TSMC on Track to Begin Volume Production of 3nm Chips in 2022

[centauratlas]

The end of Slicon node reduction is near. Maybe 1.5 nm or so. What next?

That has been the refrain since at least the 1980s, and engineers (some who I went to grad school with) have found ways around it. Eventually, it will happen, but ramping up cores has been a way to increase performance, software to use them efficiently has been the bottlenec. (There was also the CISC vs RISC debate. )

 

[rp2011]

I’m sure they’ve thought about this a while back, will be interesting to see what happens when that time arrives.

it would be nice to believe. It’s already amazing we have got this small. We may start seeing problems with tunneling start creeping in at 3nm. I am not sure how they are going to mitigate that.

 

[rp2011]

What comes after 1nm?

As it stands right now, it seems with existing technology and known physics, it’s doubtful we will get past 3nm without facing major problems with quantum tunneling.

 

[Romeo_Nightfall]

TMSC is pure madness!

crazy innovation - not found near apple.

 

[BuffaloTF]

What comes after 1nm?

SoA -- System on Amoebas.

 

[rp2011]

.

 

[Romeo_Nightfall]

As it stands right now, it seems with existing technology and known physics, it’s doubtful we will get past 3nm with facing major problems with quantum tunneling.

Which means we are close to the next innovation and major leap!

 

[rp2011]

Which means we are close to the next innovation and major leap!

Optimism is a nice thing. Different materials possibly?

 

[coolfactor]

Starting to get boring now, apple has no completion 😁😂 🏎💨

Crushing the completion with ease.

Completion or Competition?

As for 3nm... this is getting nuts! Is that even a measurable thickness! I guess so, but holy cow, batman! There ain't anywhere to go from here. I can't imagine 1nm would be possible.

 

[nikaru]

TMSC is pure madness!

crazy innovation - not found near apple.

Innovation? It is not about innovation at all, not at least at TMSC side. The only thing TMSC did was to buy the equipment required to fabricate at 10nm, then at 5nm and in the future at 3nm. Surprisingly, there is only one company in the world that makes the machines that makes these chips and TMSC bought everything. TMSC a is pure manufacturer, they are not designing the chips, nor the equipment they use to manufacture the end product. The only merit TMSC has is that it was able to sign the right agreement with the right company at the right moment.

Rise of TSMC gives windfall to Dutch chipmaking equipment giant

Sales to Taiwan triple as US-China trade war alters supply chains

asia.nikkei.com

 

[Captain Trips]

Back to 14nm 😁😂

[Obligatory Back to the Future misquote]

Doc: "Strap in Marty, once the size gets to 1 nm, we will be sent back in time to 14 nm chips."
Marty: "I always wanted to see how slow a device with a 14 nm chip would be."

 

[blazerunner]

Innovation? It is not about innovation at all, not at least at TMSC side. The only thing TMSC did was to buy the equipment required to fabricate at 10nm, then at 5nm and in the future at 3nm. Surprisingly, there is only one company in the world that makes the machines that makes these chips and TMSC bought everything. TMSC a is pure manufacturer, they are not designing the chips, nor the equipment they use to manufacture the end product. The only merit TMSC has is that it was able to sign the right agreement with the right company at the right moment.

Kind of like how Apple owes everything to Xerox.

 

[Serban55]

So next year be ready for the iphone and the mac with 3nm since, i guess all of those will come after half of 2022

 

[McScooby]

Bonkers.

 

[itexas]

That would be nice if true, but AMD bought Nuvia headed by the former Apple Silicon designer. Their chips boast the same performance as Apple’s M1’s naturally.
Apple is in no position to be cocky as the competition is only waking up, and the market for non Apple computers is far greater.
Competition is already nipping at Apple’s heels.

I think you meant to say that Qualcomm purchased Nuvia, not AMD. And their plan is to use the technology to speed up their special purpose processors like those used in their modems.

I have no doubt that AMD, Intel, and others are strategizing and working toward better performance, but I think that for the next 2-3 years Apple will have the lead. And even if you're not a MacOS/iOS user that's fine. Competition is always good and Intel and AMD make some damn fine processors. In the meantime, companies like Microsoft are hedging their bets and looking at making their own M1-style chips rather than waiting on Intel.

