> logic technology can extend for the first time below the 1 nm node, advancing the era of angstrom-level scaling, where dimensions approach the size of individual atoms. While transistor nodes now refer to a generation of manufacturing technology versus an exact physical dimension, IBM’s 0.7 nm technology—also referred to as 7 angstroms—demonstrates how continued scaling remains possible.
Continuing the well established trend of making bold claims about physical dimensions that have nothing to do with any of the structures in the chip, and the name scales better than the tech.
What they actually deliver is a "nanostack architecture" built with ~5nm features that according to them is comparable to a hypothetical real sub-1nm chip.
It's an impressive achievement nonetheless but it looks like the industry has a few too many marketers.
As it can be seen from the photos, horizontally the features are much bigger than 5 nm.
For silicon, the gate length of a FET has a lower limit somewhere between 10 nm and 15 nm.
The current CMOS manufacturing processes have not reached the limit yet. For making smaller transistors, a transition to other semiconductor materials will be necessary.
The vertical thicknesses of various layers may be of only a few nanometers or even of a fraction of a nanometer, but that does not matter directly for the circuit density.
The supposed node size refers to horizontal dimensions, not to vertical dimensions.
Vertical dimensions of around 1 nanometer or less could be achieved already many decades ago, because they depend on growth speed and on time, not on lithography, like the horizontal dimensions.
The industry should have stopped decades ago to talk about the "size" but they should have characterized a CMOS process by its density, e.g. in logic gates per square mm.
However, an actual concrete number would be disliked by marketing, because they could no longer claim that their "1 nm" process is better than the "2 nm" process of another vendor, if their density is not really better.
The scale bar on the far right "photo" (micrograph?) doesn't make sense. It is only slightly less than half the scale bar on the middle photo (10 nm), but the image is clearly scaled up by much more than 2x. Individual silicon atoms are circled in the right photo, but the covalent radius of silicon is about 0.11 nm, so they should be much smaller if the scale bar is accurate.
Unlike marketing terms, "nm density" is actually useful measure.
It describes density measure where you can compare it to planar transistors from the 28-nanometer (28 nm) node around 2010 to 2011 and before. A "0.7 nm" node has equivalent transistor density as if we could have shrunk standard flat transistor node down to 0.7 nanometers.
It's a physical quantity per some unit of spatial measurement so the units still don't match up b/c in one case the transistors are stacked per volume & in the other case per area.
> Historically, "node" sizes (like 28nm or 7nm) directly correlated to the physical length of a transistor's gate. Today, names like 3nm or 2nm reflect a marketing generation. The actual transistors are significantly larger than these nanometer labels, meaning density varies between companies
> Research organizations like IEEE have proposed new metrics, such as transistors per cubic millimeter (MTr/mm^3), to accurately map future 3D scaling. However, commercial chip foundries resist this change because it would make it harder to calculate commercial yields and thermal density limits using standard industry formulas.
It's been decades since published node sizes had any connection to actual feature size. Sadly this is just how it works in the semiconductor industry now.
My read on it was that they are trying to imply a transistor density (in a 2D plane sense) that is comparable to a 1nm process? But they achieve that through stacking (3D, not 2D) since the features aren't actually anywhere near 1nm?
I really can't see where the 0.7nm is coming from. The white line looks like it's just an edge of a feature that is "15 rows of silicon atoms", which by some quick arithmetic on Wolfram Alpha has to be AT LEAST ~1.6nm, and the way the rows of atoms appear to be packed in that image and by the provided scale, it seems to be significantly more. Using the white line as a meaningful measurement seems to me to be more misleading than any other interpretation here.
A 0.7 nm planar transistor made of silicon has no performance, because a device so small cannot function as a transistor.
The intended meaning of "0.7 nm" is that if you compare the transistor density per area of a "0.7 nm" manufacturing process with that of a "350 nm" process (like used for some Pentium II CPUs, at a time when "350 nm" was a real length), the ratio between the transistor densities is (350 nm / 0.7 nm)^2 = 500^2 = 250,000.
Comparing with the number of transistors of a Pentium II, a 0.7 nm CPU should be able to contain about 5000 billion transistors. This is consistent with the fact that the latest 3 nm NVIDIA Rubin GPU has 336 billion transistors and a 0.7 nm circuit must have a density around 16 times greater than a 3 nm circuit.
However, for many of the modern node names used by some companies even this computation is not really true, because marketing may have chosen an arbitrary name that is smaller than for the last process of the main competitor.
For now, IBM has not provided any kind of information that could prove their claim that their new CMOS process has the transistor density corresponding to "0.7 nm" (i.e. 16 times greater than the TSMC "3 nm" CMOS process).
Better metrics are transistors/mm^2, performance/watt, and raw performance, since at this point "nm" is fluff and easily game-able.
Different companies measure it differently too. This was a while ago, but I remember reading that Intel 10nm was more or less close to TSMC 7nm. I'm sure this is still true to varying degrees.
> Continuing the well established trend of making bold claims about physical dimensions that have nothing to do with any of the structures in the chip, and the name scales better than the tech.
We care about PPA (power, performance, area) and not how large or not-large features actually are. Comparing gate lengths between a 1980s planar transistor and a 2010s 3D FinFET or GAA transistor is obviously nonsense, the relatively aligned node names of the industry actually do make sense as a shortcut here.
