Chinese semiconductor thread II

tphuang

Lieutenant General
Staff member
Super Moderator
VIP Professional
Registered Member
the question for Loongson is not whether it's good or not good enough. it is good enough. The question is why would SOEs and others buy this if they can get better product from Huawei? Whether we like it or not, Huawei has 7nm capacity mostly to itself for this year and next. That's why Loongson is stuck to 12nm process for 3A6000. Huawei core was said to be close to Zen 3 level and it has a much larger ecosystem due to Kunpeng developer pool and ARM
 

measuredingabens

Junior Member
Registered Member
Please, Log in or Register to view URLs content!

Please, Log in or Register to view URLs content!

Researchers overcome lattice mismatch issue to advance optoelectronic applications

Aligned with the third-generation detector SWaP3 concept (Size, Weight, Power, Price, Performance), the latest generation of optoelectronic devices is trending towards miniaturization, flexibility and intelligence, Professor Ho emphasized. "The lattice-mismatch-free construction of core-shell heterostructure nanowires holds great promise for next-generation ultrasensitive SWaP3 optoelectronics," he said.

This pioneering research encompasses innovative material design, novel process development, and the exploration of new optoelectronic applications. The initial focus involves the investigation of an amorphous shell composed of chalcogenide covalent-bond networks, strategically employed to address the lattice mismatch issue surrounding the III–V core.

The successful achievement of effective lattice-mismatch-free construction in the core-shell heterostructure introduces unconventional optoelectronic properties. Notably, these properties include bi-directional photoresponse, visible light-assisted infrared photodetection, and enhanced infrared photodetection.
 

ht1688

New Member
Registered Member
Please, Log in or Register to view URLs content!

Bloomberg) -- The Chinese government has quietly asked electric-vehicle makers from BYD Co. to Geely Automobile Holdings Ltd. to sharply increase their purchases from local auto chipmakers, part of a campaign to reduce reliance on Western imports and boost China’s domestic semiconductor industry. The country’s tech overseer previously set an informal target for automakers to source a fifth their chips locally by 2025, but has grown dissatisfied with the pace of progress, the people said.

The ministry is now directly instructing firms to avoid foreign semiconductors if at all possible, said the people, who asked to remain anonymous discussing sensitive information. That means overseas chip firms effectively have to manufacture their silicon through a local foundry such as Semiconductor Manufacturing International Corp. or Hua Hong Semiconductor Ltd., one of the people said. During a recent tender conducted by a major domestic brand, one foreign bidder failed to secure a contract despite offering a price they estimated was 30% lower than the eventual winner, another person said.

The developments reflect Beijing’s stepped-up efforts to galvanize its own tech sector, a response to US attempts to curtail Chinese chip development via sanctions and restrictions on the sale of advanced technology. The Chinese directive on cars casts uncertainty over the business of companies from Nvidia Corp. and NXP Semiconductor NV to Renesas Electronics Corp. and Texas Instruments Inc., which compete with local firms to supply the world’s largest EV market.

Many local brands buy a hodge-podge of other components from power management chips to microcontrollers and display ICs. Some carmakers from leader BYD to upstarts like Nio Inc. rely on Nvidia processors for instance to coordinate functions in their hyper-connected cars.

Beijing is increasingly intent on supporting a domestic chip industry it knows is lagging well behind foreign rivals, but critical to supporting the economy and maintaining a geopolitical advantage. The directive is in line with pledges this year to mobilize all means at its disposal to wrest technological supremacy from the US and other nations.

The central government may also be bracing for curbs in the auto sector specifically.

Washington is probing potential data and cyber-security risks posed by internet-connected cars from China, including EVs. US Commerce Secretary Gina Raimondo has expressed concern that data collected by the cars could wind up in Beijing’s hands, and is overseeing an investigation that could result in restrictions on the sale of those cars in the US — measures that would go beyond a potential increase in tariffs.

A mandate in favor of Chinese chips also raises the prospect of further retaliation as Washington ramps up its containment campaign.

At the Two Sessions meeting of lawmakers this month, Beijing hammered home a longstanding goal to break an American stranglehold on key spheres. Under Xi Jinping’s watch, China has expanded state control of strategically critical areas from semiconductor manufacturing to quantum computing.

Against that backdrop, Beijing has endorsed the efforts of local champions such as Huawei Technologies Co. through subsidies and targeted policies. It’s now establishing a $27 billion-plus fund to drive chip investments and hiking spending on nationwide research to more than $51 billion.

