology Leap is expected to be available in 2019, but Intel's 7-nanometer chip may have to wait until 2020.

Has Moore's Law expired?

In the eyes of chip companies such as Nvidia, Moore's Law began to fail 10 years ago, and the artificial intelligence revolution has arrived. The sharp rise in CPU transistors and energy has led to a small increase in application performance. Recently, its performance has only increased by 10% per year, compared with 50% in the past. Nvidia believes that the Dennard scaling effect has encountered a bottleneck in component physics.

Still, Intel, which has always used Moore's Law as the guiding light, is still convinced that it can still produce smaller, faster, and cheaper chips. In Intel's Technology and Manufacturing Day in the first half of this year, Intel Vice President Stacy Smith said: "Moore's Law is not dead, at least for us."

In fact, for Intel, their strategic direction is gradually changing. Mark Bohr, senior researcher at Intel's Process Architecture and Integration, said: "Intel doesn't want to play 'digital' games with Samsung or TSMC. Intel will use density metrics to define process nodes."

If this standard is used to calculate, Intel has increased transistor density at twice the speed in recent years. Bohr exemplified that when 22 nm evolved to 14 nm, the transistor density increased by 2.5 times, and when 14 nm evolved to 10 nm, the density increased by 2.7 times. "The most important thing is that the 10 nanometer chip has made great progress in computing speed and power consumption," Bohr said.

Intel has been punished by investors and analysts for failing to adhere to the practices of previous decades, that is, to compress the size of transistors every two years. In the past five years, Intel's share price has only increased by 28%, less than half of the S&P 500 index.

Gartner analyst Sheng Linghai told the First Financial Reporter: "Intel uses Moore's Law to promote technology development while achieving a monopoly on general-purpose processors. But now, its core production technology following Moore's Law is not suitable. The fragmented market, in addition to the volume and profit requirements of the investment sector is also very high."

Public information shows that some of the projects Intel has invested in have not achieved much benefit, and have abandoned applications such as application processors, tablet chips, mobile phone chips and digital TV.

Integration acceleration

Shenglinghai told the First Financial Reporter: "As the technological progress in production slows down, the design and software capabilities are even more important, which also accelerates the integration of the industry."

Sheng Linghai explained that the profit distribution is still good, the branches in each industry chain can live very well, but now it is continuously integrated, because the profit is reduced by reducing the synergy of competitors or integrated products. .

In the past two years, the integration of chip industry giants has been frequent. At the end of 2015, Intel spent $16.7 billion to acquire programmable chip maker Altera, the largest acquisition in Intel's history. In October last year, Qualcomm acquired Europe's semiconductor giant NXP for $47 billion. In 2015, NXP acquired another in-vehicle semiconductor giant, Freescales, for $11.8 billion. In July last year, ADI (Yardno Semiconductor) acquired Linear Technology for $14.8 billion.

This year, the acquisition of chip giants has revived. In March of this year, Intel acquired Israeli information technology company Mobileye for $15.3 billion. On May 26, Qualcomm teamed up with Datang Telecom, a subsidiary of Datang Telecom, and Jianguang Capital and Zhilu Capital to establish JLQTechnology, a joint venture company, to enter the low-end market of mobile phone chips, thereby countering the shares of Intel, which is owned by Intel. Spreadtrum and Rideco. As early as February of this year, Jianguang Capital has completed the acquisition of NXP's standard parts business for 2.75 billion US dollars.

At present, Japanese semiconductor manufacturer Toshiba is also eager to sell semiconductor business. As the world's second largest flash memory chip manufacturer, many giants are aiming at Toshiba Semiconductor. Among them are not only the Toshiba partner Western Digital, but also the US chip maker and Apple supplier Broadcom; there are also Korean chip maker SKHynix (Hellex) and Taiwan's Foxconn parent company Hon Hai.

According to Bloomberg News, Apple also intends to join Hon Hai to acquire Toshiba. Earlier, Hon Hai said it was willing to pay 3 trillion yen for the acquisition. However, Western Digital prevented Toshiba from selling chips to third parties on the grounds of “default” and initiated arbitration. The results will be seen in the second half of this year.

Chip layout dispute

After a large-scale integration, the number of companies in the semiconductor industry capable of manufacturing the most advanced chips has changed from a dozen or so of 10 years ago to a few of them. In addition to Intel, the current list is left with Samsung, TSMC and Grofund, which was split from AMD in 2009.

Samsung is fighting with Intel, they all want to be the world's most advanced chip maker. Samsung is the first to embrace 10 nanometer technology, but Intel is delayed due to manufacturing problems. Now Intel is still making chips on the 14nm process.

A recent report by McLean shows that Intel’s position as the chip industry leader for more than 20 years will be broken. Samsung's chip sales in the first quarter of 2017 exceeded Intel for the first time, boosted by the growing market demand for memory and flash memory. And the future industry trends driven by deep learning and machine learning will further fuel the demand for smart chips, and Intel's successful transformation becomes critical.

Samsung's first-generation 10nm chip has been used in the Galaxy S8 released earlier this year. Qualcomm's Opteron 835 chip is also manufactured at Samsung's existing 10nm factory.

Samsung is still improving the performance of 10 nanometer chips. The company is already developing a second-generation 10nm chip and announced that it will use 10nm LPP (low-powerplus) technology to produce chips in the fourth quarter of this year. According to reports, the latest 10 nanometer LPP chip uses an enhanced 3D architecture, which is 10% faster than existing chips and can save 15% energy.

Although Samsung did not disclose which customers will place orders to manufacture new chips, Qualcomm is likely to continue to be a major customer of Samsung's second-generation 10nm chips. Apple and Qualcomm's chip demand accounted for about half of Samsung's overall shipments.

螳螂 螳螂 蝉, the oriole is behind. Taiwan Semiconductor Manufacturing Corporation TSMC is also actively participating in the highest-end chip manufacturing field. Wei Zhejia, CEO of TSMC, revealed that the company is developing 7-nanometer chips and has started to manufacture 12 products and started mass production in 2018. The industry expects that the 10 nanometer chip produced by TSMC will become the exclusive supplier of the latest iPhone that will be released in the year.

Gartner senior research analyst David Christensen told the First Financial reporter: "Fast 4G migration and more powerful processors make the wafer size larger than the previous generation of application processors, requiring OEMs to provide more 28nm, 16/14nm and 10 Nano-wafers. The original process will continue to grow strongly in the areas of highly integrated display driver chips and fingerprint ID chips and active matrix organic light-emitting diode (AMOLED) display driver integrated circuits (ICs)."


Switch Cabinet Intelligent Operating Device

Switch Cabinet Intelligent Operating Device,Wireless Temperature Measurement,Temperature Measuring Device,Wireless Temperature Measuring

zhejiangjinyidianqiyouxiangongsi , https://www.jooeei.com