1. Cell phone/laptop-take-all Arm CPU performance keeps pace with Core i5;
As mobile computing becomes more complex and various new forms of virtual experiences are applied, the application of artificial intelligence and machine learning is increasing. It is necessary to promote the efficiency and efficiency of mobile phone processors. To this end, Arm released a new generation of Cortex A76 mobile processing chips to provide the equivalent of Intel. (Intel) Core i5 high-performance processor, to meet the market demand for mobile phones and laptop applications.
Peter Greenhalgh, vice president and academician of Arm Technology, said that the demand for single-linear computing in smartphones is constantly increasing. The high-performance required for the smart phone requires a notebook-grade rating, prompting Arm to create a processor with a notebook-size capability. The processor is imported into the mobile application.
Ian Smythe, Arm's senior marketing director, said that smart phones have become a mobile device for people to use every day. Everyone has been accustomed to the rapid response of smart phones, but they do not realize that smart phones need a very high computing power. In addition, there are many businesses. Due to the needs of the workforce, the mobile phone uses a lot of Office 356 cloud service functions. The mobile phone also indirectly becomes another vehicle to replace the computer work.
According to Smythe analysis, when OEMs use Arm's high-end applications processors, they often depend on cost, processing performance, and considerations for different applications. The new processors will provide better memory capacity, improve processing performance, and be highly efficient. The product design; However, Cortex A76 does not replace the existing processor, but to provide a new choice. For example, some OEM manufacturers believe that the Cortex A75 and Cortex A73 can meet most of the application scenarios, will continue to use In addition, the Cortex A75 and Cortex A73 are still smaller than the Cortex A76 in terms of overall processor footprint, and are still superior to the Cortex A76 for cost-sensitive manufacturers.
On the other hand, in addition to applications on mobile phones, the Cortex A76 is an attempt to replace the components of the notebook processor. Arm observed that the trend of the notebook Always On, is likely to produce a new size class device, and this is the Cortex A76 The type of product that may be the first to be introduced, providing OEMs to import larger mobile devices to increase overall device productivity.
Overall, Greenhalgh revealed that there are currently two SoC vendors introducing the Cortex A76 into chip design, and Arm also expects that the processor will be seen on networked laptops in 2019.
2. The equipment/capacity investment is hot. The semiconductor industry is still a good year.
In 2017, the semiconductor industry paid a surprising number of transcripts. The overall revenue for the first time exceeded US$400 billion, breaking the view that the semiconductor industry is already mature. However, this boom has not yet ended, and the scale of semiconductor revenue will be Continue to refresh historical records, but the growth rate is slightly converged. It is estimated that by 2019, global semiconductor revenue will challenge the 500 billion US dollars mark.
The global semiconductor industry’s revenue in 2017 has achieved super-brilliant growth of up to 20% year-on-year, which not only broke the view that the semiconductor industry is already mature and has limited growth potential. At the same time, its revenue scale has also exceeded the US$400 billion mark. This figure has caused many research institutes and market participants to break their glasses. In the semiconductor industry, in the past 10 years, except that the financial tsunami has pushed down the base period, which led to a surge in growth rate to more than 30% in 2010, the semiconductor industry has seen an annual revenue growth rate of more than half. Below 10%, there was even a slight decline of 1%. Therefore, the performance of the semiconductor industry in 2017 is indeed remarkable.
The memory market returns to normal
Looking into 2018 and 2019, the global semiconductor industry will continue to grow, but the growth rate will slow to 8% and 6%, return to normal level. The revenue scale will continue to set a new historical record, and it is expected to challenge 500 billion US dollars in 2019 However, some research institutions believe that the revenue growth rate of the semiconductor industry in 2018 will still maintain a high level of 16%, but SEMI believes that taking into account the memory market prices fell back to normal levels, 8% should be a more reasonable expectation.
Memory is the locomotive for the growth of the semiconductor industry in 2017. Regardless of DRAM or NAND Flash, the quotation has continued to be high, and the revenue of each DRAM and NAND flash player has also written a record high. Samsung Electronics even squeezed it down. Intel (Intel) becomes the largest semiconductor company in global revenue.
However, as the timing has entered into 2018, NAND Flash prices have been declining, and DRAM prices are likely to be revised downwards in the second half of 2018 to early 2019 due to a large number of new production capacities. This also means that this wave of memory boom cycles. The highest point has passed.