China’s biggest auto chip and software providers include Huawei, Loongson Technology Corp., GigaDevice Semiconductor Inc. and Wingtech Technology Co. Many, like Novosense Microelectronics Co. and StarPower Semiconductor Ltd., have built significant capacity in so-called mature or lower-end nodes. The MIIT’s policy may soak up some of that capacity — at the expense of their foreign rivals.

The US and its European allies, meanwhile, have grown alarmed by China’s rush into older-generation chips. The Biden administration is mulling tariffs on those components in addition to other trade tools, but has said it is not considering restricting them using export controls.

While US rules introduced in 2022 slowed China’s development of advanced chipmaking capabilities, they left largely untouched the country’s ability to use techniques older than 14-nanometers. That has led Chinese firms to construct new plants faster than anywhere else in the world. They are forecast to build 26 fabs through 2026 that use 200-millimeter and 300-mm wafers, according to data from the trade group SEMI as of 2023. That compares with 16 fabs for the Americas.
Back in 2022, somebody posted this article on a teardown of Yuan by BYD:
Please, Log in or Register to view URLs content!


I had thought that by 2022, most of the chips would be localized. To my great disappointment, it did not look like much effort was spent to do this. As can be seen from the article, most of the chips are western, including chips from NXP, Sandisk, ADI, Lattice, Maxim, and TI. A lot of the chips are mixed signal, which I think might still be unavailable from Chinese suppliers (though they should be pretty low end in terms of performance). What surprised me most was that some of the chips are just decades old microcontrollers. They should have chinese equivalents. Porting the code would require a little effort, but if you want to localize, this is unavoidable. Porting the DSP code on the motor controller from the TI DSP to something else, however, is a little harder but still necessary if China is ever going to decouple from the US.

The Bloomberg article is a little misguided in that this really doesn't have much to do with fabs (though the processing required for self driving might require more advanced nodes). But as the teardown above shows, the US can technically shut down the Chinese auto industry if it stops these companies from selling their chips to China. Sourcing 20% locally by 2025 is a very modest goal, especially since a lot of power semiconductors are probably already sourced locally. Chinese auto makers really need to have a greater sense of urgency. China does have some leverage in that the US is in a similar situation vis-a-vis China in even more areas (eg., in pharmaceuticals). But if I were BYD, this would be one of my highest priorities.

This has bothered me since I read that article in 2022. I don't think localizing a car will be any easier than Huawei localizing a phone. There are hundreds of chips to substitute, and I don't even know if they control all the algorithms/codes. Judging from the 2022 teardown and this article, they aren't making much progress. Either they are not capable of the technical task. Or they don't think relying on western ICs is a big risk.
 

staplez

New Member
Registered Member
Back in 2022, somebody posted this article on a teardown of Yuan by BYD:
Please, Log in or Register to view URLs content!


I had thought that by 2022, most of the chips would be localized. To my great disappointment, it did not look like much effort was spent to do this. As can be seen from the article, most of the chips are western, including chips from NXP, Sandisk, ADI, Lattice, Maxim, and TI. A lot of the chips are mixed signal, which I think might still be unavailable from Chinese suppliers (though they should be pretty low end in terms of performance). What surprised me most was that some of the chips are just decades old microcontrollers. They should have chinese equivalents. Porting the code would require a little effort, but if you want to localize, this is unavoidable. Porting the DSP code on the motor controller from the TI DSP to something else, however, is a little harder but still necessary if China is ever going to decouple from the US.

The Bloomberg article is a little misguided in that this really doesn't have much to do with fabs (though the processing required for self driving might require more advanced nodes). But as the teardown above shows, the US can technically shut down the Chinese auto industry if it stops these companies from selling their chips to China. Sourcing 20% locally by 2025 is a very modest goal, especially since a lot of power semiconductors are probably already sourced locally. Chinese auto makers really need to have a greater sense of urgency. China does have some leverage in that the US is in a similar situation vis-a-vis China in even more areas (eg., in pharmaceuticals). But if I were BYD, this would be one of my highest priorities.

This has bothered me since I read that article in 2022. I don't think localizing a car will be any easier than Huawei localizing a phone. There are hundreds of chips to substitute, and I don't even know if they control all the algorithms/codes. Judging from the 2022 teardown and this article, they aren't making much progress. Either they are not capable of the technical task. Or they don't think relying on western ICs is a big risk.
Unfortunately if anything, replacing car chips would be harder than Huawei replacing the chips for their phones. This is because cars have mission critical chips which failure could mean death. Toyota was recently sued and lost because one of their fly by wire modules failed from a memory overload. This caused the car's brakes to fail and killed some people.