Just to be clear, this doesn't mean that anything on the die actually measures 0.7nm — it means that it's roughly double the density as the previous node generation. At some point the industry decided to keep talking about "nanometers" even though the actual transistor sizes have been decoupled from the node name for years.
Two big problems 1) NOBODY knows what IBM's definition of "sub 1nm" means 2) IBM bullshits so much more than anyone including Intel (remember the "teleportation" ads years ago) that nobody is going to waste time researching what they mean in reality
That's not what I meant. IBM published some scientific papers which explain the details. If you don't have the background to understand the papers and you don't trust the marketing that still doesn't justify rhetorically flipping the table.
Most of their fabs were divested to GlobalFoundries, but they still have pretty significant fab capability and capacity- I suspect at least partly to have a us-based chip-making for military ("Trusted Foundry").
The labs might not be that different from consulting, the NYT reporting on this notes they run R&D labs so they can license the tech they develop to people who actually make chips.
IBM has been the company with the most patent registrations in the US for I think 29 of the last 30 years. They're one of the largest industrial research organizations in the world. They're doing more hard science research than almost anyone else.
Which is so weird, right? Like what is IBM now and how does a research lab make sense with the rest of their business?
The money-making parts of IBM are: legacy software and hardware (declining), consulting (low margin, low leverage), enterprise software (mostly redhat, not really growing). It's hard to explain how IBM research is accretive to any of that.
Licensing is a substantial source of revenue, and their servers have very impressive (think Telum’s caching) innovations, even though they rely on third-parties for manufacturing the chips themselves.
They are also betting on quantum computing to become commercially relevant.
I don't know enough about their business to say, but I'm thrilled at even the idea that someone might actually value long-term success over quarterly earnings.
It's just a shame that none of it seems to pan out, and in the areas where I know what they're talking about, it all sounds like cynical nonsense to me.
Broadly speaking yes, this is the business model. IBM has been at this for many years with technology transfers, licensing agreements, support and other arrangements. Rapidus, Samsung, GlobalFoundries, ST, SMIC, AMD, etc. have all used IBM R&D work at various times for various nodes and products.
The cutting edge of semiconductors is a writhing mass of copulating tapeworms, and IBM lives deep inside that ball. For IBM, what this means is that when you buy one of the ASML machines to make products with this process, you'll pay IBM for the knowledge and support to actually get it working, or give them a cut, or something else, TBD, as circumstances warrant.
I’m sure they will license it. It’s better for them if everyone in the industry can innovate on everything around it. All the process tech companies will make it more cost effective, for instance, which helps IBM as well.
I always feel like I'm not quite getting quantum stuff no matter how much I read and learn: what does this advancement have to do with quantum computers?
Don't worry about not grokking quantum computing stuff, neither do any of the people who invest in it as well as many people who work on it.
1. The OP has nothing to do with quantum computers.
2. Quantum computing deals in coherent quantum states: associated with N qubits there are 2^N complex amplitudes. You can measure by sampling the square-magnitude of the complex amplitude which turns it into a Probability Distribution. Quantum computing "gates" cause interference in the complex amplitude of entangled qubits cancelling out incorrect results, such that if you maintain coherence for long enough and sample the final state and measure the probability distribution, you get a computationally useful result. The key challenge in quantum computing is extending the coherence time of a larger and larger number of qubits, which is why you hear so much about quantum error correction. Recent results from Google showed a scaling law for "surface codes" using multiple qubits to create an error-corrected topological qubit with extended lifetime. There is no telling how far this scaling law will go, but as long as Gil Kalai is in the next room, it is unlikely there will be actual useful quantum computation for a while.
Approximately everyone (at least in the F500) outside of Big Tech uses them. For example, Costco's entire inventory management system runs on IBM i (so, POWER). You can see the classic terminal look around the store. Banks run a TON of z and i. You'll never see them because they're essentially always in data centers, but I guarantee you interact with them even if it's very non-obvious because there's 50 microservices between the UI and the actual system of record.
Yes, and we're already there. We've been there for quite a while, in fact.
Once you make the gate of a transistor small/thin enough, quantum effects take over. Electrons will randomly teleport into and through the gate causing the transistor to conduct when it shouldn't. I don't have numbers to hand, but it's on the order of a few atoms wide. There's really nothing that can be done about it either, as far as we know. Electrons just aren't physical objects at this scale, you can't simply exclude them from any given volume of space. The electron wave function will simply just appear wherever it wants (within the electron probability cloud). The only way to stop it is to make your insulating junction thicker than the probability cloud.
The experiment to observe this behavior is pretty simple though (Young's double slit), and it was conducted more than 200 years ago. The explanation came much later but it's not like the phenomenon was hiding somewhere.
It’s both ridiculous and quite amazing really. The hint that there is something less random underneath it that we just haven’t figured out (and lack the resources to explore at this time) is tantalising.
Even if there isn’t, the way it seems all based on the uneven flow of state over spacetime is deeply fascinating for someone who studies computing.
> you can't simply exclude them from any given volume of space...
... inside a silicon crystal.
You can keep the electrons into as small a volume as you want, but you need something there forcing them, and doped silicon will only force them so much.
In fact, those transistors are smaller than what a silicon crystal can do, and the electrons are only held there because they are made of more materials than only silicon.
