TSMC N1 Node Chip Plant Said to be Under Planning

[TheLostSwede]

Based on news out of Taiwan, TSMC is said to be in the early planning stages of yet another chip plant, this time for its first N1 node. The new plant will reportedly be built in a science park in Taoyuan, less than an hour south west of Taipei, according to the Commercial Times. TSMC already has a pair of chip packaging and testing facilities in the science park, making it a suitable location for a chip plant. This will be TSMC's most northern chip manufacturing plant in Taiwan, although it's not expected to start pilot production until sometime in 2027. TSMC hasn't confirmed any of the details, but the company didn't outright deny the report either.

Despite the potential global downturn in the economy, TSMC appears to be fully committed to continue to build new fabs for increasingly smaller nodes. The company is set to start its first commercial production on its N3 node this quarter and is expecting the N3 node to contribute as much as four to six percent of its overall revenue in 2023. Its N2 node should enter commercial production in 2025, but not much is known about the state of the N2 node at this point in time. The N1 node might end up being a 1.4 nm node, based on TSMC's measurements, but the company is still in the very beginning of the R&D phase for this node.



View at TechPowerUp Main Site | Source

 

[Daven]

Keep shrinking those nodes until you reach the quantum realm!

 

[AusWolf]

Next: "TSMC is planning its next plant on the node 0.0000025x the size of an electron to open by 2049"

 

[THU31]

Are they not supposed to switch to graphene or something by the next decade? Silicon is reaching the limit of further shrinking possibilities, right?

 

[The red spirit]

Next: "TSMC is planning its next plant on the node 0.0000025x the size of an electron to open by 2049"

The crazy thing is that TSMC actually could build something on angstrom scale if they really wanted too, but at that point various voltage and currency leaks become really crazy and you can end up with chip that isn't any more efficient than older node chip. Another crazy thing is that such phenomena started at around 180nm nodes and the main reason why nm numbers don't mean shit no more is because number can get lowered even if node isn't shrunken at all, but instead process improved in ways that it deals with leakages better, therefore becomes denser and therefore superior. And while that makes it look like we still get some scaling from miniaturization, such technique made each improvement balloon in engineering price. That's actually called Moore's second law (https://en.wikipedia.org/wiki/Moore's_second_law). Now we make each chip perhaps cheaper than ever in terms of materials, but more expensive than ever in terms of research required. We have been playing with those limitations for two decades and it looks like we may really face a wall in miniaturization soon, but this time it won't be diminishing return wall, but a real wall. That's why there is a lot of research done in non silicon materials. But problem isn't really silicon itself, but rather the way electricity works and what makes that a problem is that it's actually one of the fastest things that we have at all. Optical computing would have failed instantly, because light is much slower than electricity. There's a lot of research done about parts of atoms, particularly photons. If you think about all of this it's damn insane what we have right now on silicon and research done about future technology is even more insane. Oh and BTW not really silicon anymore, there are many other materials researched, tested and used right now for CPUs and chips in general. I think Germanium was the last such material, but it finally lost out to silicon due to even worse leakage problems and therefore bad scaling.

 

[Steevo]

The crazy thing is that TSMC actually could build something on angstrom scale if they really wanted too, but at that point various voltage and currency leaks become really crazy and you can end up with chip that isn't any more efficient than older node chip. Another crazy thing is that such phenomena started at around 180nm nodes and the main reason why nm numbers don't mean shit no more is because number can get lowered even if node isn't shrunken at all, but instead process improved in ways that it deals with leakages better, therefore becomes denser and therefore superior. And while that makes it look like we still get some scaling from miniaturization, such technique made each improvement balloon in engineering price. That's actually called Moore's second law (https://en.wikipedia.org/wiki/Moore's_second_law). Now we make each chip perhaps cheaper than ever in terms of materials, but more expensive than ever in terms of research required. We have been playing with those limitations for two decades and it looks like we may really face a wall in miniaturization soon, but this time it won't be diminishing return wall, but a real wall. That's why there is a lot of research done in non silicon materials. But problem isn't really silicon itself, but rather the way electricity works and what makes that a problem is that it's actually one of the fastest things that we have at all. Optical computing would have failed instantly, because light is much slower than electricity. There's a lot of research done about parts of atoms, particularly photons. If you think about all of this it's damn insane what we have right now on silicon and research done about future technology is even more insane. Oh and BTW not really silicon anymore, there are many other materials researched, tested and used right now for CPUs and chips in general. I think Germanium was the last such material, but it finally lost out to silicon due to even worse leakage problems and therefore bad scaling.