Now, obviously China can make these chips, they're not all that advanced. But unlike a cellphone, these chips would require far more testing and guarantees than a chip for a phone.
 

FairAndUnbiased

Brigadier
Registered Member
Unfortunately if anything, replacing car chips would be harder than Huawei replacing the chips for their phones. This is because cars have mission critical chips which failure could mean death. Toyota was recently sued and lost because one of their fly by wire modules failed from a memory overload. This caused the car's brakes to fail and killed some people.

Now, obviously China can make these chips, they're not all that advanced. But unlike a cellphone, these chips would require far more testing and guarantees than a chip for a phone.
A wide variety of labs have accelerated life testing capabilities, temperature and humidity cycling, failure analysis, etc. These are well publicized standards and labs do this on a daily basis commercially.
 

european_guy

Junior Member
Registered Member
the question for Loongson is not whether it's good or not good enough. it is good enough. The question is why would SOEs and others buy this if they can get better product from Huawei? Whether we like it or not, Huawei has 7nm capacity mostly to itself for this year and next. That's why Loongson is stuck to 12nm process for 3A6000. Huawei core was said to be close to Zen 3 level and it has a much larger ecosystem due to Kunpeng developer pool and ARM

China is big enough for both.

IMHO it makes little sense to let Loongson die. There is a lot of know how and technical skills in that company, it is a strategic asset that shall be preserved, eventually helping Loongson to pass through these difficult years until China fab capacity for advanced nodes will be enough to serve properly Loongson too.

The investment required is not a biggie, not a lot is needed to keep Loongson moving. Just for some SOE to buy their chips.

IMHO is not a good idea to put all the eggs in one basket...also because there are not so many eggs after all.
 

tphuang

Lieutenant General
Staff member
Super Moderator
VIP Professional
Registered Member
Back in 2022, somebody posted this article on a teardown of Yuan by BYD:
Please, Log in or Register to view URLs content!


I had thought that by 2022, most of the chips would be localized. To my great disappointment, it did not look like much effort was spent to do this. As can be seen from the article, most of the chips are western, including chips from NXP, Sandisk, ADI, Lattice, Maxim, and TI. A lot of the chips are mixed signal, which I think might still be unavailable from Chinese suppliers (though they should be pretty low end in terms of performance). What surprised me most was that some of the chips are just decades old microcontrollers. They should have chinese equivalents. Porting the code would require a little effort, but if you want to localize, this is unavoidable. Porting the DSP code on the motor controller from the TI DSP to something else, however, is a little harder but still necessary if China is ever going to decouple from the US.

The Bloomberg article is a little misguided in that this really doesn't have much to do with fabs (though the processing required for self driving might require more advanced nodes). But as the teardown above shows, the US can technically shut down the Chinese auto industry if it stops these companies from selling their chips to China. Sourcing 20% locally by 2025 is a very modest goal, especially since a lot of power semiconductors are probably already sourced locally. Chinese auto makers really need to have a greater sense of urgency. China does have some leverage in that the US is in a similar situation vis-a-vis China in even more areas (eg., in pharmaceuticals). But if I were BYD, this would be one of my highest priorities.

This has bothered me since I read that article in 2022. I don't think localizing a car will be any easier than Huawei localizing a phone. There are hundreds of chips to substitute, and I don't even know if they control all the algorithms/codes. Judging from the 2022 teardown and this article, they aren't making much progress. Either they are not capable of the technical task. Or they don't think relying on western ICs is a big risk.
that was a car built in 2018. How much domestic chips do you think cars were using back then? Stop being bothered and stop making up numbers like sourcing 20% locally by 2025.

And no, neither BYD nor Chinese auto industry can be shut down overnight.

China is big enough for both.

IMHO it makes little sense to let Loongson die. There is a lot of know how and technical skills in that company, it is a strategic asset that shall be preserved, eventually helping Loongson to pass through these difficult years until China fab capacity for advanced nodes will be enough to serve properly Loongson too.

The investment required is not a biggie, not a lot is needed to keep Loongson moving. Just for some SOE to buy their chips.

IMHO is not a good idea to put all the eggs in one basket...also because there are not so many eggs after all.
I don't think Loongson will die, but Huawei is clearly going to be the dominant choice for domestic CPU procurement program in the future. I'm not saying that's necessarily a good thing
 

tokenanalyst

Brigadier
Registered Member
Back in 2022, somebody posted this article on a teardown of Yuan by BYD:
Please, Log in or Register to view URLs content!


