Chinese semiconductor industry

xypher said:

Please, Log in or Register to view URLs content!

Click to expand...

If this is true, it is happening way too fast.

For example, the 5G network build out worldwide. That will take several years. Now they have the chip in mass production at 3nm?

That 3nm chip is not going into those networks, that is for sure. That 3nm chip not going into IoT devices or automobiles.

Other than the cell phone, there will be no one using a 3nm by next year.

If TSCM really does this, then what this is, it is a preemptive strike to try to monopolize the market in the cutting edge chip.

If they can do that, then the next node is theirs too.

horse said:

If this is true, it is happening way too fast.
...
That 3nm chip is not going into those networks, that is for sure. That 3nm chip not going into IoT devices or automobiles.
Other than the cell phone, there will be no one using a 3nm by next year.

Click to expand...

It is just pilot production. If the pilot batch proves out ok, mass production might be a year or two away from that. Assuming everything works ok. The industry moves quite slowly. It takes months just to get the wafers out once you start production for example. Then they have to be tested for defects and the process might need to be improved. It is an iterative and slow process.

3nm would be useful not just for mobile but also server CPUs and large GPUs.

It is recommended to have a separate thread for Quantum. Posting here since it doesn't exist

China claimed Quantum Computing Superiority with two new Quantum 2.0 models
Google translation below of key paras

The 62-qubit "Zuchongzhi" was already the world's superconducting quantum computer with the largest number of bits mentioned in publicly published papers. Now it has completed the self-transcendence "Zuchong No. 2", realized 66 qubits, and adopted a brand-new flip-chip 3D packaging process and adjustable coupling architecture to solve the problem of large-scale bit integration and achieve fast and accurate coupling between bits Adjustable.
"Zuchongzhi No. 2" processes specific tasks 10 million times faster than current fastest supercomputers, and the difficulty of completing tasks is 2-3 orders of magnitude higher than that of Google "Plananus".
On the other route, the calculation scale and complexity of "Nine Chapters No. 2" has been significantly improved compared with that of "Nine Chapters": "Nine Chapters" builds a quantum computing prototype with 76 photons and 100 modes. "Nine Chapter 2" successfully constructed a quantum computing prototype with 113 photons and 144 modes. The task completed by "Nine Chapters" in one minute, the supercomputer "Fuyue" takes 100 million years; the task completed by "Nine Chapter 2" in one millisecond, "Fuyue" needs 30 trillion years. Most importantly, "Nine Chapter Two" also has some programmability.

Changguangchen Core successfully passed the acceptance of major nuclear high-tech projects​

Nuclear high base of major projects is one of the "Long-term Scientific and Technological Development Plan (2006-2020)" determined by national sixteen major science and technology, its purpose in the nuclear heart of electronic devices, high- end general chips and based foundation software, catch up with international technology and the rapid development of the industry. In October 2021, Changguangchen Core's major nuclear-high-base special project "8K ultra-high-definition image sensor chip and system application" successfully completed the acceptance work.
Changguang Chenxin takes the lead in the major special project "8K ultra-high-definition image sensor chip and system application" of Nuclear High-tech, and cooperates with Zhejiang Huarui Technology Co., Ltd., Shenzhen DJI Innovation Technology Co., Ltd. and Shenzhen DJI Piovan. Jointly undertaken by Technology Co., Ltd., the project was approved by the Ministry of Industry and Information Technology in 2019. This project aims to develop 8K ultra-high-definition CMOS image sensor chips and camera systems with independent intellectual property rights, and break the situation that my country's ultra-high-definition imaging chips and systems have long been dependent on foreign imports and their development is severely limited. During the project implementation period, with the support of the Ministry of Industry and Information Technology and various local governments, the four development units worked together to finally complete the development of two 8K ultra-high-definition CMOS image sensors. At the same time, based on the two image sensors, the national production and independent The development of a controllable and competitive 8K ultra-high-definition camcorder. The smooth acceptance of this project will surely bring new changes to my country's ultra-high-definition industry, and will also drive my country's ultra-high-definition display industry and 5G ultra-high-speed transmission The rapid upgrading of other industries.
Changguangchenxin is a company specializing in the design of high-end, high-performance CMOS image sensors. Its core applications include high-end scientific imaging and industrial vision imaging. The CMOS image sensor chip required in the 4K/8K ultra-high-definition camera system has high performance requirements, such as ultra-low readout noise, ultra-high dynamic range, and high frame rate. It is highly compatible with the technical route and capabilities of Changguangchen Core. Therefore, Changguangchen Core also regards professional photography as the company's next important market direction. With the support of this project, Changguangchen Core has completed the development of an 8K full-frame, back-illuminated, stacked CMOS image sensor with a resolution of 49 million pixels, a 4.3-micron pixel design, and readout noise. Less than 2e-, the single-frame dynamic range reaches 87dB; the chip uses the most advanced back-illuminated, stacking technology, under 16bit ADC output, the frame rate is up to 120fps in 8K mode, and up to 240fps in 4K mode. The chip is extremely advanced in terms of process platform selection and performance indicators, and it will also speed up the pace of my country's ultra-high-definition camera system to catch up with the international leading level.
With the support of the special project, Changguangchen Core has broken through a number of key technologies and stacking processes, and pressed the fast forward button for the company in the development of ultra-high-definition 8K/4K CMOS image sensors, forming the GCINE series for the radio and television field Chemical products to meet the needs of many industries such as 8K broadcast television, SLR cameras, drones, and high-end 8K video surveillance, and accelerate the pace of upgrading my country’s ultra-high-definition industry.

