What's new

China’s Top Chipmaker Achieves Breakthrough Despite US Curbs


India has ability to manufacture 128nm nodes and aims for sub 100nm for scaled production.

It is not sanctioned and has one foundry using only foreign equipment and tooling.

China has dozens of foundries making various chip types some use foreign equipment and tooling but there are several that are totally cut off and only use Chinese ones.

Military chips don't care about 7nm or 128nm. They care more about shielding. There is hardly an issue for military chip supply. Or space etc. At least for basic consideration. A fighter aircraft using 7nm chips will have better processing than another using 128nm tech.

India's chip foundry is legacy using only foreign equipment and processes. Truth is China's foundries are legacy, mature, and leading edge using various equipment and processes that the foundries themselves have and do not rely on anyone else. The state ones are more or less totally cut off except in some cases where they can purchase a complete tool. It has made steps to totally make everything in house and has achieved this long ago but only with legacy and mature nodes. Only in recent years has some steps for leading edge nodes become mastered and using domestic equipment.

For example Qualcomm has used China's SMIC in the past to make one their designs. Qualcomm is fabless btw. E.g. like Huawei's hisilicon used to get TSMC to make their designs.

What I'm saying is no one is fully vertically and horizontally integrated except South Korea but Japan and China and USA are the closest behind in terms of leading edge and Taiwan too. Taiwan has lots of parts missing too for example they use Dutch asml euvl.

India doesn't make any of the important tooling and equipment for even 128nm fab processes. This is well in legacy nodes but it is fine for military chips. Just not as powerful and effective as more transistors though. It can of course supply those legacy chips but ask India to make even a 28nm node chip and it cannot and who knows how long until it can.

TSMC or anyone opening a factory in India to make 5nm let's day doesn't mean this is India's own. If India makes all the equipment and understands the processes, then it can be called an Indian 5nm chip but the industrial and scientific base required to achieve that simply isn't there. One can say and believe whatever no more than Malaysia and Thailand making 28nm under contract means they can make 28nm independently.

There's much more detail in this potentially confusing topic to navigate and explain so quickly but I hope those reading understand somewhat.
 
.
Chinese pdf posters - I think it is best to ignore the Indian who joined today as he probably got banned from another id and the post style is suspiciously like that nilgiri guy who used to troll BD section day and night and his latest id has just been banned.

Think he may be staying clear of BD section for a while as we would spot him immediately and so a ban on new id.
 
.
A fighter aircraft using 7nm chips will have better processing than another using 128nm tech.
Mostly wrong. For starter, the F-22 CPU technology node is the equivalent of the early Pentium 90nm series. Giving the F-22's avionics new processors on smaller nodes without giving the avionics more work to do will not make the jet fly any better. The new James Webb telescope have a measly 68gb SSD storage.


With the James Webb Space Telescope (JWST) now powered up and snapping some spectacular images, you may wonder exactly how it's storing them. Surprisingly enough, it carries a relatively tiny 68GB SSD, according to IEEE Spectrum — enough to handle a day's worth of JWST images, but not a lot more.​

The 'scope would capture and transmit daily, so there is no need to have a larger capacity storage. The Mars rover Perseverance have an old Apple PowerPC 750 single core 233 mhz.

 
. .
Source: CCP/BBS :taz:

Imports of oxygen concentrators, supplies from China surge

Ananth Krishnan
MAY 08, 2021 20:10 IST
UPDATED: MAY 08, 2021 21:04 IST
Most of the supplies coming in from China have been imported by private Indian companies on a commercial basis.

Oxygenconcentrators.jpg



Indian companies have placed orders for more than 60,000 oxygen concentrators with Chinese medical equipment manufacturers to meet the shortage of medical oxygen in India, which is facing a devastating surge in Covid-19 cases, the top Chinese medical industry association has said


Indians should first learn to make oxygenerator. Then discuss cow dung chip and economy.
 
