Rumor Reversal: A20 Chip for iPhones Said to Use TSMC's 2nm Process

[vantelimus]

2nm is 20 Angstoms which is around 10 atoms wide (gate length). You can't shrink forever.

The Planck length is 10^-25 Å, so there is a limit. 🥸

 

[tridley68]

Tech blogs really need to educate more about how much these node names are pure marketing, because someone asks this every single time this comes up as we get down to the single digit nodes. TSMC N7 for example had 36nm fin pitches and 50nm gate widths or the other way around iirc. These are three dimensional objects with sizes on all sides in double digits and there's plenty left to shrink.

Node names USED to mean a minimum feature size, but if that's one feature in a library of thousands of transistors, how much is that really the reality of a chip when every other feature averages many times larger. Density is a better metric in mtr, but you can also trade off density for less thermal hotspotting and higher clocking etc, so it's not always highest is best.

What happens after 2/1nm? They just switch to Angstroms as Intel already will be with 20a and 18a etc.

Don't these tech writers just copy and paste from some other blog

 

[Channan]

What happens at 0nm? Do we win a prize?

No, we move to the negatives. -1nm is after that. It actually creates more space in the phone.

 

[knappeduivel]

Once again: The 2nm is just a marketing term, it does not correspond to an actual physical measurement

2nm is not just a marketing term, it does correspond to an actual physical measurement

 

[Le0M]

33%. Pretty big change. That doesn’t translate to 33% more battery life tho. Likely the clock speed will go up and the battery size could shrink a bit. It’ll be a balancing act between more battery life vs. more power for same battery life vs. smaller battery supporting design changes like bigger heat sink etc

You can't just do the math giving for granted that "3nm" and "2nm" chips are actually of that size, cuz they are not, it's just marketing like many before me said, so you can't make a proportion between these two and know the gain.

 

[tenthousandthings]

Did he also clarify his CoWoS mention? That also seemed unlikely for the A-series.

 

[Miles Teg]

What happens at 0nm? Do we win a prize?

Once 0 nm is reached next is -1 nm

 

[JitteryJimmy]

I wouldn't be surprised if the C1's successor finds its way into the A20's die. Yes, a bigger package with more thermal engineering, but a much more simple logic board and smaller overall real estate.

 

[bigandtasty]

Nice, this means 2nm for the MBP is coming for M6. Should be easy to make this M1 Max last until then.

 

[geta]

What happens at 0nm? Do we win a prize?

The world will come to an end, they'll reset everything and start over.

 

[Tyler O'Bannon]

33%. Pretty big change.

You can't just do the math giving for granted that "3nm" and "2nm" chips are actually of that size, cuz they are not, it's just marketing like many before me said, so you can't make a proportion between these two and know the gain.

It’s been pretty close to that throughout several previous node size changes

 

[gpat]

The world will come to an end, they'll reset everything and start over.

Can't wait to smell a freshly baked MOS 6502 all over again.

 

[joshwithachance]

Do we even care anymore? Like realistically how much is this ACTUALLY going to improve our user experience? Barely. It's all marketing at this point.

 

[ifxf]

What happens at 0nm? Do we win a prize?

when they start quoting 100pm they will confuse most people.

 

[bodonnell202]

How much power is this actually going to save? Like, if the iPhone 16 pro used a 2nm chip (if it were even possible to)- how much more usage time would that give the user?

In July 2022, TSMC announced that its N2 process technology was expected to offer 10–15% higher performance at iso power or 20–30% lower power at iso performance and over 20% higher transistor density compared to N3E. This means if Apple just ported the current A18 design to N2 and ran it at the same clock speeds it would use 20-30% less power to do the same amount of work. This isn't what they will do as they will design a new chip which will mostly increase performance, but may be somewhat more efficient too.

 

[bodonnell202]

What happens at 0nm? Do we win a prize?

They switch to a smaller unit of measurement and continue. Looks like the industry is moving to using Angstroms and the next process after 2 nm (N2) will be known at A16 at TSMC.

 

[bodonnell202]

I'm under the impression it's not insignificant... but Apple won't actually pass the savings onto the user. They'll do some combination of putting in smaller battery making the device smaller (and cheaper to manufacture and distribute, but Apple won't lower prices and they'll probably consider the smallness as justification to raise the prices instead) and putting in some other hardware...

Although maybe not. When was the last time Apple put in new hardware that was at all noticeable to regular users? I feel like you have to go all the way back to the iPhone 7's pressure sensor that nobody cared about and was quickly removed. NFC on the 6 might be the last significant new hardware on an iPhone... that correlates well for me with being the last time I bought a new iPhone instead of a used one. It's depressing to think about how incredibly long the iPhone has been stagnant for. Apple is so massively overdue for having a better CEO.

Except devices have been getting thicker (and later also bigger) since the iPhone 6 and they have been increasing battery capacity for years now so I'm not really sure what you are going on about.

 

[bodonnell202]

Most devices going from 4nm to 3nm saw zero improvements. Apple will say it is a more efficient processor, but those gains were not to be found in battery life tests.

That's because with newer fab generations there is a choice to be made between spending the additional "chip budget" it allows on performance or efficiency and Apple typically spends most of the budget on performance.

 

[bodonnell202]

2nm is not just a marketing term, it does correspond to an actual physical measurement

It does not - read the 2nd paragraph here: https://en.wikipedia.org/wiki/2_nm_process. Process nodes have been purely marketing terms since about 1997.

 

[Octavius8]

Absolutely useless. iPhones battery has kept the same over the years. More efficient CPU is a fallacy. Mediocre CPU and poor battery life in IPhone since forever.

 

[Homme]

Once again: The 2nm is just a marketing term, it does not correspond to an actual physical measurement

yeah like the A16 being marketed for “4nm” according to Apple when it’s actually 5nm according to TSMC ( which is annoying though)

 

[MacBH928]

say, why iphone chips still labeled as A and not M chips? I mean the iPads has the M chips and they are mobile devices. Maybe the A chips are just rebranded M chips?

They will use the angstrom as unit of measure and we start again.

not sure if sarcasm or serious. Surely we have hit the wall with physics?

 

[MacBH928]

Absolutely useless. iPhones battery has kept the same over the years. More efficient CPU is a fallacy. Mediocre CPU and poor battery life in IPhone since forever.

bruh, I may be old, but a phone that runs 10hr straight with wifi, 5g, 6 inch full color HD screen that runs console level games is mind boggling to me. I come from an era with cell phones had monochrome 4 line screens.

Sega game gear had 6 AA batteries on 3inch screen that lasted 4 hours before you had to buy new batteries , I know things advanced but we got to appreciate what we have.

 

[Siliconguy]

What happens at 0nm? Do we win a prize?

The atomic radius of silicon is 110 pm, so the diameter is 220 pm. Zero is not an option.

I'm surprised they got it down as far as they did without diffusion rearranging the atoms into non-functionality in a few months. Then again, there are a surprising number of M-series Macs for sale listed as For Parts Only. Maybe they are dying young.

 

[bodonnell202]

say, why iphone chips still labeled as A and not M chips? I mean the iPads has the M chips and they are mobile devices. Maybe the A chips are just rebranded M chips?

A chips and M chips share a common architecture. M chips just have more cpu and GPU cores than A chips. A chips are smaller and need less power.

not sure if sarcasm or serious. Surely we have hit the wall with physics?

We have not, as feature size is much larger than 2 nm, it’s only a marketing term. When they start using angstroms it will also be a marketing term.