Figure 1: The physical 8K CMOS chip

gelgoog said:

It is just pilot production. If the pilot batch proves out ok, mass production might be a year or two away from that. Assuming everything works ok. The industry moves quite slowly. It takes months just to get the wafers out once you start production for example. Then they have to be tested for defects and the process might need to be improved. It is an iterative and slow process.

3nm would be useful not just for mobile but also server CPUs and large GPUs.

Click to expand...

Yes, I agree that 3nm would be useful for server CPU, but it all depends on the price and server don't have power (battery) limitation. I don't think 3nm is able to compete in price with lets say 28 or 14nm chips

Ali's 3D stacking chip breakthrough

Alibaba successfully developed a new architecture chip. This chip is the world's first DRAM-based 3D bonded stacked storage and computing integrated AI chip, which can break through the performance bottleneck of the von Neumann architecture and meet the needs of artificial intelligence and other scenarios for high bandwidth, high capacity memory and extreme computing power. In certain AI scenarios, the performance of the chip is increased by more than 10 times, and the energy efficiency ratio is increased by up to 300 times.

antiterror13 said:

Yes, I agree that 3nm would be useful for server CPU, but it all depends on the price and server don't have power (battery) limitation. I don't think 3nm is able to compete in price with lets say 28 or 14nm chips

Click to expand...

But they do. If you have a huge datacenter with millions of CPUs the power consumption is a major part of the expense of running it.

foofy said:

Ali's 3D stacking chip breakthrough

Alibaba successfully developed a new architecture chip. This chip is the world's first DRAM-based 3D bonded stacked storage and computing integrated AI chip, which can break through the performance bottleneck of the von Neumann architecture and meet the needs of artificial intelligence and other scenarios for high bandwidth, high capacity memory and extreme computing power. In certain AI scenarios, the performance of the chip is increased by more than 10 times, and the energy efficiency ratio is increased by up to 300 times.

Please, Log in or Register to view URLs content!

Click to expand...

Did I read this correctly they stacked computing and memory? The mention of storage makes it sounds a bit confusing, I see a lot of people and sometimes do it myself using both term memory and storage interchangeably.

Sounds like what AMD is doing with their compute and Cache stacking in their zen chips.

gelgoog said:

But they do. If you have a huge datacenter with millions of CPUs the power consumption is a major part of the expense of running it.

Click to expand...

depends on how much you care about upfront cost, or in the case of China, about the data center existing. and we can do a calculation of run cost vs. cost of the chips needed, and we may find that it isn't a big deal until decades in the future. I don't have the time to do that right now but just look up KWH cost of electricity vs per wafer cost of EUV chips.