SEMI believes that under the influence of memory price revision, the growth rate of global semiconductor industry revenue in 2018 and 2019 will be closer to normal, and it will not be easy to see growth of more than 10%. However, this does not mean that the memory industry is good. The days have come to an end. Looking at NAND Flash, the current price revision of the product mainly reflects the decline in manufacturing costs, but not the competition between the companies; the DRAM prices will loosen and will be the inevitable result of the new production capacity. The memory industry's investment in production capacity is quite rational, so the profit margin of related companies can still hold the basic disk.
In addition to memory, Optoelectronics, Sensors, Discrete and Analog components also performed well in 2017 and are expected to grow in the coming years in applications such as IoT, 5G, artificial intelligence and automotive electronics. There is good performance. Figure 1 shows SEMI's analysis of the semiconductor industry trends and future growth engine.
Figure 1 Future semiconductor revenue growth engine and major application market size forecast
Equipment investment creates a new record high material market performance
SEMI updated the contents of the global fab forecast report at the end of 2017 and pointed out that in 2017, the related expenditures for equipment investment in fabs will be revised up to US$57 billion, which will record a new historical high. As the demand for chips is strong, the memory pricing remains high. Factors such as fierce competition have continued to drive up investment in fabs, and many companies have invested in new fabs and related equipment in unprecedented ways. The amount of global fab equipment spending over the years (Figure 2).
Figure 2 Global Semiconductor Capacity/Equipment Capital Expenditure Trend
Although many companies such as Intel, Micro, Toshiba, Western Digital, and Globalfoundries all increased their fab investment in 2017 and 2018, the increase in fab equipment spending It is also the two companies in South Korea's Samsung and SK Hynix.
SEMI data shows that the overall increase in South Korea’s investment in 2017 was mainly due to the large growth in Samsung’s spending, which is expected to increase by 128%, from US$8 billion to US$18 billion. SK hynix’s fab equipment spending also increased approximately. 70%, up to 5.5 billion U.S. dollars, setting a record for the company's history. Samsung and SK hynix spent most of their spending in South Korea, but still some of them invested in the Chinese mainland and the United States, thus driving the spending in these two regions. The growth of SEMI forecasts that the investment amount of these two companies will continue to remain high in 2018.
In 2018, many fabs completed in 2017 in China are expected to enter the equipment-installation stage. However, most of the fab investment in mainland China is still from foreign manufacturers. In the year of 2018, fab equipment spending by mainland component manufacturers in mainland China The amount will grow by a big margin, reaching about US$4.5 billion, while foreign companies are expected to invest US$6.4 billion. Many new entrants, including Changjiang Storage, Fujian Jinhua, Huali, Hefei Changxin, etc., are planning to be located in China. Great investment in setting up factories.
As regards semiconductor materials, although new production capacity has not yet been developed, due to the fact that silicon wafers and other material companies have not actively expanded their production capacity in the past few years, the situation of the shortage of materials is very evident in the case of a significant recovery in semiconductor economy. There are a wide variety of materials, including silicon wafers, gases, and chemicals, among which silicon wafers are the largest. Therefore, the performance of the silicon wafer industry has the greatest impact on the overall performance of semiconductor materials.
The average unit price (ASP) of semi-conductor silicon wafers has been in a slow decline for a long time, and the driving force for revenue growth has been the increase in shipments. However, this has not been the case since 2017, due to the insufficiency of supply of silicon wafers. The successful escalation of quotations by suppliers of silicon wafers has also reversed the long-term trend of the decline of ASP. Coupled with the continued historical record of silicon wafer shipments, the size of the silicon wafer market in 2017 has grown by 17% compared to 2016. It also led to a 10% annual growth in the overall semiconductor material market (Figure 3).
Figure 3 Global semiconductor capacity/equipment capital expenditure trends
In order to protect the hard-won good prices, looking forward to the next few years, the silicon wafer manufacturers' capacity expansion will not be too fast in order to create as many bargaining chips as possible with customers. However, due to the strong bargaining power of semiconductor silicon wafer customers, Therefore, the future trend of the average price of silicon wafers needs to be observed.
SEMI estimates that the global semiconductor silicon wafer production capacity in 2018 and 2019 will only increase by 3.6% and 3.2%. The overall semiconductor material market size in 2018 is estimated to grow by 4%.