I mean, you can't get smaller than an atom, there is some amount of plausibility of using individual atoms as at least the occasional computing element.
Beyond that, engineering a quark-gluon plasma as a processor? I'd watch that Star Trek episode. (we might fantasize about stuff like that but we're roughly monkeys smashing rocks together in a cave vs. building an iPhone sort of gap away from that kind of thing unless somebody has a really good idea)
You could, in principle, use photons and/or electrons. We got pretty damn close in the vacuum tube era, and photonic computing has been a popular research topic for a while.
You also have quantum computing, which I think can/does use subatomic particles? Not sure about that one
It depends on the type of quantum computer. In some a physical qubit is a single atom, but then to make it reliable they need to add error correction resulting in logical qubits consisting of at least 100 or so physical qubits.
Another type of quantum computer uses qubits consisting of "quantum circuits" which are actually huge macroscopic constructions (> 1mm).
You can’t make smaller chips features with photonics. Visible light photons have a wavelength between 400 and 800 nm, much larger than current chip features. When you go to higher frequencies they get smaller, but they are really difficult to produce and control.
> You could, in principle, use photons and/or electrons. We got pretty damn close in the vacuum tube era, and photonic computing has been a popular research topic for a while
Wait, what? How does this work in principle for storage? You can store electrons but you're saying you can store photons too?
We do use electrons. That's what flows through transistors to do computations. Or, vaguely, the distribution of the electric field....
>We got pretty damn close in the vacuum tube era
Uh, what?
There's only so many fundamental interactions in the Universe. Computing requires you to be able to distinguish two states and our current methodology is built around some sort of black box three input machine that can output either state, a switch.
That switch is the part that cannot be scaled down infinitely. The reality we are familiar with doesn't exist at atomic scales. "Things" don't even have properly defined boundaries at a certain level, and thermal noise is a huge issue.
IMO a much more direct limiter of our current computing capability is lack of manufacturing ability, and heat. We were lucky that transistors were so amenable to lithography as a concept, that they work so well in 2D and as a surface feature, as that is what drove our advances the past 100 years and enabled computing to be such a normal thing. The combination of a "Solid state" effect, the electric force having very convenient properties, and lithography being so amenable to scaling things in various directions is how we got here.
But lithography doesn't scale into 3D. We've been hacking around that by doing more layers but that scales awfully, has very strict limitations, and makes the heat problem infinitely worse, to the point of making it impossible to work around.
If we could assemble things atom by atom exactly how we want, we could vastly improve our theory and practice, and build really intricate processor chunks with effective cooling channels or something, and computing would scale so much more. Maybe. Maybe some other problem would suddenly start dominating in that world.
Biology literally is nanotechnology, but it takes massive tradeoffs in exchange. It might never be possible to manufacture, at scale, stuff atom by atom. The Universe doesn't promise us infinite progress in technology. Quite the opposite.
> IBM and its partners conduct this work at a leading semiconductor research facility in Albany, New York, which will soon be home to a High Numerical Aperture Extreme Ultraviolet (High NA EUV) lithography tool, essential for the future of logic scaling. Developed by ASML, this technology enables ultra‑precise circuit printing, supporting the creation of smaller, more powerful chips.
I'm guessing that this is the technology that is developed by Cymer (ASML subsidiary) in California, correct? Is there competing technology? I know xLight is trying to make some inroads on their own version of this EUV tech. I have not heard about any progress though.
The industry does use a collection of more practical measurements, like transistor density. Marketing pieces for the news tend to use this kind of jargon precisely because it can be fudged & it sounds like it means something else than it really does to the average person. It's also simple enough to avoid needing to really explain what kinds of numbers are impressive etc, everyone just knows less than 1 nm is tiny and they've heard X nm for decades to compare to at this point.
I wonder at what point does it become deceptive in the legally biting way to market process nodes like this. One can wax and wane a bunch about industry terminology status quo, but this is mental.
For anyone who needs it, a friendly reminder that CPU nm marketing is a complete fabrication and the physical size of transistors has zero relation to the marketing claims. These are not, in fact, physically sub 1 nm, despite the bombastic claims.
At some point in the transistor scaling, the electrons started leaking across the gate, we've switched from 2D design to 3D structures to prevent that, so the actual physical gate pitch for like the TSMC 3nm is around 45 nm in distance.
Currently thrown around numbers mean the "equivalent performance/density" or something like that.
They don't describe the exact physical size (that would rather defeat the point of the marketing), but you can see the photographs at the bottom have a scale measured in tens of nm.
IBM regularly announces silicon breakthroughs like this but I'm not aware of those ever becoming products. Is IBM mainly in the business of licensing their technology to big silicon manufacturers with stuff like this? Is it just marketing for their consulting business?
IBM Z series mainframe Telum CPUs are designed by IBM but manufactured by Samsung. IBM no longer owns any fabs. I assume they have some kind of technology licensing deal.
It's a lab. It's where ASML brings up the prototype machine and gets it working, with IBM talent working out the problems and getting it ready for commercial operation. They won't make chips at scale there: the facility isn't designed for that part. The thing to understand here is that isn't a simple, clean, comprehensible business arrangement. The Albany facility is highly subsidized by the state. IBM has their hooks deep in the operation and occupation of the site. Such facilities are extraordinary with capabilities that talent that are unique and fabulously expensive. That's why ASML is there, and not just doing it in some village in the Netherlands. It's why when Obama, Biden, Trump or whomever tells ASML to whom they will and won't be selling hardware, ASML listens.
