Opinionated - China Chipping Away to Semiconductor Dominance

[JSCh]

China MEMS foundry startup boosts monthly output to 70,000 wafers

Semiconductor Manufacturing Electronics (ShaoXing), a China-based MEMS foundry startup formed by Semiconductor Manufacturing International (SMIC) and the local government, has managed to boost its monthly production output to 70,000 wafers from 40,000 in November 2020, according to industry sources.

www.digitimes.com

 

[JSCh]

22:14, 17-Apr-2021
'Historic breakthrough': Chinese company unveils self-developed CPU architecture
CGTN

File photo of a Loongson chip. /CFP

Loongson Technology, a leading Chinese chip company based in Beijing, unveiled on Thursday its fully self-developed CPU (central processing unit) architecture that uses an original instruction set, which has been hailed as a "historic breakthrough" by industry insiders.

The new infrastructure is named "LoongArch", short for "Loongson Architecture". It has passed the assessment of a leading third-party intellectual property evaluating agency, the company said.

An instruction set architecture (ISA) is the interface between computer hardware and software. It's the underlying architecture that supports the running of an operating system and a whole software ecosystem.

For example, the Windows operating system (OS) runs on U.S. company Intel Corporation's X86 architecture, and the Android mobile OS runs on ARM architecture developed by UK-based chip designer ARM Holdings.

The two best-known ISAs have been dominating on computers and smartphones worldwide since 2010, and both have very large industrial ecosystems including hardware like chips and software such as various applications.

Foreign CPU manufacturers have been using their ISAs as a tool to control the ecosystems and exclude competitors. Other companies have to pay royalties to get the licenses to develop CPUs compatible with them. If Chinese companies rely on those foreign ISAs, it's not possible for them to build an independent industrial ecosystem, according to the company.

"If we compare chip designing to writing articles, an ISA is like the language we use. Chinese people can write an article in English, but can never develop our national culture based on English," Hu Weiwu, chairman and chief scientist of Loongson, was quoted as saying by the Beijing Daily. He's also the chief engineer of the Institute of Computing Technology under the Chinese Academy of Sciences (CAS).

The "LoongArch", which is an outcome of the company's 20 years of hard work and experience in CPU development and ecosystem building, has nearly 2,000 proprietary instructions, the company said.

It's also compatible with other mainstream ISAs, allowing the efficient operation of complicated programs written for other architectures.

So far, Loongson's CPU chip based on the new architecture, named "3A5000", has been taped out and is under internal test. Its processing performance is very close to that of mainstream products on the market, Hu said. A complete OS based on the "LoongArch" has also run stably on computers using the chip.

"Only with our proprietary ISA can we build a new pattern in the information technology industry and form our own industrial chain," Hu told the Science and Technology Daily.

Loongson Technology was jointly funded and set up in 2010 by the Beijing municipal government and the CAS, with the goal of industrializing the technological achievements of the CAS in chip research and development.

IT之家​

7-23 09:47 来自 IT之家​

​

【龙芯 3A5000 正式发布：首款采用自主指令系统 LoongArch 的处理器芯片】日前，龙芯中科技术股份有限公司正式发布龙芯 3A5000 处理器。该产品是首款采用自主指令系统 LoongArch 的处理器芯片，性能实现大幅跨越，代表……详情点击：龙芯 3A5000 正式发布：首款采用自主指令系统 LoongArch 的处理器芯片​

IT House
7-23 09:47 From IT House

[Loongson 3A5000 officially released: the first processor chip using own instruction set architecture LoongArch]

Recently, Loongson Technology Co., Ltd. officially released the Loongson 3A5000 processor. This product is the first processor chip that uses Loongson own instruction set architecture LoongArch, and its performance has achieved a significant leap, representing...Click for details: Loongson 3A5000 officially released: the first processor chip that uses the autonomous command system LoongArch

 

[JSCh]

https://twitter.com/x/status/1418637368802701314

 

[Beidou2020]

Huawei's first flexible OLED driver chip enters trial production, report says
Yvonne Tang
July 19, 2021

Huawei's chip division HiSilicon's first flexible OLED driver chip has entered trial production and is expected to be delivered by the end of this year, ijiwei.com reported Sunday, adding that Huawei's products are also expected to use it.