The semiconductor industry is good for the next two years
Judging from the data on semiconductor equipment and materials currently acquired by SEMI, 2018 and 2019 are still considered to be a good year for the semiconductor industry. Although the US-China trade war has recently warmed up, it coincides with ZTE’s sanctions imposed by the US embargo. It seems that the war on trade war seems to have delayed the possibility of burning into the technology industry, but this should only be an independent event. After all, traditional industries such as steel and agricultural products are the main battlefields of trade warfare.
On the other hand, after the ZTE was banned by the US technology, the Chinese government raised its concern for the development of the semiconductor industry to a higher level. In the future, China may provide more supportive policies for the semiconductor industry to try to resolve the lack of "core." However, Due to the recent occurrence of the ZTE event, the speech made by Chinese President Xi Jinping during his visit to Wuhan Xinxin, the implementation of the policy adjustment was not going to be so fast. Therefore, it is still difficult to assert that China’s investment in the semiconductor industry will increase. How much. But for the global semiconductor industry, China's every move in the future will inevitably be the focus of continuous attention.
(This article was dictated by Zeng Ruiyan, senior manager of SEMI Taiwan Industrial Research, Huang Jikuan, finishing) New Electronics
3. CAGR of MEMS MEMS components from 2018 to 2023 reached 17.5%;
The Micro-Electro-Mechanical (MEMS) market will grow at a CAGR of 17.5% between 2018 and 2023. The market size will reach USD 31 billion by 2023. The research institute Yole Développement (Yole) pointed out that RF components play a key role in the development of the MEMS industry. Role, excluding RF, the growth rate of the MEMS market during the same period fell to 9%. With the complexity of the transition to 5G and the higher bandwidth it brings, the demand for 4G/5G RF filters is increasing. RF MEMS (mainly BAW filter) has become the most rapidly developing field of MEMS applications.
In many existing MEMS components, inkjet heads will continue to grow, and the consumer market will account for more than 70% of the print head market demand. This market recovered in the first half of 2017, and this trend was confirmed in the second half of 2018. This recovery has been observed in both disposable and stationary printheads.
Many pressure sensor applications also help the market expand. In fact, it is interesting that although it is one of the oldest MEMS technologies, pressure sensors are still growing. In the automotive field, pressure sensors have the highest number of applications, Advantages of smart tires that are subject to toxic emissions and harsh environments, higher precision, and more tire status messages. For consumers, mobile phones and smart phones still account for 90% of the sales volume of pressure sensors, reducing costs. This is a priority, because the size is already very small. And new applications are emerging: smart homes, electronic cigarettes, drones and wearable devices, and more.
MEMS microphones have emerged as one of the highest CAGRs for MEMS technology in the past five years. In 2008, the scale reached US$105 million. The market size in 2012 reached US$402 million. In 2016, it officially exceeded the milestone of US$1 billion. Currently, MEMS buy winds. Each year, nearly 4.5 billion shipments are shipped. The main application is mobile phones, which account for 85% of shipments and 98% of the consumer market. Tablet PCs and personal computers/laptops account for the second and third largest shipments respectively. The amount of 5% and 3.2%. In 2016, the revenue of the top 30 MEMS manufacturers exceeded $9.238 billion. In 2017, it further increased to $98.81.
4. Automotive semiconductor manufacturing requirements vary greatly Fab random defect rate is critical
In the 1950s, the electronics used in automobile manufacturing were less than 1% of the total manufacturing cost. Today, the cost of electronic products can already be as much as 35% of the total cost and is expected to increase to 50% by 2030. The rapid growth of electronic products is mainly driven by the following four aspects: • System monitoring and control (electronic fuel injection, gas-electric hybrid, etc.) • Safety systems (anti-lock braking, airbags, etc.) • Advanced driver assistance systems (drift lanes) Warning, parking assistance, blind spot monitoring, adaptive cruise control, etc.) Convenient driving (satellite navigation, infotainment, etc.)
Semiconductor components are the core of the electronics assembly in automobiles. According to different manufacturers and models of automobiles, modern automobiles may require as many as 8,000 chips, and this number will only increase with the popularity of self-driving cars. Additional The electronics subsystem and the integrated circuits it uses will provide the driverless sensors, radar and artificial intelligence needed for the driverless car.