> It's why when Obama, Biden, Trump or whomever tells ASML to whom they will and won't be selling hardware, ASML listens.
My understanding is that ASML's acquisition of Cymer in California (the actual EUV light source technology) in 2014 was only permitted under a strict technology sharing and export agreement with the US government. And that the technology development and production had to remain within the US.
The USA CHIPS Act and NY State have provided $100 billion+ in funding with the expectation that ASML's core R&D and "prototyping" like this will be done in the US in partnership with US companies (like IBM).
A little bit of a nitpick, but wouldn't that be a picometer instead of angstrom node? Like, isn't a "pico-" the next magnitude smaller than "nano-", or am i wrong?
Otherwise, that chip tech sounds really awesome - at least for the future!
Useless fact I just learned from Wikipedia: Ångström/Angstrom (in Sweden of course we still use the original spelling) has its own UNICODE symbol, Angstrom sign: Å (U+212B) not to confuse with the Swedish letter Å (U+00C5). Looks slightly different in my browser.
Looks like that's deprecated. From the next sentence:
However, version 5 of the standard already deprecates that code point and has it normalized into the code for the Swedish letter U+00C5 Å `latin capital letter a with ring above`
You had the right idea. Angstroms are not an SI unit. The SI units jump by three orders of magnitude at this scale: picometer, nanometer, micrometer, millimeter.
(In the same way that meter jumps three orders of magnitude to kilometer[1], or millions to billions to trillions, etc.)
[1] Technically there are intermediate SI units between meter and km but nobody uses them. There are not intermediate SI units between the tiny ones.
Everyday necessity. The gap between mm and m is too large, there are many things in daily life that are better expressed in cm. SI units must strike a balance between three factors: not having so many denominations nobody can remember them; not having so few denominations that using them adds too much wordiness to daily life (150mm or 0.15m are wordier than 15cm); and a degree of familiarity with the everyday units people used before metric, to smooth the transition and encourage adoption.
> logic technology can extend for the first time below the 1 nm node, advancing the era of angstrom-level scaling, where dimensions approach the size of individual atoms. While transistor nodes now refer to a generation of manufacturing technology versus an exact physical dimension, IBM’s 0.7 nm technology—also referred to as 7 angstroms—demonstrates how continued scaling remains possible.
Continuing the well established trend of making bold claims about physical dimensions that have nothing to do with any of the structures in the chip, and the name scales better than the tech.
What they actually deliver is a "nanostack architecture" built with ~5nm features that according to them is comparable to a hypothetical real sub-1nm chip.
It's an impressive achievement nonetheless but it looks like the industry has a few too many marketers.
As it can be seen from the photos, horizontally the features are much bigger than 5 nm.
For silicon, the gate length of a FET has a lower limit somewhere between 10 nm and 15 nm.
The current CMOS manufacturing processes have not reached the limit yet. For making smaller transistors, a transition to other semiconductor materials will be necessary.
The vertical thicknesses of various layers may be of only a few nanometers or even of a fraction of a nanometer, but that does not matter directly for the circuit density.
The supposed node size refers to horizontal dimensions, not to vertical dimensions.
Vertical dimensions of around 1 nanometer or less could be achieved already many decades ago, because they depend on growth speed and on time, not on lithography, like the horizontal dimensions.
The industry should have stopped decades ago to talk about the "size" but they should have characterized a CMOS process by its density, e.g. in logic gates per square mm.
However, an actual concrete number would be disliked by marketing, because they could no longer claim that their "1 nm" process is better than the "2 nm" process of another vendor, if their density is not really better.
Just get better marketers to say your 2nm process has more gates per sqmm than your competition 1nm process.
Exactly. WatsonX AI quantum angstroms for e-business.
At some point, people would not know if they have the deal or not have the deal in this new era of quantum e-business ...
Now with Tivoli Monitoring!
The scale bar on the far right "photo" (micrograph?) doesn't make sense. It is only slightly less than half the scale bar on the middle photo (10 nm), but the image is clearly scaled up by much more than 2x. Individual silicon atoms are circled in the right photo, but the covalent radius of silicon is about 0.11 nm, so they should be much smaller if the scale bar is accurate.
Unlike marketing terms, "nm density" is actually useful measure.
It describes density measure where you can compare it to planar transistors from the 28-nanometer (28 nm) node around 2010 to 2011 and before. A "0.7 nm" node has equivalent transistor density as if we could have shrunk standard flat transistor node down to 0.7 nanometers.
Yeah, the actual sizes are right there in the pictures, and never < 1 nm.
Why not use something absolute, like nand-gates per volume?
I am not a chip designer, doesn't area matter way more than volume? Vertical space is basically free; it's horizontal space that is at a huge premium.
Density is mass per volume so how are you comparing it to a planar transistor? Your units don't even match.
density is quantity per unit measure.
mass per volume is one example.
Not all densities is mass per volume. eg. population density.
It's a physical quantity per some unit of spatial measurement so the units still don't match up b/c in one case the transistors are stacked per volume & in the other case per area.