The chip uses a 40nm process and is scheduled for mass production in the first half of next year, with a monthly capacity of 200 to 300 wafers, the report said, citing Isaiah Research CEO Zeng Mengbin, adding that samples have been sent to BOE, Huawei and Honor for testing.

OLED screens have a large number of pixel dots, all independently illuminated, but need to be lit by OLED driver chips for each pixel dot.

The flexible OLED driver chip is the main direction of HiSilicon, Magirror Research analyst Si Maqiu pointed out.

HiSilicon's entry into the OLED driver chip market is a deliberate move under the changing landscape, as it does not require advanced processes, the report said.


Since the US set a chip ban on Huawei, global foundries cannot produce chips using 14nm, 7nm, 5nm and other advanced processes for HiSilicon, and the supply of 28nm and above mature process foundry process also requires permission from the United States.

At present, OLED driver chips mainly use 28nm, 40nm and 55nm mature process technology, and the US is likely to license foundries to produce for Huawei, the report noted.

Huawei's first flexible OLED driver chip enters trial production, report says - CnTechPost

Huawei's chip division HiSilicon's first flexible OLED driver chip has entered trial production and is expected to be delivered by the end of this year, ijiwei.com reported.

cntechpost.com

 

[Beidou2020]

Chinese chipmaker YMTC's 128-layer memory chip achieves mass production
Global Times
05:22 Jul 30 2021

Photo: ymtc.com
Wuhan-based Yangtze Memory Technologies Co (YMTC) has put into mass production a 128-layer memory chip, a feat of domestic chipmaking in terms of stacking density, addressing China's missing link in high-end solid state drives (SSDs), according to media reports.

Powev Electronic Technology Co, a Shenzhen-based storage and memory manufacturer, recently took the wraps off its newest upscale SSD under its subsidiary brand, Asgard, based on YMTC's 128-layer 3D three-level cell (TLC) NAND flash memory. According to industry insiders, the announcement suggests YMTC's 128-layer technology has officially been massed produced, Chinese news site guancha.cn reported on Thursday.

The Wuhan-based 3D NAND flash memory maker was only established in July 2016. In April 2020 it announced that its 128-layer technology had passed sample verification on the SSD platform, less than a year after its 64-layer TLC 3D NAND flash memory was produced in volume.

Higher density will allow greater storage capacity. With the 129-layer lineup, YMTC is closing the gap with South Korean memory giants SK Hynix and Samsung, among other established chipmakers.

SK Hynix announced in June 2019 that it began mass-producing the world's first 128-layer 4D NAND.

In a radical breakthrough, Micron, a smaller rival to the two South Korean conglomerates, revealed in November 2020 that it has commenced volume shipments of the world's first-ever 176-layer 3D NAND flash memory, boasting industry-pioneering density.

However, YMTC's notable progress over the past five years remained insufficient to give it a boost in the global memory rankings, industry data showed.

In the first quarter of the year, Samsung took the top spot in revenue in the NAND flash market across the world with a 33.5 percent share of the global market, according to market intelligence provider TrendForce.

Japanese memory manufacturer Kioxia ranked No.2 with 18.7 percent of the market, followed by US-based Western Digital with 14.7 percent, SK Hynix with 12.3 percent, Micron with 11.1 percent, and Intel with 7.5 percent, according to TrendForce data.

This means the six major memory firms held nearly 98 percent of the entire global market of NAND flash memories.

 

[jamahir]

OK when yr Microsoft and Linux beating OS comes out and your 10GHZ processor is ready, don't forget to ring me. And for the last time professor, you should tell me why there is a 2000 line code.

Hi, read this post of mine which has two article quotings that explain why the supposedly indigenous LoongArch Chinese processor has 2500 instructions. Turns out the processor's base design is not Chinese at all.

 

[Globenim]

OK when yr Microsoft and Linux beating OS comes out and your 10GHZ processor is ready, don't forget to ring me. And for the last time professor, you should tell me why there is a 2000 line code.

