Mainstream advanced process technology review
Latest update time：2019-09-30
Reads：
At present, the mainstream advanced semiconductor process technology that has been mass-produced has reached 7nm, and next year, 5nm will also be mass-produced. From the perspective of process technology development, 28nm is generally used as a watershed to distinguish between advanced processes and traditional processes. Below, we will sort out the development of mainstream advanced process technology in the industry.
28nm
Since improving cost-effectiveness has always been regarded as the core meaning of Moore's Law, the rising cost of processes below 20nm was once considered a sign that Moore's Law was beginning to fail. As the most cost-effective process technology, 28nm has a long life cycle.
With design costs rising, only a few customers can afford to switch to advanced nodes. According to Gartner, the average IC design cost for 16nm/14nm chips is about $80 million, while 28nm bulk silicon process devices are about $30 million, and designing 7nm chips requires $271 million. IBS data shows that the design cost of 28nm bulk silicon devices is roughly around $51.3 million, while 7nm chips require $298 million. For most customers, switching to 16nm/14nm FinFET processes is too expensive.
In terms of unit chip cost, 28nm has obvious advantages and will maintain a longer life cycle. On the one hand, compared with 40nm and earlier processes, the 28nm process has obvious advantages in frequency regulation, power consumption control, heat dissipation management and size compression. On the other hand, since 16nm/14nm and more advanced processes use FinFET technology, it is more difficult to maintain high parameter yield and low defect density, and the cost of each logic gate is higher than that of the 28nm process.
28nm is between 32nm and 22nm. The industry introduced the high-k insulating layer/metal gate (HKMG) process at the earlier 45nm stage, and introduced the second-generation high-k insulating layer/metal gate process at 32nm, which laid the foundation for the gradual maturity of 28nm. In terms of the subsequent advanced processes, FinFET (fin field effect transistor) was adopted from 22nm. It can be seen that 28nm is at the key point of process transition, which is also an important reason for its high cost performance.
At present, the 28nm process in the industry is mainly competed among TSMC, GF (GlobalFoundries), UMC, Samsung and SMIC. In addition, Huali Microelectronics, a subsidiary of Huahong, which announced the mass production of MediaTek's 28nm chips at the end of 2018, has also begun to join the competition.
Although the high-end market will be occupied by 7nm, 10nm and 14nm/16nm processes, 40nm, 28nm and other processes will not withdraw. For example, 28nm~16nm processes are still an important part of TSMC's revenue, especially the foundries built in mainland China, which are mainly based on 16nm. SMIC is continuously improving the yield of 28nm.
14/16nm
The 14nm process is mainly used for the manufacture of mid-to-high-end AP/SoC, GPU, mining machine ASIC, FPGA, automotive semiconductors, etc. For various manufacturers, this process is also the main source of income, especially Intel, 14nm is its current main process technology. Considering the size of the company, the income it brings can be imagined. For mainland China's local wafer foundries, especially SMIC and Huahong, they are developing 14nm process technology, and the time for mass production is not far away.
At present, there are seven major manufacturers that have or will soon have 14nm process production capacity, namely: Intel, TSMC, Samsung, GlobalFoundries, UMC, SMIC and Huahong.
Both are 14nm processes. Since Intel strictly pursues Moore's Law, the level and rigor of its process are the highest. Judging from the technology that has been released so far, Intel's continuously updated 14nm process is roughly the same as TSMC's 10nm.
In May this year, Intel said it would increase 14nm process production capacity in the third quarter to solve the shortage problem in the CPU market.
However, Intel's own 14nm production capacity is already fully loaded, so the company invested $1.5 billion to expand 14nm production capacity, and is expected to increase output in the third quarter of this year. Its 14nm process chips are mainly produced in the D1X wafer fab in Arizona and Oregon, and the overseas 14nm wafer fab is Fab 24 in Ireland, which is currently upgrading the 14nm process.
Samsung announced the official mass production of 14nm FinFET process in 2015, and has manufactured high-end mobile phone processors for Apple and Qualcomm. At present, its 14nm production capacity market share is second only to Intel and TSMC.
TSMC started mass production of 16nm FinFET process in the second half of 2015. Compared with Samsung and Intel, although their node names are different, Samsung and Intel are 14nm, and TSMC is 16nm, they are at the same generation in terms of actual process technology level.
In August 2018, GlobalFoundries announced that it would abandon the research and development of the 7nm LP process and invest more resources in 12nm and 14nm processes.
GlobalFoundries has developed two process roadmaps: one is FinFET. In this regard, the company has 14LPP and the new 12LPP (a transitional version from 14LPP to 7LP); the other is FD-SOI. GlobalFoundries is currently producing 22FDX and will release 12FDX when customers need it.
As for UMC, its 14nm process accounts for only about 3% and is not its main production line. This is directly related to the company's development strategy. UMC focuses on the development of special processes. Whether it is an 8-inch plant or a 12-inch plant, the company will focus on the development of various new special processes.
As for SMIC, its 14nm FinFET has entered the customer trial stage. In the second quarter of 2019, new equipment was put into use in the Shanghai factory, and it is planned to enter the mass production stage in the second half of the year. In the future, its first 14nm process customer is likely to be a mobile phone chip manufacturer. It is reported that in 2019, SMIC's capital expenditure increased from US$1.8 billion in 2018 to US$2.2 billion.
As for Huali Microelectronics, at the SEMICON China 2019 Advanced Manufacturing Forum at the beginning of the year, the company's vice president of R&D Shao Hua said in a speech that Huali Microelectronics will mass-produce the 28nm HKC+ process by the end of this year and the 14nm FinFET process by the end of 2020.