> Historically, "node" sizes (like 28nm or 7nm) directly correlated to the physical length of a transistor's gate. Today, names like 3nm or 2nm reflect a marketing generation. The actual transistors are significantly larger than these nanometer labels, meaning density varies between companies
> Research organizations like IEEE have proposed new metrics, such as transistors per cubic millimeter (MTr/mm^3), to accurately map future 3D scaling. However, commercial chip foundries resist this change because it would make it harder to calculate commercial yields and thermal density limits using standard industry formulas.
https://share.google/aimode/Z5BqUjlZWFNphm6Z6
https://en.wikipedia.org/wiki/Transistor_count#Transistor_de...
Density is this reply.
It's been decades since published node sizes had any connection to actual feature size. Sadly this is just how it works in the semiconductor industry now.
Who started it?
My read on it was that they are trying to imply a transistor density (in a 2D plane sense) that is comparable to a 1nm process? But they achieve that through stacking (3D, not 2D) since the features aren't actually anywhere near 1nm?
If they're adding a dimension, the marketing should reflect that.
I know they won't go for an anything that makes as much sense as 5nm3, so I vote for "1nm hyper space"
“TeraThread”
yeah, where on the pictures is the 0.7nm feature? The linespacing is around 5nm. Is it the white line which is 0.7nm?
I really can't see where the 0.7nm is coming from. The white line looks like it's just an edge of a feature that is "15 rows of silicon atoms", which by some quick arithmetic on Wolfram Alpha has to be AT LEAST ~1.6nm, and the way the rows of atoms appear to be packed in that image and by the provided scale, it seems to be significantly more. Using the white line as a meaningful measurement seems to me to be more misleading than any other interpretation here.
It's the equivalent performance of a 0.7 nm planar transistor. It's not about the feature size.
A 0.7 nm planar transistor made of silicon has no performance, because a device so small cannot function as a transistor.
The intended meaning of "0.7 nm" is that if you compare the transistor density per area of a "0.7 nm" manufacturing process with that of a "350 nm" process (like used for some Pentium II CPUs, at a time when "350 nm" was a real length), the ratio between the transistor densities is (350 nm / 0.7 nm)^2 = 500^2 = 250,000.
Comparing with the number of transistors of a Pentium II, a 0.7 nm CPU should be able to contain about 5000 billion transistors. This is consistent with the fact that the latest 3 nm NVIDIA Rubin GPU has 336 billion transistors and a 0.7 nm circuit must have a density around 16 times greater than a 3 nm circuit.
However, for many of the modern node names used by some companies even this computation is not really true, because marketing may have chosen an arbitrary name that is smaller than for the last process of the main competitor.
For now, IBM has not provided any kind of information that could prove their claim that their new CMOS process has the transistor density corresponding to "0.7 nm" (i.e. 16 times greater than the TSMC "3 nm" CMOS process).
Better metrics are transistors/mm^2, performance/watt, and raw performance, since at this point "nm" is fluff and easily game-able.
Different companies measure it differently too. This was a while ago, but I remember reading that Intel 10nm was more or less close to TSMC 7nm. I'm sure this is still true to varying degrees.
On the otherhand, no investor really cares what it's called, they just need to know it's next gen.
> Continuing the well established trend of making bold claims about physical dimensions that have nothing to do with any of the structures in the chip, and the name scales better than the tech.
We care about PPA (power, performance, area) and not how large or not-large features actually are. Comparing gate lengths between a 1980s planar transistor and a 2010s 3D FinFET or GAA transistor is obviously nonsense, the relatively aligned node names of the industry actually do make sense as a shortcut here.
You have to admit it's getting progressively sillier though.
Just to be clear, this doesn't mean that anything on the die actually measures 0.7nm — it means that it's roughly double the density as the previous node generation. At some point the industry decided to keep talking about "nanometers" even though the actual transistor sizes have been decoupled from the node name for years.
Two big problems 1) NOBODY knows what IBM's definition of "sub 1nm" means 2) IBM bullshits so much more than anyone including Intel (remember the "teleportation" ads years ago) that nobody is going to waste time researching what they mean in reality
> remember the "teleportation" ads years ago
Never heard of this, care to elaborate?
I know what it means. Something isn't automatically bullshit because it's outside your field of expertise.
“I’m able to parse a marketing term” isn’t a great claim here. Pointing out it’s an unclear term largely abused is valid.
That's not what I meant. IBM published some scientific papers which explain the details. If you don't have the background to understand the papers and you don't trust the marketing that still doesn't justify rhetorically flipping the table.
The most surprising part for me is that IBM still somehow owns silicon labs, I was sure it's effectively a consulting company by now
Most of their fabs were divested to GlobalFoundries, but they still have pretty significant fab capability and capacity- I suspect at least partly to have a us-based chip-making for military ("Trusted Foundry").
The labs might not be that different from consulting, the NYT reporting on this notes they run R&D labs so they can license the tech they develop to people who actually make chips.
IBM has been the company with the most patent registrations in the US for I think 29 of the last 30 years. They're one of the largest industrial research organizations in the world. They're doing more hard science research than almost anyone else.
Which is so weird, right? Like what is IBM now and how does a research lab make sense with the rest of their business?
The money-making parts of IBM are: legacy software and hardware (declining), consulting (low margin, low leverage), enterprise software (mostly redhat, not really growing). It's hard to explain how IBM research is accretive to any of that.