Last edited:
.
Mostly wrong. For starter, the F-22 CPU technology node is the equivalent of the early Pentium 90nm series. Giving the F-22's avionics new processors on smaller nodes without giving the avionics more work to do will not make the jet fly any better. The new James Webb telescope have a measly 68gb SSD storage.


With the James Webb Space Telescope (JWST) now powered up and snapping some spectacular images, you may wonder exactly how it's storing them. Surprisingly enough, it carries a relatively tiny 68GB SSD, according to IEEE Spectrum — enough to handle a day's worth of JWST images, but not a lot more.​

The 'scope would capture and transmit daily, so there is no need to have a larger capacity storage. The Mars rover Perseverance have an old Apple PowerPC 750 single core 233 mhz.

ah okay thanks for the correction. I assumed smaller nodes and higher density for a military platform means higher processing power.

Anyway point is I'm sure something like even J-20 in PLAAF would be using nothing near 7nm nodes even if production is available. firstly due to when the node is ready for scaled production but also because designers care more about hardening against electromagnetic interference and the rare background radiation effect where an electron could cause a malfunction (can't remember the details) but basically military chips are quite different.
 
.
ah okay thanks for the correction. I assumed smaller nodes and higher density for a military platform means higher processing power.
You assumed correctly, but for what?

We can continue using the F-22 as example. The F-22 have one flight controls feature that others in the US inventory do not have: thrust vectoring. But as far as other flight controls capabilities in the three axes, there is no difference between the F-22 and F-16. The F-22 is called an 'antenna farm' because it has many conformal antennas all over its body. Each antenna is mostly passive in operation, meaning all the antennas simply scoops up signals. So when they designed the F-22, they used the Raytheon Common Integrated Processors that...


The U.S. Air Force's new F-22 Raptor advanced tactical fighter is finally preparing to move into production after more than a decade of development. In the process its avionics architecture has passed through at least three cycles of obsolescence and relies on an Intel microprocessor — the i960MX — that went out of production four years ago.

The reason why the F-22 have the reputation of having two Cray supercomputers is because the avionics package do not have to process much of the extraneous demands like graphics or keyboards. The two CIPs are not available commercially because no one is going to rewrite their game or finance codes for them. Today's smart phones have 10 times the raw processing power the CIPs does, but the CIPs are not burdened like the smart phones are. Making the CIPs in lower nm tech nodes will have no benefits for the F-22 currently locked in capabilities.

Anyway point is I'm sure something like even J-20 in PLAAF would be using nothing near 7nm nodes even if production is available. firstly due to when the node is ready for scaled production but also because designers care more about hardening against electromagnetic interference and the rare background radiation effect where an electron could cause a malfunction (can't remember the details) but basically military chips are quite different.
When I started with Micron Technology, it was in the Burn-In dept. That is where packaged chips are literally baked in an oven while under electrical stress tests. Micron literally test all their products prior to ship, at the wafer level to the packaged level, while most companies does sampling. There was a separate line for military products and all of them were at least two nodes older. Eventually, the separate military line was discarded, but that was for fab efficiency improvement, not because there was no need for a separate product list for unique applications.

The one thing that most people do not understand about tech nodes is that not all customers want the latest. That is why I chuckled at threads like this one. China reached X nm...!!! Every PDF Chinese starts wanking and posting inane comments about how China will take over the semicon industry. Lower node means smaller die structures means smaller dies means smaller packaging means more DRAM or NAND chips on the same PCB. Great achievement. Not going to dispute it. But not every sub-sector of the tech economy want it, let alone need it. Smaller die structures means decreased structural durability and longevity over time of electrical stresses. Conversely, larger die structures equals higher durability and longevity. So which sub-sector of the tech industry want the older nodes, maybe even a decade older? How about smart car? In automotive and aviation, or just transportation overall, an accident is usually catastrophic. Train derailment or a smart car run over a pedestrian. The manufacturers do not care about smaller packaging of chips. There is plenty of room in the car or airplane electronics compartments. The manufacturers want reliability and durability. So if all Chinese semicon companies reached 7nm node and move all their lines to that node, they WILL lose customers. The PC and console gamers and the cell phone companies would remain because their product lines need smaller packaging. But other customers whose products are proven reliable on older tech nodes, they will move to some other suppliers such as Micron who has a history of keeping old products always available for startup.
 
