Chinese semiconductor industry

Status
Not open for further replies.

gelgoog

Lieutenant General
Registered Member
If China wants to break the tool and materials blockade, it is pretty simple, the weakest link is Japan. Japan can do pretty much everything by themselves and their major trade partner by far is China. The US tools and materials manufacturers should be kept out of the Chinese market as much as possible. China and HK are like 25.66% of Japanese exports and 26.1% of Japanese imports. In comparison the US is 18% and 10.9% respectively.
 
Last edited:

olalavn

Senior Member
Registered Member
If China wants to break the tool and materials blockade, it is pretty simple, the weakest link is Japan. Japan can do pretty much everything by themselves and their major trade partner by far is China. The US tools and materials manufacturers should be kept out of the Chinese market as much as possible. China and HK are like 25.66% of Japanese exports and 26.1% of Japanese imports. In comparison the US is 18% and 10.9% respectively.
China doesn't need to do like U.S.... they just leisurely do their job... Japan isn't stupid enough to ignore the China market..
 

tokenanalyst

Brigadier
Registered Member

[Winning the bid] Jiaxin Caneng, a subsidiary of Wanye Enterprise, won the bid for 3 sets of thin film deposition equipment for the Shanghai Jita Project​


According to news from Jiwei.com, on February 28, Shanghai Jita Semiconductor Co., Ltd. (hereinafter referred to as "Shanghai Jita") special process production line construction project announced the results of three winning bids.

It is reported that the successful bidders of the three projects are Jiaxin Caneng Semiconductor Equipment Technology (Zhejiang) Co., Ltd., a subsidiary of Wanye Enterprise, and the subject matter involved is 1 boron-phosphorus-doped silicon dioxide film chemical deposition equipment and 2 titanium/titanium nitride deposition equipment.

Previously, Jiashan official news showed that the Jiaxin semiconductor project under Wanye Enterprise was undergoing secondary structural construction. It is expected to complete the civil engineering construction by the end of March and transfer to the stage of interior decoration and equipment installation. It will be completed and put into production by the end of the year.

Please, Log in or Register to view URLs content!
 

tokenanalyst

Brigadier
Registered Member

Microelectronics has made important progress in in-memory computing of graph networks​

Manuscript source: Shang Dashan and Zhang Kangwei from key laboratories Release time: 2023-02-24
As an important engine of AI ,   deep learning technology has received widespread attention and rapid development in recent years. Graph Neural Network ( Graph Neural Network ) is a relatively new deep learning technology that can be used to process more complex unstructured data, and is widely used in application scenarios such as social networks, e-shopping, drug prediction, and human-computer interaction. With the rapid expansion of the amount of data, the efficiency of running graph neural networks in traditional CMOS digital hardware systems needs to be improved urgently. The training process of graph neural networks is becoming more and more complicated, which makes training energy consumption remain high. Although the in-memory computing technology based on Resistive Memristor ( RRAM ) can significantly alleviate the von Neumann bottleneck in traditional digital hardware systems and further improve computing efficiency, it is still subject to high power consumption for erasing and writing, delay and certain limitations. Limitations of device non-ideal characteristics such as programming resistance randomness.

  In response to the above problems, Researcher Shang Dashan from Academician Liu Ming's team of the Key Laboratory of Microelectronic Devices and Integration Technology of the Institute of Microelectronics and Dr. Wang Zhongrui from the Department of Electronic Engineering of the University of Hong Kong have developed a graph structure using Reservoir Computing technology . A technique for data classification - Echo State Graph Network ( ESGNN ). Reservoir calculation is a simplified form of recurrent neural network, which can transform the time-series input signal into a high-dimensional space through the nonlinear activation function of neurons, and then effectively read it out through a simple linear regression method. In the pool calculation, the weight of the recurrent connection layer is always fixed, and only the weight of the output layer needs to be trained, which can minimize the training complexity and training time. In terms of hardware, the team used the intrinsic randomness of RRAM to construct a large-scale random resistor array (Figure 1a-b ), which is used as the initialization weight of the storage pool network, which has the advantages of low cost and scalability. In terms of software, ESGNN cleverly uses the physical random projection brought by the random resistor array, and completes the graph embedding process by means of in-memory calculation, which greatly reduces the training cost of the graph neural network. The team also realized the graph classification of MUTAG and COLLAB datasets on the FPGA -based board-level test platform through software - hardware collaborative optimization technology , and carried out node classification simulation of larger-scale CORA datasets. Compared with the traditional digital hardware system, the energy efficiency has been improved by 2.16 , 35.42 and 40.37 respectivelytimes. This work demonstrates the great potential of RRAM arrays in building edge graph learning systems, and also provides a reference for developing more efficient intelligent hardware systems using the rich physical and chemical properties of nature.

1677586096886.png
Please, Log in or Register to view URLs content!
 

tphuang

Lieutenant General
Staff member
Super Moderator
VIP Professional
Registered Member
Please, Log in or Register to view URLs content!
This should surprise no one, but cutting all Qualcomm sales to Huawei would have an immediate affect on Huawei's mobile & tablet division as well as auto division.
Now, they better stock up those CPUs while they still can. It should've been obvious when this was first announced that this is coming.
 

gadgetcool5

Senior Member
Registered Member
stopped reading at the part where he said it will take at least 10 years to build a domestic 28nm line....

then I double-checked, oh he's a finance guy.
10 years for 28nm is not that far off from other estimates. For example:

Wu Zihao, a former top Taiwanese chip engineer who has worked in China for two decades,
Please, Log in or Register to view URLs content!
that even in the best-case scenario China might be able to produce 28nm chips with its own lithography tools within six years.

He says this forecast is based on the most optimistic scenario, in which SMEE can still source foreign parts, which is increasingly difficult in the face of US sanctions and pressure.