I had thought that by 2022, most of the chips would be localized. To my great disappointment, it did not look like much effort was spent to do this. As can be seen from the article, most of the chips are western, including chips from NXP, Sandisk, ADI, Lattice, Maxim, and TI. A lot of the chips are mixed signal, which I think might still be unavailable from Chinese suppliers (though they should be pretty low end in terms of performance). What surprised me most was that some of the chips are just decades old microcontrollers. They should have chinese equivalents. Porting the code would require a little effort, but if you want to localize, this is unavoidable. Porting the DSP code on the motor controller from the TI DSP to something else, however, is a little harder but still necessary if China is ever going to decouple from the US.

The Bloomberg article is a little misguided in that this really doesn't have much to do with fabs (though the processing required for self driving might require more advanced nodes). But as the teardown above shows, the US can technically shut down the Chinese auto industry if it stops these companies from selling their chips to China. Sourcing 20% locally by 2025 is a very modest goal, especially since a lot of power semiconductors are probably already sourced locally. Chinese auto makers really need to have a greater sense of urgency. China does have some leverage in that the US is in a similar situation vis-a-vis China in even more areas (eg., in pharmaceuticals). But if I were BYD, this would be one of my highest priorities.

This has bothered me since I read that article in 2022. I don't think localizing a car will be any easier than Huawei localizing a phone. There are hundreds of chips to substitute, and I don't even know if they control all the algorithms/codes. Judging from the 2022 teardown and this article, they aren't making much progress. Either they are not capable of the technical task. Or they don't think relying on western ICs is a big risk.
You don't understand how this works. Let me repeat myself "The only thing that US politicians needed to do to stop China semiconductor industry is just to sell them everything they want", They saw what happen to ZTE and then Huawei and then SMIC, they get warned "Hey you should derisk you supply chain, at least try to increase the source your components more locally" "maybe getting into partnerships with local companies like Huawei did ensure quality and availability" But no the CEO will always contort backwards and "NO THEY WILL RESPECT THE MARKET", "NOT GOOD ENOUGH" ,"ONLY SLOW AND STEADY DEVELOPMENT", "IT WILL TAKE A MILLIONS TO REACH X COMPANY LEVEL", "THIS INDUSTRY IS DIFFERENT", "LET NOT REPEAT THE MISTAKE OF THE PAST", the quick money, is easier to work with chips that already have a nice easy to access development ecosystem, that build something strong and more immune to external factors from the ground up.

But meanwhile Mr. Elon Musk probably is already conspiring in secret with BIS idiots on how to get BYD in blacklist, because is "National Security Threat", that they helping the PLA to build electric vehicles or something crazy.

And then BAM, the hammer comes, hard as steel, If they are lucky, they got a few month notice to stockpile enough to survive another year, people running like crazy, phone ringing everywhere, NOW, they are calling Chinese IC companies to see if they have equivalent chips, at the same time begging in their knees to see if BIS can allow them to buy chips again, licenses that will take years, the China MFA will release an statement saying " US unfairness, undoing market rules, National Security excuse, bla bla bla" things US politicians give a rat ass.

Then CEO goes head looking down, to a government official begging for money that could have gone to a semiconductor equipment company or a semiconductor material company. So the company now has to build everything almost from the ground up, the software development ecosystems, the IC partnerships and so on.

If American politicians and their think tankers were not so dumb, so ideologically blinded or at least try to put an minimal effort to justify the money they are getting overpaid for, they would understand this dynamic, day one. but no, too much effort.​
 

tphuang

Lieutenant General
Staff member
Super Moderator
VIP Professional
Registered Member
Please, Log in or Register to view URLs content!

Last year, news came out that CETC's 28nm ion implantation machine was going into production lines or something like that. This article talks about it a little bit

Says in there that through hard work, they improved up time by 70% and WPH by a lot. Reaching international leading level. That seems to be the key here. Not only do the machines need to work, they also need to work reliably in HVM.
 

tokenanalyst

Brigadier
Registered Member
Competition is heating between Naura and AMEC. Naura announced to their investors their Damascus Etching tool, while AMEC is getting their validated. Staying in the same old process nodes is not an option know.

Northern Huachuang officially released a new product - the 12-inch capacitively coupled plasma dielectric etching machine Accura LX. This machine mainly covers multiple technology nodes in the logic field, including AIO (double Damascus etching process), Contact (contact hole) It represents the key dielectric etching process and can be extended to CMOS (complementary metal oxide semiconductor) related processes in the storage field. It is suitable for the etching process of Low-K (low dielectric constant) dielectric and conventional dielectric materials.

 
Top