Licensing is a substantial source of revenue, and their servers have very impressive (think Telum’s caching) innovations, even though they rely on third-parties for manufacturing the chips themselves.
They are also betting on quantum computing to become commercially relevant.
I don't know enough about their business to say, but I'm thrilled at even the idea that someone might actually value long-term success over quarterly earnings.
It's just a shame that none of it seems to pan out, and in the areas where I know what they're talking about, it all sounds like cynical nonsense to me.
How does IBM commercialize this? Do they license this out to fabs?
> Do they license this out to fabs?
Broadly speaking yes, this is the business model. IBM has been at this for many years with technology transfers, licensing agreements, support and other arrangements. Rapidus, Samsung, GlobalFoundries, ST, SMIC, AMD, etc. have all used IBM R&D work at various times for various nodes and products.
The cutting edge of semiconductors is a writhing mass of copulating tapeworms, and IBM lives deep inside that ball. For IBM, what this means is that when you buy one of the ASML machines to make products with this process, you'll pay IBM for the knowledge and support to actually get it working, or give them a cut, or something else, TBD, as circumstances warrant.
They licensed 2 nm to Rapidus so yes.
I’m sure they will license it. It’s better for them if everyone in the industry can innovate on everything around it. All the process tech companies will make it more cost effective, for instance, which helps IBM as well.
boost sales for their systems division, POWER CPUs, mainframes, maybe Quantum stuff
I always feel like I'm not quite getting quantum stuff no matter how much I read and learn: what does this advancement have to do with quantum computers?
Don't worry about not grokking quantum computing stuff, neither do any of the people who invest in it as well as many people who work on it.
1. The OP has nothing to do with quantum computers.
2. Quantum computing deals in coherent quantum states: associated with N qubits there are 2^N complex amplitudes. You can measure by sampling the square-magnitude of the complex amplitude which turns it into a Probability Distribution. Quantum computing "gates" cause interference in the complex amplitude of entangled qubits cancelling out incorrect results, such that if you maintain coherence for long enough and sample the final state and measure the probability distribution, you get a computationally useful result. The key challenge in quantum computing is extending the coherence time of a larger and larger number of qubits, which is why you hear so much about quantum error correction. Recent results from Google showed a scaling law for "surface codes" using multiple qubits to create an error-corrected topological qubit with extended lifetime. There is no telling how far this scaling law will go, but as long as Gil Kalai is in the next room, it is unlikely there will be actual useful quantum computation for a while.
Sit on a patent and try to scrape earnings from others, maybe? That is, license or litigate.
Keep hearing that IBM makes these incredible chips but don’t see anyone using IBM chips. What do they do with them?
Approximately everyone (at least in the F500) outside of Big Tech uses them. For example, Costco's entire inventory management system runs on IBM i (so, POWER). You can see the classic terminal look around the store. Banks run a TON of z and i. You'll never see them because they're essentially always in data centers, but I guarantee you interact with them even if it's very non-obvious because there's 50 microservices between the UI and the actual system of record.
Their line of POWER chips are used in their mainframes
The product here is the research and licensing the tech.
usgov
One of the images has "15 rows of Si atoms".
Is there a limit to how small things can go? A single atom?
Is there a physical/molecular limit to Moore's Law?
Yes, and we're already there. We've been there for quite a while, in fact.
Once you make the gate of a transistor small/thin enough, quantum effects take over. Electrons will randomly teleport into and through the gate causing the transistor to conduct when it shouldn't. I don't have numbers to hand, but it's on the order of a few atoms wide. There's really nothing that can be done about it either, as far as we know. Electrons just aren't physical objects at this scale, you can't simply exclude them from any given volume of space. The electron wave function will simply just appear wherever it wants (within the electron probability cloud). The only way to stop it is to make your insulating junction thicker than the probability cloud.
>"The electron wave function will simply just appear wherever it wants (within the electron probability cloud)."
I don't know which is more ridiculous, the fact that reality works like this, or, that a species of apes was able to figure this out.
The experiment to observe this behavior is pretty simple though (Young's double slit), and it was conducted more than 200 years ago. The explanation came much later but it's not like the phenomenon was hiding somewhere.
It’s both ridiculous and quite amazing really. The hint that there is something less random underneath it that we just haven’t figured out (and lack the resources to explore at this time) is tantalising.
Even if there isn’t, the way it seems all based on the uneven flow of state over spacetime is deeply fascinating for someone who studies computing.
> you can't simply exclude them from any given volume of space...
... inside a silicon crystal.
You can keep the electrons into as small a volume as you want, but you need something there forcing them, and doped silicon will only force them so much.
In fact, those transistors are smaller than what a silicon crystal can do, and the electrons are only held there because they are made of more materials than only silicon.
https://en.wikipedia.org/wiki/There%27s_Plenty_of_Room_at_th...
https://en.wikipedia.org/wiki/Landauer%27s_principle
> https://en.wikipedia.org/wiki/There%27s_Plenty_of_Room_at_th...
Yes, single-atom manipulation has already been demonstrated:
* https://en.wikipedia.org/wiki/IBM_(atoms)
Can you make transistors using that technique? Can you smaller?
I mean, you can't get smaller than an atom, there is some amount of plausibility of using individual atoms as at least the occasional computing element.