Its just the usual conartist routine trying to establish themself as some insider and professional with repeated vague anecdotes about scrubbing someone buttsweat off a pilotseat and holding the stick on a Cesna with a trained co-pilot for 5 minutes sold as "I worked on some aircraft and piloted a real aircraft" to make themself out as a fighter pilot on the internet and dumb "just figure yourself out what im trying to say" wikipedia and google snippets because they cant actually articulate their textbook and google knowledge to support any of their ignorant opinions beyond their fleeting undergrad and remote viewer experience, while not actually ever saying anything conrete that supports their vapid nationalistic denial and baseless opinions. You get these "I have 10 years experience" types everywhere in the West and South Asia.

 

[jamahir]

Its just the usual conartist routine trying to establish themself as some insider and professional with repeated vague anecdotes about scrubbing someone buttsweat off a pilotseat and holding the stick on a Cesna with a trained co-pilot for 5 minutes sold as "I worked on some aircraft and piloted a real aircraft" to make themself out as a fighter pilot on the internet and dumb "just figure yourself out what im trying to say" wikipedia and google snippets because they cant actually articulate their textbook and google knowledge to support any of their ignorant opinions beyond their fleeting undergrad and remote viewer experience, while not actually ever saying anything conrete that supports their vapid nationalistic denial and baseless opinions. You get these "I have 10 years experience" types everywhere in the West and South Asia.

Instead of this long ridiculing of me you could have spent time to prepare a rebuttal of the sources I have presented and also posted link to some website where the supposedly-Chinese LoongArch processor's architecture is posted.

 

[serenity]

Loongarch includes a instruction set architecture that is MIPS compatible.

It would be stupid to not. That's like starting audio player product that is not compatible with existing files for music.

Loongarch is both MIPS and RISC compatible. Including of course similarities to both.

The architecture design can be using the same as MIPS, RISC V. Why create a different design? The design is like the shape for the wheel. Recreating it has no advantage for Loongson.

Creating something that can be made and useful is the purpose. The implementation is based on MIPS and RISC V. When the Americans banned Chinese from using American made chips for supercomputers in the 2010s, we used our own a few years later. The speeds were still topping the supercomputer list for two cycles between US and China for top spots. Only since 2019, it is barely reported anymore from both sides since it led to some speed war and Japan then took lead. Both US and China stopped announcing their newest.

I don't think there is any advantage or purpose for Chinese or even anyone else new to this to develop their unique architecture. It would have no point. Why invent another 1,2,3,4,5... number system when this one works. Whatever you come up with is unlikely to be better because physics is the same regardless for a new architecture. How the 1,2,3,4,5... architecture is used is where the important difference is.

 

[jamahir]

Loongarch includes a instruction set architecture that is MIPS compatible.

It would be stupid to not. That's like starting audio player product that is not compatible with existing files for music.

Loongarch is both MIPS and RISC compatible. Including of course similarities to both.

The architecture design can be using the same as MIPS, RISC V. Why create a different design? The design is like the shape for the wheel. Recreating it has no advantage for Loongson.

Creating something that can be made and useful is the purpose. The implementation is based on MIPS and RISC V. When the Americans banned Chinese from using American made chips for supercomputers in the 2010s, we used our own a few years later. The speeds were still topping the supercomputer list for two cycles between US and China for top spots. Only since 2019, it is barely reported anymore from both sides since it led to some speed war and Japan then took lead. Both US and China stopped announcing their newest.

I don't think there is any advantage or purpose for Chinese or even anyone else new to this to develop their unique architecture. It would have no point. Why invent another 1,2,3,4,5... number system when this one works. Whatever you come up with is unlikely to be better because physics is the same regardless for a new architecture. How the 1,2,3,4,5... architecture is used is where the important difference is.

My base point was that the Chinese claim of having designed a new ISA has turned out to be false and I was notifying this to @Han Patriot who was disparaging to me and in his hurry to do this he spoke about the LoongArch processor having "2000 lines of code" as if it is software ( or hardware under design ) instead of "2000 instructions". He should have known the difference since he said that he is an Electronics and Electrical engineer.

Secondly, 1,2,3,4,5... is permanent but processor ISA and architectures can evolve in technological and economic areas. Why then has the open source RISC-V processor architecture come up in recent years ? You yourself mention RISC-V.

Thirdly, about unique processor architecture, other than any economic advantage it can present also an intellectual challenge. Do it for this. Do it for fun and adventure. What else would be the use of having those computing PhDs in Tsinghua University, who Han Patriot so arrogantly trumpeted, if all they do is present others' work as their own ?