.
You assumed correctly, but for what?

We can continue using the F-22 as example. The F-22 have one flight controls feature that others in the US inventory do not have: thrust vectoring. But as far as other flight controls capabilities in the three axes, there is no difference between the F-22 and F-16. The F-22 is called an 'antenna farm' because it has many conformal antennas all over its body. Each antenna is mostly passive in operation, meaning all the antennas simply scoops up signals. So when they designed the F-22, they used the Raytheon Common Integrated Processors that...


The U.S. Air Force's new F-22 Raptor advanced tactical fighter is finally preparing to move into production after more than a decade of development. In the process its avionics architecture has passed through at least three cycles of obsolescence and relies on an Intel microprocessor — the i960MX — that went out of production four years ago.

The reason why the F-22 have the reputation of having two Cray supercomputers is because the avionics package do not have to process much of the extraneous demands like graphics or keyboards. The two CIPs are not available commercially because no one is going to rewrite their game or finance codes for them. Today's smart phones have 10 times the raw processing power the CIPs does, but the CIPs are not burdened like the smart phones are. Making the CIPs in lower nm tech nodes will have no benefits for the F-22 currently locked in capabilities.


When I started with Micron Technology, it was in the Burn-In dept. That is where packaged chips are literally baked in an oven while under electrical stress tests. Micron literally test all their products prior to ship, at the wafer level to the packaged level, while most companies does sampling. There was a separate line for military products and all of them were at least two nodes older. Eventually, the separate military line was discarded, but that was for fab efficiency improvement, not because there was no need for a separate product list for unique applications.

The one thing that most people do not understand about tech nodes is that not all customers want the latest. That is why I chuckled at threads like this one. China reached X nm...!!! Every PDF Chinese starts wanking and posting inane comments about how China will take over the semicon industry. Lower node means smaller die structures means smaller dies means smaller packaging means more DRAM or NAND chips on the same PCB. Great achievement. Not going to dispute it. But not every sub-sector of the tech economy want it, let alone need it. Smaller die structures means decreased structural durability and longevity over time of electrical stresses. Conversely, larger die structures equals higher durability and longevity. So which sub-sector of the tech industry want the older nodes, maybe even a decade older? How about smart car? In automotive and aviation, or just transportation overall, an accident is usually catastrophic. Train derailment or a smart car run over a pedestrian. The manufacturers do not care about smaller packaging of chips. There is plenty of room in the car or airplane electronics compartments. The manufacturers want reliability and durability. So if all Chinese semicon companies reached 7nm node and move all their lines to that node, they WILL lose customers. The PC and console gamers and the cell phone companies would remain because their product lines need smaller packaging. But other customers whose products are proven reliable on older tech nodes, they will move to some other suppliers such as Micron who has a history of keeping old products always available for startup.
This is really my blind spot of knowledge. Thank you for pointing out my mistake.
 
.
You assumed correctly, but for what?

We can continue using the F-22 as example. The F-22 have one flight controls feature that others in the US inventory do not have: thrust vectoring. But as far as other flight controls capabilities in the three axes, there is no difference between the F-22 and F-16. The F-22 is called an 'antenna farm' because it has many conformal antennas all over its body. Each antenna is mostly passive in operation, meaning all the antennas simply scoops up signals. So when they designed the F-22, they used the Raytheon Common Integrated Processors that...