Beyond that, engineering a quark-gluon plasma as a processor? I'd watch that Star Trek episode. (we might fantasize about stuff like that but we're roughly monkeys smashing rocks together in a cave vs. building an iPhone sort of gap away from that kind of thing unless somebody has a really good idea)
You could, in principle, use photons and/or electrons. We got pretty damn close in the vacuum tube era, and photonic computing has been a popular research topic for a while.
You also have quantum computing, which I think can/does use subatomic particles? Not sure about that one
It depends on the type of quantum computer. In some a physical qubit is a single atom, but then to make it reliable they need to add error correction resulting in logical qubits consisting of at least 100 or so physical qubits.
Another type of quantum computer uses qubits consisting of "quantum circuits" which are actually huge macroscopic constructions (> 1mm).
You can’t make smaller chips features with photonics. Visible light photons have a wavelength between 400 and 800 nm, much larger than current chip features. When you go to higher frequencies they get smaller, but they are really difficult to produce and control.
> You could, in principle, use photons and/or electrons. We got pretty damn close in the vacuum tube era, and photonic computing has been a popular research topic for a while
Wait, what? How does this work in principle for storage? You can store electrons but you're saying you can store photons too?
We do use electrons. That's what flows through transistors to do computations. Or, vaguely, the distribution of the electric field....
>We got pretty damn close in the vacuum tube era
Uh, what?
There's only so many fundamental interactions in the Universe. Computing requires you to be able to distinguish two states and our current methodology is built around some sort of black box three input machine that can output either state, a switch.
That switch is the part that cannot be scaled down infinitely. The reality we are familiar with doesn't exist at atomic scales. "Things" don't even have properly defined boundaries at a certain level, and thermal noise is a huge issue.
IMO a much more direct limiter of our current computing capability is lack of manufacturing ability, and heat. We were lucky that transistors were so amenable to lithography as a concept, that they work so well in 2D and as a surface feature, as that is what drove our advances the past 100 years and enabled computing to be such a normal thing. The combination of a "Solid state" effect, the electric force having very convenient properties, and lithography being so amenable to scaling things in various directions is how we got here.
But lithography doesn't scale into 3D. We've been hacking around that by doing more layers but that scales awfully, has very strict limitations, and makes the heat problem infinitely worse, to the point of making it impossible to work around.
If we could assemble things atom by atom exactly how we want, we could vastly improve our theory and practice, and build really intricate processor chunks with effective cooling channels or something, and computing would scale so much more. Maybe. Maybe some other problem would suddenly start dominating in that world.
Biology literally is nanotechnology, but it takes massive tradeoffs in exchange. It might never be possible to manufacture, at scale, stuff atom by atom. The Universe doesn't promise us infinite progress in technology. Quite the opposite.
> IBM and its partners conduct this work at a leading semiconductor research facility in Albany, New York, which will soon be home to a High Numerical Aperture Extreme Ultraviolet (High NA EUV) lithography tool, essential for the future of logic scaling. Developed by ASML, this technology enables ultra‑precise circuit printing, supporting the creation of smaller, more powerful chips.
I'm guessing that this is the technology that is developed by Cymer (ASML subsidiary) in California, correct? Is there competing technology? I know xLight is trying to make some inroads on their own version of this EUV tech. I have not heard about any progress though.
Cymer builds the EUV light source, but the biggest enabler for High NA EUV is using anamorphic optics (ie asymmetric horizontal and vertical magnification) from Zeiss: https://www.asml.com/en/news/stories/2024/5-things-high-na-e...
Correct
Why doesn't the industry use something like transistor density per cubic cm? This would extend to 3d cases and impossible to fake
The industry does use a collection of more practical measurements, like transistor density. Marketing pieces for the news tend to use this kind of jargon precisely because it can be fudged & it sounds like it means something else than it really does to the average person. It's also simple enough to avoid needing to really explain what kinds of numbers are impressive etc, everyone just knows less than 1 nm is tiny and they've heard X nm for decades to compare to at this point.
IBM debuts a PDK and some test structures, as they have no production fabs
They have small scale fabrication capabilities or they wouldn’t be able to validate the technology enough to sell it.
has anyone found a paper with details?
also, I was expecting to see cfets mentioned.
I wonder at what point does it become deceptive in the legally biting way to market process nodes like this. One can wax and wane a bunch about industry terminology status quo, but this is mental.
It'll be a brave man who takes on the IBM legal department over terminology in widespread use.
Since a transistor can't be smaller than a single atom, maybe it's time to start optimizing our software again.
For anyone who needs it, a friendly reminder that CPU nm marketing is a complete fabrication and the physical size of transistors has zero relation to the marketing claims. These are not, in fact, physically sub 1 nm, despite the bombastic claims.
>These are not, in fact, physically sub 1 nm, despite the bombastic claims.
Why? What's their real size?
Not doubting you, just trying to understand and also trying to assess how exaggerated the marketing is.
At some point in the transistor scaling, the electrons started leaking across the gate, we've switched from 2D design to 3D structures to prevent that, so the actual physical gate pitch for like the TSMC 3nm is around 45 nm in distance.
Currently thrown around numbers mean the "equivalent performance/density" or something like that.