12nm
From the current wafer foundry market, there are very few manufacturers with 12nm process technology capabilities, mainly TSMC, GlobalFoundries, Samsung and UMC. UMC announced in 2018 that it would stop the research and development of 12nm and more advanced process technologies. Therefore, at present, the main players in the global wafer foundry market for 12nm are TSMC, GlobalFoundries and Samsung.
TSMC's 16nm process has gone through three generations: 16nm FinFET, 16FF+, and 16FFC. After that, it entered the fourth generation of 16nm process technology. At this time, TSMC changed its strategy and launched a revised process, namely 12nm technology, to attract more customer orders and thus improve the capacity utilization rate of 12-inch wafer fabs. Therefore, TSMC's 12nm process is its fourth generation of 16nm technology.
In 2018, GlobalFoundries announced its withdrawal from the research and development of 10nm and more advanced processes, so the company's most advanced process is 12nm. The company walks on two legs, namely FinFET and FD-SOI, which is also fully reflected in the 12nm process. In terms of FinFET, the company has 12LP technology, and in terms of FD-SOI, there is 12FDX. 12LP is mainly aimed at applications such as artificial intelligence, virtual reality, smartphones, and network infrastructure. It utilizes the expertise of GlobalFoundries' Fab 8 in Saratoga County, New York. The factory has been mass-producing GlobalFoundries' 14nm FinFET products since early 2016.
Since many connected devices require both high integration and more flexible performance and power consumption, which is difficult to achieve with FinFET, 12FDX provides an alternative path that can achieve lower power consumption, lower cost and better RF integration than FinFET products.
Samsung's foundry roadmap originally did not include a 12nm process, only 11nm LPP. However, Samsung's 11 LPP and GlobalFoundries' 12nm LP are actually from the same school, both are improvements on Samsung's 14nm, with little change in transistor density and increased performance. Therefore, GlobalFoundries' 12nm LP and Samsung's 12nm process have a lot in common, which may be one of the reasons why AMD chose Samsung to manufacture its 12nm products.
As for SMIC, not only has the 14nm FinFET process entered the customer risk mass production stage, but also in the first quarter of 2019, its 12nm process technology development entered the customer introduction stage, and the second-generation FinFET N+1 research and development has made a breakthrough, with progress exceeding expectations. At the same time, the Shanghai SMIC South FinFET factory has been successfully built and entered the capacity deployment stage. This means that it will not be long before a new 12nm process player will join the battle.
10nm
At the 10nm node, the only industry players left are TSMC, Samsung and Intel.
In general, TSMC is still in the lead, and its typical product is the A11 processor manufactured for Apple in 2017. Samsung is also keeping up with the pace. At the 10nm point, the progress of the two parties is not much different, but the overall level is still slightly better.
This year, Intel's old rival AMD staged a comeback. With the 7nm Ryzen 3000 series processors manufactured by TSMC, AMD surpassed Intel in CPU processor manufacturing technology for the first time.
At present, Intel's mainstream process is 14nm. However, news came not long ago that after years of research and development, the company has finally solved the technical difficulties of the 10nm process and has begun mass production.
However, Intel's rigorous pursuit of process nodes is very commendable. From the perspective of specific performance indicators, especially PPA and transistor density, Intel's 10nm has advantages over TSMC's 10nm.
7nm
At present, there are only TSMC and Samsung in the 7nm process, and Samsung's mass production time lags significantly behind TSMC, which forces Samsung to skip 7nm and go directly to 7nm EUV. As a result, TSMC has almost snatched away orders from major 7nm process customers such as Apple, Huawei, AMD, and Nvidia. With this first-mover advantage, TSMC's 7nm production capacity is already a bit overwhelmed. In terms of 7nm EUV mass production, TSMC is also one step ahead. Huawei's Kirin 990, which is manufactured by TSMC, has been commercialized, and Qualcomm's new generation of processors manufactured by Samsung's 7nm EUV are also in production and are expected to be available soon.
As for Intel, after 10nm, the company said it will launch the 7nm process in 2021. It is reported that its 7nm process is on the right track, and its power consumption and performance look very good. According to previous news, the 7nm process will be launched on the data center GPU in 2021.
Conclusion
The above is an explanation of the development of mainstream advanced process technologies that have been mass-produced in the industry, as well as the progress of related manufacturers. The more advanced 5nm, 3nm, 2nm, etc. have not yet entered the mass production stage, so I will not elaborate on them. There are few players in these process nodes. Currently, there are only two companies, TSMC and Samsung. TSMC said it will mass-produce 5nm next year, while Samsung seems to skip 5nm and go directly to 3nm.
*Disclaimer: This article is originally written by the author. The content of the article is the author's personal opinion. Semiconductor Industry Observer reprints it only to convey a different point of view. It does not mean that Semiconductor Industry Observer agrees or supports this point of view. If you have any objections, please contact Semiconductor Industry Observer.
Today is the 2083rd issue of content shared by "Semiconductor Industry Observer" for you, welcome to follow.
Recommended Reading
★ In-depth analysis of the reasons behind semiconductor mergers and acquisitions
★ Chip-level disassembly of iPhone 11 Pro Max, BOM list exposed
★
Ranking of the top 20 global industrial chip manufacturers: Two European giants are the highlights
Semiconductor Industry Observation
" The first vertical media in semiconductor industry "
Real-time professional original depth
Scan the QR code , reply to the keywords below, and read more
AI｜ RF ｜ Huawei ｜ Chinese Chip ｜Wafer｜ Compound ｜ Storage ｜Alibaba
Reply
Submit your article
and read "How to become a member of "Semiconductor Industry Observer""
Reply Search and you can easily find other articles that interest you!
京公网安备 11010802033920号