 

[serenity]

My base point was that the Chinese claim of having designed a new ISA has turned out to be false and I was notifying this to @Han Patriot who was disparaging to me and in his hurry to do this he spoke about the LoongArch processor having "2000 lines of code" as if it is software instead of "2000 instructions". He should have known the difference since he said that he is an Electronics and Electrical engineer.

Secondly, 1,2,3,4,5... is permanent but processor ISA and architectures can evolve in technological and economic areas. Why then has the open source RISC-V processor architecture come up in recent years ? You yourself mention RISC-V.

Thirdly, about unique processor architecture, other than any economic advantage it can present also an intellectual challenge. Do it for this. Do it for fun and adventure. What else would be the use of having those computing PhDs in Tsinghua University, who Han Patriot so arrogantly trumpeted, if all they do is present others' work as their own ?

Mmm Yes I understand your point but this is also a bit like saying spend the money and time to try and reinvent the number system. There really is no point. There are other areas that desperately need funding and spending doing for fun and trying etc.

The phds from top universities do a lot as well. It's just not redesigning the wheel. Even for example in producing these chips there is thousands of man hours in each engineer's dedication towards. We are not wealthy enough to "waste" time and money to do certain things. For others like pointless space exploration, they are spending that type of money and energy because of future application to start developing the technology course today for example.

On ISA, I don't think China has ever claimed Loongarch as one type is unique from RISC and MIPS. They said it is designed to be compatible with both MIPS and RISC. The important goal with this is to have something they can use and make themselves.

Just like for example American engineers use Hindu number systems to do maths right. They don't need to invent their own or use Roman numerals again to do calculations. The important pride isn't about what number system you use but what the product of that is. For this ISA case the product is to get an ability to manufacture and use its own ISA even if it is using similar architecture.

The academic and for fun argument I understand but please also understand in this area, there is no point to do so. I think the Chinese aim is to develop the product of the tool. This is just the tool itself.

 

[jamahir]

For others like pointless space exploration, they are spending that type of money and energy because of future application to start developing the technology course today for example.

About space exploration please watch this interview of an American company named Relativity Space which is developing 3D Printed rockets to first going up into Low Earth Orbit with its first generation rocket and eventually in the next one to Mars and the Moon. While the co-founder does talk about industrialization on Mars he towards the end also speaks of his philosophical reasoning for this effort. Same for Elon Musk. It is about contributing to humanity's future instead of creating some product for current temporary needs.

 

[serenity]

About space exploration please watch this interview of an American company named Relativity Space which is developing 3D Printed rockets to first going up into Low Earth Orbit with its first generation rocket and eventually in the next one to Mars and the Moon. While the co-founder does talk about industrialization on Mars he towards the end also speaks of his philosophical reasoning for this effort. Same for Elon Musk. It is about contributing to humanity's future instead of creating some product for current temporary needs.

When I talk about product of an effort, I mean the end result. So for China it is to be able to produce and supply its own chips for certain needs of industries. It cannot match the best from Taiwan and South Korea right now but on design there is no issue. On fabrication it is lagging about 2 generations or roughly 5 to 10 years.

For contributing to humanity's future? Well I would say things like agricultural biotechnology advances made by China since 1970s already helps in this. It's not only western minds that can do this. Attempt to reverse pollution damage. Planting entire forests, stopping half of China's old coal plants for energy and turning more modern ones into far more efficient ones for scrubbing carbon away from being emitted. More reusable sources. Even enforcing one child policy. All these are contributing isn't it.

3D printing parts has been done in China as well since 3D printing became more well known. 3D printing aircraft parts and rocket parts too just not the whole thing. That Relativity Space company may be doing it in a certain niche area which suits their business model. Doesn't mean they are the only one thinking of different ways to do things.
I should say people like Elon Musk work for their personal vision of world where they rule and hold onto most wealth. They use their abilities and wealth to further this.

They talk in public with grand designs and views of how to shape the world. That's cool okay.

Doesn't mean no one else has dreams.

Also doesn't mean only they are working to better humanity even though they use capitalist models to achieve this which while is fine, often is problematic and almost always begins and ends with everything about profit and more power from increased wealth.