Photons propagate at the speed of electromagnetism, just like electrons (to be completely clear, not the electrons themselves, but the electromagnetic force). The issue with optical computing is we can't store light in a bottle yet, so it must be converted to electrical signals, and there we are, back at the same wall.

 

[Minus Infinity]

Photons propagate at the speed of electromagnetism, just like electrons (to be completely clear, not the electrons themselves, but the electromagnetic force). The issue with optical computing is we can't store light in a bottle yet, so it must be converted to electrical signals, and there we are, back at the same wall.

Um memo, we can indeed do all optical computing and have been able to do so for a long time albeit very primitive circuits. There is no covnersion to electrical signals, as the circuits use bandgap materials that are electrically insulating.

 

[Steevo]

Um memo, we can indeed do all optical computing and have been able to do so for a long time albeit very primitive circuits. There is no covnersion to electrical signals, as the circuits use bandgap materials that are electrically insulating.

Optical computing - Wikipedia

en.wikipedia.org

Can it play Crysis? Seriously, we ahve some parts of optical computers, but not a functioning useable system that is coherent and complete.

 

[The red spirit]

Um memo, we can indeed do all optical computing and have been able to do so for a long time albeit very primitive circuits. There is no covnersion to electrical signals, as the circuits use bandgap materials that are electrically insulating.

If it's super primitive and can't scale, then it's far away from any mainstream or even corporate adoption, basically just one of failed technologies.

 

[Steevo]

If it's super primitive and can't scale, then it's far away from any mainstream or even corporate adoption, basically just one of failed technologies.

I can see the appeal, once we have a whole plug in system it will be functional, not sure how it will scale to home use though and it’s not going to be the same as CPUs and GPUs now, it will be like buying a Server that is complete and you pay for the performance level desired.

I’m more interested in what 3D packaging will bring in the next few years, TSVs and the ability to use chiplets, more mature processes, look at what Intel is doing on the same node with refinements.

 

[The red spirit]

I can see the appeal, once we have a whole plug in system it will be functional, not sure how it will scale to home use though and it’s not going to be the same as CPUs and GPUs now, it will be like buying a Server that is complete and you pay for the performance level desired.

It's so primitive so far that it's useless for anything. It's too primitive even for trailing edge chips, which are already 30-50 year old lithographies.

 

[darakian]

N0 coming up next /s

 

[Minus Infinity]

1.4nm based on measurements. Measurements of what exactly????? Point me to a feature, any feature that will be 1.4nm. Not even a simple line trace. Certainly not even any part of a GAAFET or whatever it's called.

 

[Wirko]

Photons propagate at the speed of electromagnetism, just like electrons (to be completely clear, not the electrons themselves, but the electromagnetic force). The issue with optical computing is we can't store light in a bottle yet, so it must be converted to electrical signals, and there we are, back at the same wall.

Photons also are not infinitely small. They are not small at all. To use a pure optical computer made on a fine node, one that requires an EUV source for lithography, you'd also need an EUV light source at home because DUV photons are simply too fat.

 

[Minus Infinity]

If it's super primitive and can't scale, then it's far away from any mainstream or even corporate adoption, basically just one of failed technologies.

It's still in it's infancy but there are many all optical compnents required to build a photonic computer already being made. But a long way to go to what one would call a computer.

Optical Computing: Status and Perspectives - PMC

For many years, optics has been employed in computing, although the major focus has been and remains to be on connecting parts of computers, for communications, or more fundamentally in systems that have some optical function or element (optical ...

www.ncbi.nlm.nih.gov

 

[The red spirit]

It's still in it's infancy but there are many all optical compnents required to build a photonic computer already being made. But a long way to go to what one would call a computer.

Optical Computing: Status and Perspectives - PMC

For many years, optics has been employed in computing, although the major focus has been and remains to be on connecting parts of computers, for communications, or more fundamentally in systems that have some optical function or element (optical ...

www.ncbi.nlm.nih.gov

But it has to be competitive in some market or niche, if it's slow or has too many drawbacks or just one fatal drawback, then they make no sense.

 

[Minus Infinity]

But it has to be competitive in some market or niche, if it's slow or has too many drawbacks or just one fatal drawback, then they make no sense.

Well it's probably less advanced than quantum computing. I used to work in photonics and I had real concerns about scaling. You cannot just use smaller wavelengths, because of materials we need for guiding having extreme losses as wavelengths get smaller. Now ay we could even use UV IMO. Telecommunications uses IR 1550nm or so for optical fibre as that the minimum loss in silicon glass. For guiding light and making photonic computers they have been using photonic bandgap materials, but they have a huge set of problems. Too much to get into.

 

[The red spirit]

Well it's probably less advanced than quantum computing. I used to work in photonics and I had real concerns about scaling. You cannot just use smaller wavelengths, because of materials we need for guiding having extreme losses as wavelengths get smaller. Now ay we could even use UV IMO. Telecommunications uses IR 1550nm or so for optical fibre as that the minimum loss in silicon glass. For guiding light and making photonic computers they have been using photonic bandgap materials, but they have a huge set of problems. Too much to get into.

BTW isn't speed of light slower than of electrons?

 

[Wirko]

BTW isn't speed of light slower than of electrons?

Can be, in very specific circumstances (look up Cherenkov radiation, it occurs in water around nuclear reactors I think).

The ordered movement of electrons in wires is very, very, VERY slow. Bamboo grows faster, up to 10 cm per day. Random thermal movement of the same electrons is very, very fast but this doesn't matter in this discussion.

Are they not supposed to switch to graphene or something by the next decade? Silicon is reaching the limit of further shrinking possibilities, right?

Unfortunately, scientists are only able to tell about as much as you did. Graphene or something.

Better materials may solve some problems but the limits of manufacturing remain the same. ASML scanners still won't draw lines thinner than EUV wavelength approximately, which is 13.5 nm. Building stacked (CFET) logic transistors is still just a vague idea, a few years out at best. A processed wafer will still cost 20,000 dollars. Designing a chip will be no cheaper.

 

[Minus Infinity]

BTW isn't speed of light slower than of electrons?

In what? Not in vacuum. Also electrons don't have an inherent speed, they have to be accelerated, photons alwasy move at c in vacuum and c/n in materials where n is refractive index. Electrons can never have a speed of c since they have mass. Now of course in semiconductors and photonic bandgap materials thinks are far more complicated. Electrons aren't even localised and can't be called point masses and photons can move much slower than c/n. We have even managed to make photons stop.

 

[Wirko]

In what? Not in vacuum. Also electrons don't have an inherent speed, they have to be accelerated, photons alwasy move at c in vacuum and c/n in materials where n is refractive index. Electrons can never have a speed of c since they have mass. Now of course in semiconductors and photonic bandgap materials thinks are far more complicated. Electrons aren't even localised and can't be called point masses and photons can move much slower than c/n. We have even managed to make photons stop.

Photons aren't localised, either. I know nothing about photonic bandgap materials - can photons be controlled by using structures smaller, even much smaller, than their wavelength?