They don't describe the exact physical size (that would rather defeat the point of the marketing), but you can see the photographs at the bottom have a scale measured in tens of nm.
The marketing nm better represent the density and performance of the transistors than the actual feature size, especially in this case.
So the title should be corrected. The did not debut sub nm chips at all.
That ship sailed long ago. I think it was around 32nm-22nm node when the marketing term started diverging from the physical feature size.
IBM regularly announces silicon breakthroughs like this but I'm not aware of those ever becoming products. Is IBM mainly in the business of licensing their technology to big silicon manufacturers with stuff like this? Is it just marketing for their consulting business?
My understanding is they are largely an IP business. That said this release mentioned an ASML machine on prem, so?
IBM's contributions to computing hardware and software are incalculable.
So many breakthroughs in hard drives, chips, transistor density, and other aspects of computing have come out of their labs.
Great to see them continuing to innovate.
But, yeah, usually they partner and license. Over the years, they've spun off more and more of their hardware businesses.
It's great that they found a working business model for a pure r&d lab, and with such awesome results.
I wonder why isn't this more common.
Don't forget copper interconnects for ICs. https://www.chiphistory.org/ibm-s-development-of-copper-inte...
I believe that IBM makes the chips for their Z Series mainframes. I mean, that's low volume production, but they need small feature size.
IBM Z series mainframe Telum CPUs are designed by IBM but manufactured by Samsung. IBM no longer owns any fabs. I assume they have some kind of technology licensing deal.
https://www.ibm.com/products/z/telum
> IBM no longer owns any fabs
Per IBM: "IBM Research at Albany [...] includes more than 100,000 square feet of semiconductor fabrication space"
I guess that is technically a R&D fab not a production one, but they definitely have in house fabrication capability
It's a lab. It's where ASML brings up the prototype machine and gets it working, with IBM talent working out the problems and getting it ready for commercial operation. They won't make chips at scale there: the facility isn't designed for that part. The thing to understand here is that isn't a simple, clean, comprehensible business arrangement. The Albany facility is highly subsidized by the state. IBM has their hooks deep in the operation and occupation of the site. Such facilities are extraordinary with capabilities that talent that are unique and fabulously expensive. That's why ASML is there, and not just doing it in some village in the Netherlands. It's why when Obama, Biden, Trump or whomever tells ASML to whom they will and won't be selling hardware, ASML listens.
> It's why when Obama, Biden, Trump or whomever tells ASML to whom they will and won't be selling hardware, ASML listens.
My understanding is that ASML's acquisition of Cymer in California (the actual EUV light source technology) in 2014 was only permitted under a strict technology sharing and export agreement with the US government. And that the technology development and production had to remain within the US.
The USA CHIPS Act and NY State have provided $100 billion+ in funding with the expectation that ASML's core R&D and "prototyping" like this will be done in the US in partnership with US companies (like IBM).
A little bit of a nitpick, but wouldn't that be a picometer instead of angstrom node? Like, isn't a "pico-" the next magnitude smaller than "nano-", or am i wrong?
Otherwise, that chip tech sounds really awesome - at least for the future!
There are 3 orders of magnitude between nano (^-9) and pico (^-12). An Angstrom is ^-10m.
Useless fact I just learned from Wikipedia: Ångström/Angstrom (in Sweden of course we still use the original spelling) has its own UNICODE symbol, Angstrom sign: Å (U+212B) not to confuse with the Swedish letter Å (U+00C5). Looks slightly different in my browser.
https://en.wikipedia.org/wiki/Angstrom
Looks like that's deprecated. From the next sentence:
However, version 5 of the standard already deprecates that code point and has it normalized into the code for the Swedish letter U+00C5 Å `latin capital letter a with ring above`
Aaahhh, ok, thanks!
You had the right idea. Angstroms are not an SI unit. The SI units jump by three orders of magnitude at this scale: picometer, nanometer, micrometer, millimeter.
(In the same way that meter jumps three orders of magnitude to kilometer[1], or millions to billions to trillions, etc.)
[1] Technically there are intermediate SI units between meter and km but nobody uses them. There are not intermediate SI units between the tiny ones.
Why above 1mm do we go by tens instead of thousands?
We have centimeter (10 mm) then decimeter (100mm) then meter (1000mm). Then we jump to thousand again (kilometer).
>We have centimeter (10 mm) then decimeter (100mm)
Does anyone actually use those? I think I would throw up a little in my mouth if I saw either of those on a mechanical drawing.
Centimeter is the commonly used metric for small distances in everyday parlance, just like an inch.
Answer that question and you'll get the whole impetus for logarithmic scales.
Everyday necessity. The gap between mm and m is too large, there are many things in daily life that are better expressed in cm. SI units must strike a balance between three factors: not having so many denominations nobody can remember them; not having so few denominations that using them adds too much wordiness to daily life (150mm or 0.15m are wordier than 15cm); and a degree of familiarity with the everyday units people used before metric, to smooth the transition and encourage adoption.
Because 1 angstrom equals 10⁻¹⁰ meters and 1 picometer equals 10⁻¹² meters, the relationship is:
1 Å = 100 pm. 1 pm = 0.01 Å.
1 picometer = 0.001 nanometers, 0.01 angstrom
1 angstrom = 0.1 nanometers, 100 picometers
1 nanometer = 10 angstroms, 1000 picometers