Elon Musk talks about solar roads and solar roofs but stole money from investors to deliver nothing.

He also talks about Mars colonizing which is cool but he's being financed by the US government to achieve this personal dream and US political goal. China's doing this as well as a purely political goal. No profit model required. Just long term vision and developing technology required so one day it becomes more and more achievable.

 

[jamahir]

On ISA, I don't think China has ever claimed Loongarch as one type is unique from RISC and MIPS.

From post# 1397 :

[Loongson 3A5000 officially released: the first processor chip using own instruction set architecture LoongArch]

We are not wealthy enough to "waste" time and money to do certain things.

But wealthy enough to build up a huge military.

So for China it is to be able to produce and supply its own chips for certain needs of industries. It cannot match the best from Taiwan and South Korea right now but on design there is no issue. On fabrication it is lagging about 2 generations or roughly 5 to 10 years.

Technically any country if it spends some billions can set up fabrication plants to produce chips. China has fabs too but yes I think China has produced some of the machinery needed to produce the wafer masks and the actual etchers for the chips. Actually in two or so years China may well catch up with Taiwan and South Korea on producing the currently fashionable 5 nm chips or go below that.

But China is not looking at things like newer materials that can make commercial chips resilient to environmental factors like radiation ( common laptops will not work the Earth's magnetosphere because of the radiation ). Other than materials China is not looking at newer ways now a processor's transistor can be formed. NASA is experimenting to adopt the old vacuum tube methodology for this. All this requires innovation, not just using popular means of production.

For contributing to humanity's future? Well I would say things like agricultural biotechnology advances made by China since 1970s already helps in this.

I don't know about genetic modification experiments for agriculture in China but I do know that China is modernizing the wrong way to go about producing crops. China is putting to use its vast drone manufacturing means and robotic tractors to monitor and manage the farms. Instead of this China should have looking at countrywide adoption of Vertical Farms and general Urban Farms to actually advance agriculture.

 

[serenity]

From post# 1397 :





But wealthy enough to build up a huge military.



Technically any country if it spends some billions can set up fabrication plants to produce chips. China has fabs too but yes I think China has produced some of the machinery needed to produce the wafer masks and the actual etchers for the chips. Actually in two or so years China may well catch up with Taiwan and South Korea on producing the currently fashionable <5 nm chips or go below that.

But China is not looking at things like newer materials that can make commercial chips resilient to environmental factors like radiation ( common laptops will not work the Earth's magnetosphere because of the radiation ). Other than materials China is not looking at newer ways now a processor's transistor can be formed. NASA is experimenting to adopt the old vacuum tube methodology for this. All this requires innovation, not just using popular means of production.



I don't know about genetic modification experiments for agriculture in China but I do know that China is modernizing the wrong way to go about producing crops. China is putting to use its vast drone manufacturing means and robotic tractors to monitor and manage the farms. Instead of this China should have looking at countrywide adoption of Vertical Farms and general Urban Farms to actually advance agriculture.

They called it the Loongarch ISA. This is accurate. The ISA is known as Loongarch. It uses certain architectures from MIPS and RISC. They never hid this in fact it would be so stupid not to. Since it needs to be compatible with MIPS and RISC.

On military, well China is wealthy enough to build up its military in commensurate way right. Last time it did not, the western powers invaded and took over a lot of China and its policies. Why shouldn't China develop its military just a little bit? We honestly haven't even really put that much into this as US has been for about 80 years now. I suppose China is always going to be questioned by you so no need to explain. My point on wealth is to say there's simply no point to do totally new architecture.

Let me say it one more time, there is no benefit to invent a new MIPS or RISC alternative architecture for Godson or any other for this time being. So why would anyone waste money and time to develop their own architecture? Do you realize the MIPS and RISC ones are already highly efficient.

If and when there is better alternative ideas worth putting energy into, then it might be something do to. For now there is simply no point to do own architecture yet when first step is more important than second step. You are saying China should fly when it is trying to learn to walk.

Creating new architecture totally different from MIPS and RISC, is not the hardest thing honestly but it is a relatively pointless endeavor. Of course just because China chooses not to pursue some more wasteful endeavors you are first to say look they cannot do it. lol okay whatever.