The U.S. Air Force's new F-22 Raptor advanced tactical fighter is finally preparing to move into production after more than a decade of development. In the process its avionics architecture has passed through at least three cycles of obsolescence and relies on an Intel microprocessor — the i960MX — that went out of production four years ago.

The reason why the F-22 have the reputation of having two Cray supercomputers is because the avionics package do not have to process much of the extraneous demands like graphics or keyboards. The two CIPs are not available commercially because no one is going to rewrite their game or finance codes for them. Today's smart phones have 10 times the raw processing power the CIPs does, but the CIPs are not burdened like the smart phones are. Making the CIPs in lower nm tech nodes will have no benefits for the F-22 currently locked in capabilities.


When I started with Micron Technology, it was in the Burn-In dept. That is where packaged chips are literally baked in an oven while under electrical stress tests. Micron literally test all their products prior to ship, at the wafer level to the packaged level, while most companies does sampling. There was a separate line for military products and all of them were at least two nodes older. Eventually, the separate military line was discarded, but that was for fab efficiency improvement, not because there was no need for a separate product list for unique applications.

The one thing that most people do not understand about tech nodes is that not all customers want the latest. That is why I chuckled at threads like this one. China reached X nm...!!! Every PDF Chinese starts wanking and posting inane comments about how China will take over the semicon industry. Lower node means smaller die structures means smaller dies means smaller packaging means more DRAM or NAND chips on the same PCB. Great achievement. Not going to dispute it. But not every sub-sector of the tech economy want it, let alone need it. Smaller die structures means decreased structural durability and longevity over time of electrical stresses. Conversely, larger die structures equals higher durability and longevity. So which sub-sector of the tech industry want the older nodes, maybe even a decade older? How about smart car? In automotive and aviation, or just transportation overall, an accident is usually catastrophic. Train derailment or a smart car run over a pedestrian. The manufacturers do not care about smaller packaging of chips. There is plenty of room in the car or airplane electronics compartments. The manufacturers want reliability and durability. So if all Chinese semicon companies reached 7nm node and move all their lines to that node, they WILL lose customers. The PC and console gamers and the cell phone companies would remain because their product lines need smaller packaging. But other customers whose products are proven reliable on older tech nodes, they will move to some other suppliers such as Micron who has a history of keeping old products always available for startup.
China's achievement of 7nm does not mean abandoning the production line of old chips, but only means increasing the supply of new customers, such as mobile phone manufacturers.
 
.
You assumed correctly, but for what?

We can continue using the F-22 as example. The F-22 have one flight controls feature that others in the US inventory do not have: thrust vectoring. But as far as other flight controls capabilities in the three axes, there is no difference between the F-22 and F-16. The F-22 is called an 'antenna farm' because it has many conformal antennas all over its body. Each antenna is mostly passive in operation, meaning all the antennas simply scoops up signals. So when they designed the F-22, they used the Raytheon Common Integrated Processors that...


The U.S. Air Force's new F-22 Raptor advanced tactical fighter is finally preparing to move into production after more than a decade of development. In the process its avionics architecture has passed through at least three cycles of obsolescence and relies on an Intel microprocessor — the i960MX — that went out of production four years ago.

The reason why the F-22 have the reputation of having two Cray supercomputers is because the avionics package do not have to process much of the extraneous demands like graphics or keyboards. The two CIPs are not available commercially because no one is going to rewrite their game or finance codes for them. Today's smart phones have 10 times the raw processing power the CIPs does, but the CIPs are not burdened like the smart phones are. Making the CIPs in lower nm tech nodes will have no benefits for the F-22 currently locked in capabilities.


When I started with Micron Technology, it was in the Burn-In dept. That is where packaged chips are literally baked in an oven while under electrical stress tests. Micron literally test all their products prior to ship, at the wafer level to the packaged level, while most companies does sampling. There was a separate line for military products and all of them were at least two nodes older. Eventually, the separate military line was discarded, but that was for fab efficiency improvement, not because there was no need for a separate product list for unique applications.

The one thing that most people do not understand about tech nodes is that not all customers want the latest. That is why I chuckled at threads like this one. China reached X nm...!!! Every PDF Chinese starts wanking and posting inane comments about how China will take over the semicon industry. Lower node means smaller die structures means smaller dies means smaller packaging means more DRAM or NAND chips on the same PCB. Great achievement. Not going to dispute it. But not every sub-sector of the tech economy want it, let alone need it. Smaller die structures means decreased structural durability and longevity over time of electrical stresses. Conversely, larger die structures equals higher durability and longevity. So which sub-sector of the tech industry want the older nodes, maybe even a decade older? How about smart car? In automotive and aviation, or just transportation overall, an accident is usually catastrophic. Train derailment or a smart car run over a pedestrian. The manufacturers do not care about smaller packaging of chips. There is plenty of room in the car or airplane electronics compartments. The manufacturers want reliability and durability. So if all Chinese semicon companies reached 7nm node and move all their lines to that node, they WILL lose customers. The PC and console gamers and the cell phone companies would remain because their product lines need smaller packaging. But other customers whose products are proven reliable on older tech nodes, they will move to some other suppliers such as Micron who has a history of keeping old products always available for startup.

First of all thank you for posting this. Actually informative and good information at least to learn and a post that isn't as bad as a few indian ones earlier on lol. But most of that I was aware of... maybe others not so much. Ignoring CIPs, F-22 still uses chips. I was referring to those chips in those isolated context. Anyway still constructive.

I agree with the parts in bold. This is actually not that important tactically for China but has importance arguably as a strategic issue.

We haven't started wanking or talking about taking over the semiconductor industry. I think that's you feeling this way about us. Firstly absolutely not every industry needs this at all... not even 28nm. That's what I've been saying for years now lol. It wasn't like China was selling billions worth of 7nm and 14nm products like Taiwan, Japan, and South Korea. Most Chinese smartphones used Hisilicon from Huawei fabbed by TSMC. Others used Qualcomm fabbed by whoever Qualcomm wanted to supply depending on cost and other business factors. Qualcomm even got SMIC to make chips at one point. So yeah wasn't anything lost but certainly prevented Chinese industry from buying Chinese 7nm chips and so is important for the blacklisted companies but not for companies that were not blacklisted.

It is simply about having overcome a generation barrier. That in itself is something to celebrate even if it is not for the side you are cheering for.

The customer side of things is small money for China at the moment and unless China becomes a major seller of gaming devices which it isn't yet, then it's nothing lost... it's more a something not gained kind of thing in the past. Anyway again this is more overcoming a limitation that existed before. It doesn't change anything effectively since the 7nm ban itself didn't really change that much to begin with.

Lower threat before (since China could make chips just of 14nm and past) and overcoming the 7nm barrier is overcoming a lower threat.

The issue here is money and has always been. Income from certain industries and in the past, potential income. For Huawei this is obviously good. For most, it means little.

For those cheering China it means a little more.

As for tech war, the real competition here is rarely mentioned, that of alternative tech to silicon. We are reaching the atomic limit of 2nm. Who will commercially bring the next has a slight head start. Then again all this is a marathon rather than a sprint and even then, a sprint still has a back and forth. I will put money on either USA or China reaching commercialization of silicon's replacement. Given where Europe is now and how much funding USA and China have been surely pouring into this. Japan too is in the running as is Taiwan but Taiwan is the master assembler sort of thing with the skills of the processes certainly but perhaps lacking in the depth of funding needed to complete development of the next.
 
Last edited:
. .
Mostly wrong. For starter, the F-22 CPU technology node is the equivalent of the early Pentium 90nm series. Giving the F-22's avionics new processors on smaller nodes without giving the avionics more work to do will not make the jet fly any better. The new James Webb telescope have a measly 68gb SSD storage.


With the James Webb Space Telescope (JWST) now powered up and snapping some spectacular images, you may wonder exactly how it's storing them. Surprisingly enough, it carries a relatively tiny 68GB SSD, according to IEEE Spectrum — enough to handle a day's worth of JWST images, but not a lot more.​

The 'scope would capture and transmit daily, so there is no need to have a larger capacity storage. The Mars rover Perseverance have an old Apple PowerPC 750 single core 233 mhz.
Me agreeing with you? It's the end of days...
 
.
Amazon Web Services uses NAND in the 40 nm and even older nodes. AWS must be up 24/7/365 for a variety of customers, from small town businesses to the Pentagon. Amazon must minimize server downtime for any reason, from scheduled maintenance to migration to custom build for a new customer. Multi level NAND in the older nodes are proven. Amazon WILL leave any supplier who does not have what they need.
 
.
You assumed correctly, but for what?

We can continue using the F-22 as example. The F-22 have one flight controls feature that others in the US inventory do not have: thrust vectoring. But as far as other flight controls capabilities in the three axes, there is no difference between the F-22 and F-16. The F-22 is called an 'antenna farm' because it has many conformal antennas all over its body. Each antenna is mostly passive in operation, meaning all the antennas simply scoops up signals. So when they designed the F-22, they used the Raytheon Common Integrated Processors that...


The U.S. Air Force's new F-22 Raptor advanced tactical fighter is finally preparing to move into production after more than a decade of development. In the process its avionics architecture has passed through at least three cycles of obsolescence and relies on an Intel microprocessor — the i960MX — that went out of production four years ago.

The reason why the F-22 have the reputation of having two Cray supercomputers is because the avionics package do not have to process much of the extraneous demands like graphics or keyboards. The two CIPs are not available commercially because no one is going to rewrite their game or finance codes for them. Today's smart phones have 10 times the raw processing power the CIPs does, but the CIPs are not burdened like the smart phones are. Making the CIPs in lower nm tech nodes will have no benefits for the F-22 currently locked in capabilities.


When I started with Micron Technology, it was in the Burn-In dept. That is where packaged chips are literally baked in an oven while under electrical stress tests. Micron literally test all their products prior to ship, at the wafer level to the packaged level, while most companies does sampling. There was a separate line for military products and all of them were at least two nodes older. Eventually, the separate military line was discarded, but that was for fab efficiency improvement, not because there was no need for a separate product list for unique applications.

The one thing that most people do not understand about tech nodes is that not all customers want the latest. That is why I chuckled at threads like this one. China reached X nm...!!! Every PDF Chinese starts wanking and posting inane comments about how China will take over the semicon industry. Lower node means smaller die structures means smaller dies means smaller packaging means more DRAM or NAND chips on the same PCB. Great achievement. Not going to dispute it. But not every sub-sector of the tech economy want it, let alone need it. Smaller die structures means decreased structural durability and longevity over time of electrical stresses. Conversely, larger die structures equals higher durability and longevity. So which sub-sector of the tech industry want the older nodes, maybe even a decade older? How about smart car? In automotive and aviation, or just transportation overall, an accident is usually catastrophic. Train derailment or a smart car run over a pedestrian. The manufacturers do not care about smaller packaging of chips. There is plenty of room in the car or airplane electronics compartments. The manufacturers want reliability and durability. So if all Chinese semicon companies reached 7nm node and move all their lines to that node, they WILL lose customers. The PC and console gamers and the cell phone companies would remain because their product lines need smaller packaging. But other customers whose products are proven reliable on older tech nodes, they will move to some other suppliers such as Micron who has a history of keeping old products always available for startup.
Most Chinese fab revenue is from legacy nodes though. in particular there's large scale production for automotive, analog and power components. BYD has an in house fab for automotive and power chips, and there's also China Resource Micro. Even SMIC has most of their revenue from older nodes despite having the capability for newer nodes.
 
.

Latest posts

Back
Top Bottom