And speed isn't everything. I like playing on my Nintendo Switch just as much as the PS5.

 

[ftaok]

What comes after 1nm?

nano is 10^-9 and the next one down is pico, which is 10^-12. So 0.5 nm is 500 pm, which sounds kinda silly.

Angstrom is 10^-10 meters, but it's not an SI unit. But it would sound pretty cool to have a chip process at 8 Angstroms.

 

[ftaok]

The only merit TMSC has is that it was able to sign the right agreement with the right company at the right moment.

That, in and of itself, is a sort of achievement. TSMC has resisted the urge to compete against Apple, so as not to rock that apple cart. If TSMC were to start designing their own chips and SOC's to sell to Apple's competitors, you would see Apple start to reduce orders from them.

As it stands, I'm sure Apple has contingency plans in case that happens. Heck, Apple might do it anyway, just to bring as much of the widget production in house.

 

[mschmalenbach]

That, in and of itself, is a sort of achievement. TSMC has resisted the urge to compete against Apple, so as not to rock that apple cart. If TSMC were to start designing their own chips and SOC's to sell to Apple's competitors, you would see Apple start to reduce orders from them.

As it stands, I'm sure Apple has contingency plans in case that happens. Heck, Apple might do it anyway, just to bring as much of the widget production in house.

Agreed On your first paragraph. I work for a major semiconductor manufacturer and we have our own fabs, unusually. Fabricating these devices, even on the much larger process nodes we use, is a lot more involved & complex than simply buying the equipment, installing it and switching it on. It takes considerable engineering, physics and chemistry skills to tune the operation to the point where yields are high enough and consistent enough to achieve the kinds of margins needed for commercial viability.

I'm pretty sure this is a major reason for Apple not building their own fab to make their own devices in - even though on paper they have more than enough cash to do so.

 

[cmaier]

What comes after 1nm?

0.6 nm

People used to ask “what comes after 1 micron”

It’s like people never heard of fractions, or something.

 

[ateslik]

A couple of excerpts from the internet to put this into perspective....

A nanometer is equal to one-billionth of a meter. Atoms are smaller than a nanometer. One atom measures 0.1 nm-0.3 nm, depending upon the element.

The width of a human hair is approximately 75,000 nm. Objects that small require scanning electron microscopy to be seen. A strand of human DNA is 2.5 nanometers in diameter.

A sheet of paper is about 100,000 nanometers thick. Your fingernails grow 1 nanometer every second.

The Covid-19 virus is about 120nm. Shaquille O'Neal is 2,160,000,000 nanometers tall.

 

[jlocker]

Looking at the insides of my Texas Instruments TI-1200 from 1975 and the large computer chips and these small chips being produced. Amazing how far we have come in such little time. And it still works to this day And it only cost my dad $120 dollars in todays money

 

[jlocker]

0.6 nm

People used to ask “what comes after 1 micron”

It’s like people never heard of fractions, or something.

We are almost at the end of the electron world and we are transitioning to the optical chips. Light travels faster than electrons on a chip.

 

[Captain Trips]

So, I really had to resist the urge to add a post saying "Paging Doctor @cmaier , paging Doctor @cmaier !"

Because I knew he would chime in once he saw this article and had something he wanted to say.

 

[cmaier]

We are almost at the end of the electron world and we are transitioning to the optical chips. Light travels faster than electrons on a chip.

Light only travels twice as fast as the electrons. And most of the time delay on a chip is not due to the time-of-flight of electrons, but due to the time it takes to charge capacitors.

Also, we are not almost at the end of the electron world. No actual transistors are as small as the node size. The node size is simply a reflection of the resolution of the process, not of the dimensions of anything on the die. We’ve got quite a ways to go before transistor gates become small enough where we are at the end of that road and where there aren’t geometrical tricks to be played. Heck, you can even switch to HBTs instead of FETs at that point, and then you rotate the axis that matters by ninety degrees.

 

[Captain Trips]

What comes after 1nm?

Why, the SOFAC - System On Fish And Chips. I hear Apple's British engineering branch (I have no idea if Apple really has a technical location in England) is working on this: