TSMC working towards a future with trillion-transistor chips, 1nm-class manufacturing

[emorphy]

The big picture: The global company showed how it plans to continue increasing transistor density over the next several years. It also believes that the semiconductor industry will transition to chiplet-based designs as well.

At the recent IEDM conference, TSMC unveiled a product roadmap for its semiconductors and next-generation production nodes that culminates in eventually delivering multiple 3D-stacked collections of chiplet designs (3D Hetero Integration) with one trillion transistors on a single chip package. Advancements in packaging technologies, such as CoWoS, InFO and SoIC, will allow it to reach that goal and by 2030 it believes that its monolithic designs could reach 200 billion transistors.

Nvidia's 80-billion-transistor GH100 is one of the most sophisticated monolithic processors currently on the market. However, as the size of these processors continues to grow and become more costly, TSMC believes that manufacturers will adopt multi-chiplet architectures, such as AMD's recently-launched Instinct MI300X and Intel's Ponte Vecchio, which has 100 billion transistors.

For now, TSMC will continue to develop 2nm-class N2 and N2P production nodes and 1.4nm-class A14 and 1nm-class A10 fabrication processes. The company expects to start 2nm production by the end of 2025. In 2028, it will move onto a 1.4nm A14 process, and by 2030, it expects to be producing 1nm transistors.

TSMC presentation slide

Meanwhile, Intel is working on its 2nm process (20A), and 1.8nm (18A), which it roughly expects to launch during the same timeframe. One advantage of the new technology is backside power delivery, called PowerVia, which should allow for higher logic densities, higher boost clock speeds, and lower power leakage, resulting in more energy-efficient designs that could outperform TSMC's offerings.

As the world's largest foundry, TSMC is confident its process nodes will outperform anything from Intel. In an earnings call, TSMC CEO C.C. Wei said that an internal assessment confirmed the enhancements of the N3P technology and that its 3nm-class manufacturing node demonstrated "comparable PPA" to Intel's 18A node. He expects N3P to be even better, boasting a more competitive "technology maturity" and having significant cost advantages.

Not one to ignore fighting words, Intel CEO Pat Gelsinger claimed that its 18A process node will outperform TSMC's 2nm chips despite launching a year earlier.

The market will soon be able to determine which is better. The Taiwan-based chip giant expects to bring N3P into mass production in the second half of 2024 alongside its 20A and 18A products.

Permalink to story.

https://www.techspot.com/news/101364-tsmc-working-towards-future-trillion-transistor-chips-1nm.html

 

[noel24]

Good for Them, unfortunately This is one of Those "corporation level - professional use" products, as regular people will be stuck with sub 50 billion, multichip GPUs.

 

[Kashim]

Intel's 18A process is snake oil. It's probably going to be closer to 4nm or maybe 3nm if they get really lucky. Intel is years behind TSMC and no amount of boasting is going to make that gap disappear in just a year or two. This isn't the type of problem that can be solved by throwing more money and more talent at it. It's like The Mythical Man-Month; a woman can make a baby in nine months, but nine women can't make a baby in one month.

Intel was stuck on 14nm for like 5 years, but we're expected to believe that suddenly they're going to leapfrog the industry leader like it's nothing. Please. Intel's new strategy is to simply beat the competition using deceptive naming schemes. They call their current "state of the art" process Intel 4 when in fact it's closer to 7nm.

In the end, the proof will be in the pudding. I can't wait to see Intel's excuses and updated roadmap when things don't pan out the way they expect. I'm sure they'll pay off a lot of these review sites to celebrate modest gains as if they were some kind of miracle.

 

[WhoCaresAnymore]

There just has to be alien life, how we advanced so fast from the 70's is beyond me. Tech is great, but it's making mankind lazy. Less physical labor and moving around = increased waistband and lower life expectancy.

 

[Kashim]

There just has to be alien life, how we advanced so fast from the 70's is beyond me. Tech is great, but it's making mankind lazy. Less physical labor and moving around = increased waistband and lower life expectancy.

Your theory is nonsense. People aren't fatter now because they have to do less physical labor than 50 years ago. The two are completely unrelated. There are plenty of people now who do physical labor all day and yet they're fat anyway because they shovel fast food and other garbage into their bodies every chance they get.

People were generally thinner in 70's, that much is true. The reason people were thinner back than is because we had far less processed foods and this wonderful thing we call capitalism hasn't taken off yet. Now everything is about consumerism and shoving as much useless garbage down people's throats as possible all in the name of making a bigger profit than the competition. We now get bombarded with fast food, TV dinners, ultra-processed foods, GMO garbage, etc. Next time you're shopping try and find something that doesn't have a bunch of chemicals in it and has less than 10 ingredients. Good luck.

 

[toooooot]

People were generally thinner in 70's, that much is true. The reason people were thinner back than is because we had far less processed foods and this wonderful thing we call capitalism hasn't taken off yet. Now everything is about consumerism and shoving as much useless garbage down people's throats as possible all in the name of making a bigger profit than the competition. We now get bombarded with fast food, TV dinners, ultra-processed foods, GMO garbage, etc. Next time you're shopping try and find something that doesn't have a bunch of chemicals in it and has less than 10 ingredients. Good luck.

I am usually against government dictating everything, but at this point I think they should do something. The rate at which obesity grows and even more horrifying the sad shape of these people who are literally killing their once healthy bodies .
Them doing it seems very wrong, just like using meth.
Maybe a good way to start would be slowly limiting and then removing most unhealthy foods.

 

[Zsoltblabla]

Your theory is nonsense. People aren't fatter now because they have to do less physical labor than 50 years ago. The two are completely unrelated. There are plenty of people now who do physical labor all day and yet they're fat anyway because they shovel fast food and other garbage into their bodies every chance they get.

People were generally thinner in 70's, that much is true. The reason people were thinner back than is because we had far less processed foods and this wonderful thing we call capitalism hasn't taken off yet. Now everything is about consumerism and shoving as much useless garbage down people's throats as possible all in the name of making a bigger profit than the competition. We now get bombarded with fast food, TV dinners, ultra-processed foods, GMO garbage, etc. Next time you're shopping try and find something that doesn't have a bunch of chemicals in it and has less than 10 ingredients. Good luck.

Exactly.
Also, thanks to all that you mentioned - cancer rates are through the roof.

 

[Zsoltblabla]

I am usually against government dictating everything, but at this point I think they should do something. The rate at which obesity grows and even more horrifying the sad shape of these people who are literally killing their once healthy bodies .
Them doing it seems very wrong, just like using meth.
Maybe a good way to start would be slowly limiting and then removing most unhealthy foods.

That wont happen anytime soon….
The food industry makes huge profits of the processed garbage they sell. Then, the healthcare industry make huge profits of “curing” us… this is all carefully designed snd executed. All in the name of ever greater profits.

 

[mailpup]

Back on topic please. Thank you.

 

[Hupster]

Intel's 18A process is snake oil. It's probably going to be closer to 4nm or maybe 3nm if they get really lucky. Intel is years behind TSMC and no amount of boasting is going to make that gap disappear in just a year or two. This isn't the type of problem that can be solved by throwing more money and more talent at it. It's like The Mythical Man-Month; a woman can make a baby in nine months, but nine women can't make a baby in one month.

Intel was stuck on 14nm for like 5 years, but we're expected to believe that suddenly they're going to leapfrog the industry leader like it's nothing. Please. Intel's new strategy is to simply beat the competition using deceptive naming schemes. They call their current "state of the art" process Intel 4 when in fact it's closer to 7nm.

In the end, the proof will be in the pudding. I can't wait to see Intel's excuses and updated roadmap when things don't pan out the way they expect. I'm sure they'll pay off a lot of these review sites to celebrate modest gains as if they were some kind of miracle.

They bought 6 of the 10 high na euv machines from asml for 2nm through to 1nm development. So they are at least showing some serious intent