His advice to Chinese scientists is that they tackle problems of making 90nm or 130nm chips before focusing on more advanced 28nm ones.

Please, Log in or Register to view URLs content!
 

tokenanalyst

Brigadier
Registered Member
10 years for 28nm is not that far off from other estimates. For example:

Wu Zihao, a former top Taiwanese chip engineer who has worked in China for two decades,
Please, Log in or Register to view URLs content!
that even in the best-case scenario China might be able to produce 28nm chips with its own lithography tools within six years.

He says this forecast is based on the most optimistic scenario, in which SMEE can still source foreign parts, which is increasingly difficult in the face of US sanctions and pressure.

His advice to Chinese scientists is that they tackle problems of making 90nm or 130nm chips before focusing on more advanced 28nm ones.

Please, Log in or Register to view URLs content!
I still remember when "experts" said that the first flight of the J-20 was going to be in 2030.

How many of those 10 years China already had walked? How many of todays China below 28nm tools are just half of their validation process? How many of todays China semiconductor materials are just half of their validation process? How many Chinese tools ended their development process? How many parts for tools are being developed? day after day we see progress way more faster in the last 5 years than in the last 40 years of development in the Chinese semiconductor industry. Many years following this thing.​
 

measuredingabens

Junior Member
Registered Member
10 years for 28nm is not that far off from other estimates. For example:

Wu Zihao, a former top Taiwanese chip engineer who has worked in China for two decades,
Please, Log in or Register to view URLs content!
that even in the best-case scenario China might be able to produce 28nm chips with its own lithography tools within six years.

He says this forecast is based on the most optimistic scenario, in which SMEE can still source foreign parts, which is increasingly difficult in the face of US sanctions and pressure.

His advice to Chinese scientists is that they tackle problems of making 90nm or 130nm chips before focusing on more advanced 28nm ones.

Please, Log in or Register to view URLs content!
Wu has already been discussed in this thread before when his blog was posted here. A lot of his statements directly contradict the ones made by havok, an active SMEE employee and insider and so the former's remarks are taken with a teaspoon of salt.
 

tokenanalyst

Brigadier
Registered Member

Xinyun Intelligent accelerates the layout of semiconductor automation equipment field.​


In recent years, with the unprecedented rapid development of my country's semiconductor industry, the demand for domestic chip production capacity has increased rapidly, and domestic fabs have therefore increased their production capacity layout. As an indispensable part of the semiconductor production process, semiconductor automation equipment shoulders the important mission of improving the utilization rate and product yield of the fab. However, looking at the domestic market, this field has long been monopolized by foreign companies. As an independent and controllable semiconductor automation equipment manufacturer with leading core technology, Wuxi Xinyun Intelligent Technology Co., Ltd. (hereinafter referred to as "Xinyun Intelligent" or the "Company") conducts product layout around the Automatic Material Handling System (AMHS) and is committed to Continue to empower the improvement of semiconductor production efficiency.

As one of the few teams in China who have participated in the R&D, installation, operation and maintenance of the first domestic AMHS system, the Xinyun intelligent team has deep technical accumulation. Relying on the in-depth understanding of semiconductor equipment and semiconductor factories, Xinyun Intelligent, which was established in April 2022, chose the wafer transfer module EFEM as the entry point, and opened the advanced road of AMHS system research and development. Based on a solid technical foundation and years of experience in solutions in the industry, the company successfully designed, developed and manufactured EFEM equipment with independent intellectual property rights and delivered it to customers in just 4 months, which has won high recognition from customers for the team and products. Based on this, the company has also laid a solid foundation for gradually breaking the foreign monopoly in the field of AMHS, accelerating the replacement of localized equipment, and realizing the independent and controllable development of China's chip industry.

In addition to the precise positioning track, the talent advantages of the technical team composed of senior practitioners from the semiconductor industry, combined with the advantages of independent intellectual property rights, also constitute the "moat" for the rapid development of Xinyun Intelligent. Today, the company's main products cover semiconductor factory AMHS automatic material handling system, semiconductor factory wafer handling automation equipment, and localized replacement of core components of automation equipment, which have been verified and recognized by well-known companies such as HTKJ, ZKZX, and CDYX. In the long-term strategic planning in the future, the company will continue to increase investment in research and development, continuously improve the core competitiveness of products in technology, and is committed to providing customers with complete solutions, while achieving world-class benchmarking of key technologies for product performance.

1677595517064.png
Please, Log in or Register to view URLs content!
 

tokenanalyst

Brigadier
Registered Member

Chinese chip companies cement IP defense by aggressive patent filings​


As semiconductors gain significance strategically and commercially, global patent filing statistics reveal a changing trend in which major economies are vying for dominant status in the relevant IP, with industrial policies from China, the US, and the EU expected to lead to more fierce competition.

According to European IP law firm Mathys & Squire, a record of 69,194 patents for semiconductors was filed between October 2021 and September 2022, up 10% from a year earlier and 59% from five years ago, highlighting intense competition with the sector.

TSMC was the largest filer, accounting for 7% of all patents worldwide in 2021-2022.

37,865 (55%) of the semiconductor patents were filed in China for the same period, compared with 18,223 (26%) in the US, as China intends to reduce dependence on western technologies. Meanwhile, Applied Materials and SanDisk, the top two patent filers in the US, filed 209 and 50 patents in the country.

Edd Cavanna, the managing associate at Mathys & Squire, said that governments are increasingly concerned about the fragility of global supply chains and are taking steps to promote semiconductor research and production domestically, with global powers such as the US, China, and the EU competing to be leaders in the sector.

Please, Log in or Register to view URLs content!
 
Status
Not open for further replies.
Top