The annual production of cars and light trucks is more than 88 million, and thousands of chip products are installed in each car. The impact of the automotive industry on semiconductor manufacturing has begun to appear. A simple fact is the thousands of chips used in cars. Nothing can fail.
The reliability of automotive semiconductor parts is critical. Any chip that fails while the vehicle is in motion may result in expensive warranty repairs and product recalls, and may damage the automaker’s brand image. In extreme cases it may lead to personal injury. Injury is even life-threatening.
If an ordinary car has 5,000 chips and an automaker produces 25,000 cars per day, then even a one-millionth (ppm) chip failure rate will lead to over 125 cars a day because of the reliability of chip quality. problem.
Since semiconductors are the primary problem in automakers' fault-scheduling diagrams, first-class automotive system suppliers are now demanding that semiconductor quality can reach a one-billionth (ppb) level, and the current trend is that regardless of the number of chips, the more The more suppliers come to define the 'maximum allowed number of faults'.
The current method of detecting reliability failures relies excessively on testing and burn-in tests. As a result, the quality objectives cannot be achieved and are far apart. At the same time, auditing standards are becoming more and more challenging, and fabs have been found to be reliable at the source of chip manufacturing. Sexuality problems, because it is the lowest cost to find problems and take corrective measures at this time. To enter this growing market area, or simply to maintain market share, IC manufacturers must actively respond to this change in chip reliability requirements. .
Fortunately, for semiconductor manufacturers, the reliability of the chip is highly related to what they know: Random defects.
In fact, for well-designed processes and products, early chip reliability problems (extrinsic reliability) are dominated by random defects. Killer defects (defects that affect yield) are caused by components at time t = 0 (finally Test) Failures. Potential defects (imperfections that affect chip reliability) are defects that cause the component to fail at t>0 (after aging).
The relationship between killer defects (benefit) and potential defects (reliability) is found to affect reliability by observing the same defect type that affects yield. Both are mainly based on the size of the defects and their appearance on the component structure. Location to distinguish. Figure 1 shows examples of killers and potential defects that lead to open and short circuits.
Figure 1 The same defect type that affects the yield rate also affects reliability. It is mainly based on the size of the defects and their position on the pattern structure.
The relationship between yield and reliability defects is not limited to specific defect types; any type of defect that may cause loss of yield may also cause reliability problems. Failure analysis shows that most reliability defects are actually process-related. The defects are traceable to the fab. Because yield and reliability defects have the same root cause, improving yield (by reducing yield-related defects) will increase reliability.
The A curve in Figure 2 shows the typical yield curve. If we only consider the chip yield, then at some point further investment in this process may not be cost-effective, so the yield rate tends to increase over time. For smoothness, the dotted line B in Figure 2 shows the curve of the same factory that manufactures the same product. However, if they want to supply the automotive industry, they must also consider the cost of insufficient reliability. In this case, Further investment is needed to further reduce the defect density, which can both increase the yield and enhance the reliability required by automotive suppliers.
Figure 2 Yield curves for different types of fabs (yield vs. time). The A-curve is applicable to fabs in the non-automotive industry. The main concern is the fab's profitability. At some point, the yield has been High enough, it is not practical to continue to try to reduce the defect rate. The dotted line B also includes the yield curve of reliability. For integrated circuit products used in the automotive supply chain, additional investments must be made to ensure high reliability, which is related to benefits. closely related.
The transition from a generic chip supplier to an automotive supplier requires a paradigm shift in fab management. Successful semiconductor manufacturers in the automotive industry have already adopted the following strategy: The best way to reduce potential (reliability) defects is to reduce wafers The overall random defect level of the plant. This means that there is a world-class strategy to reduce defects, including: improving baseline yields, reducing the incidence of anomalies, detecting them quickly when they are anomalous and repairing them online, and using crystal screening to eliminate suspicious The grains.
(The writer is Dr. David W. Price and Jay Rathert is the senior director of KLA-Tencor. Douglas Sutherland is the chief scientist of KLA-Tencor. Over the past 15 years, they have worked directly with more than 50 semiconductor IC manufacturers. Optimize overall process control strategies for a variety of specific markets, including automotive reliability, traditional fab cost and risk optimization, and best listing methods for advanced design specifications.) New Electronics
