Only iPhone 14 Pro Models to Get 'A16' Chip, Standard Models to Retain A15 - rumour reported by Ming-Chi Kuo

[M. Y. Zuo]

“
Only the iPhone 14 Pro models will have the "A16" chip, while the standard iPhone 14 models will retain the A15 Bionic chip from the iPhone 13, according to insightful Apple analyst Ming-Chi Kuo.

In a tweet, Kuo said that the 6.1-inch "‌iPhone 14 Pro‌" and the 6.7-inch "‌iPhone 14 Pro‌ Max" will get the A16 chip, while the 6.1-inch "‌iPhone 14‌" and the 6.7-inch "‌iPhone 14‌ Max" will retain the same A15 chip from the ‌iPhone 13‌ lineup.

The two more affordable iPhone models retaining the same chip as the previous year could be a major new point of differentiation between the standard and "Pro" ‌iPhone‌ models. Going forward, it seems plausible that Apple could only offer a new chip with the "Pro" models, before it subsequently trickles down to the two cheaper ‌iPhone‌ models the following year.

Kuo added that all four of the ‌iPhone 14‌ models are likely to come with 6GB of memory, with the standard ‌iPhone 14‌ models having LPDDR 4X memory and the ‌iPhone 14 Pro‌ models having LPDDR 5 memory.

Currently, the ‌iPhone 13‌ mini and ‌iPhone 13‌ feature 4GB of memory, while the iPhone 13 Pro and ‌iPhone 13 Pro‌ Max feature 6GB of memory. These amounts are unchanged from the iPhone 12 lineup. For the ‌iPhone 14‌ lineup, Kuo suggests that all models will feature 6GB of RAM, but the ‌iPhone 14 Pro‌ and ‌iPhone 14 Pro‌ Max's LPDDR 5 memory will be up to one and a half times faster and up to 30 percent more power efficient.


Kuo's claim stands in contrast to a rumor from Haitong International Securities' Jeff Pu, which suggested suggested the ‌iPhone 14 Pro‌ models will feature 8GB of RAM, the same amount as the Samsung Galaxy S22 models. That being said, Pu has a mixed track record with Apple rumors. For example, he accurately claimed that 16-inch MacBook Pro and iPad Pro models with mini-LED displays would launch in 2021, but he was incorrect about HomePods with 3D sensing cameras launching in 2019. This may bring the 8GB RAM rumor into question now that Kuo, a more established analyst in the Apple space with a better track record, is claiming otherwise.
“

From: https://www.macrumors.com/2022/03/13/kuo-only-iphone-14-pro-models-to-get-a16-chip/

Ming-Chi Kuo is well known in Apple circles to only report on highly reliable rumours, I don’t think he’s gotten any wrong in the last 5 years.

It seems this rumour matches up with rumours from TSMC that N3 will be delayed and much more expensive than N5 per wafer due to SRAM barely scaling at all, EUV, etc.

 

[Daniel Nenni]

Apple has never done this before so I doubt it but would not rule it out completely. According to the other rumors the iPhone 14 will be 4nm silicon. Are we past that now? Either iPhone 14s will be half and half or all 3nm?

 

[IanD]

Apple has never done this before so I doubt it but would not rule it out completely. According to the other rumors the iPhone 14 will be 4nm silicon. Are we past that now? Either iPhone 14s will be half and half or all 3nm?

The rumour is that TSMC 3nm delays made it too late for any of the iPhone 14, and that for the A16 chip Apple are using TSMC 4nm instead which is a slightly shrunk and slightly faster 5nm process (smaller improvement than 3nm). IIRC the A15 chip uses an "Apple special" version of TSMC 5nm.

 

[Daniel Nenni]

The rumour is that TSMC 3nm delays made it too late for any of the iPhone 14, and that for the A16 chip Apple are using TSMC 4nm instead which is a slightly shrunk and slightly faster 5nm process (smaller improvement than 3nm). IIRC the A15 chip uses an "Apple special" version of TSMC 5nm.

TSMC has N3 Apple, N3, and N3E. Are you saying all are delayed or just N3E? All Apple processes are special versions so general news and rumors about N5, N4, and N3 do not necessarily apply.

 

[IanD]

TSMC has N3 Apple, N3, and N3E. Are you saying all are delayed or just N3E? All Apple processes are special versions so general news and rumors about N5, N4, and N3 do not necessarily apply.

I think you know that N3E is intended to be the "real TSMC 3nm" process, with better manufacturing margins but also a bit lower density and better performance than N3 (or N3 Apple), because TSMC found that N3 (and presumably Apple N3) was short of target performance and difficult to make -- which closed up the cost/performance/density gap with N5, and even more so with N4/N4P (which can just be a simple shrink of N5).

So N3E is delayed (too late for A16 chip?) compared to the N3 target timescale, and I guess that TSMC don't really want to make lots of chips in N3 for Apple or anyone else because it's difficult to make.

 

[Daniel Nenni]

I think you know that N3E is intended to be the "real TSMC 3nm" process, with better manufacturing margins but also a bit lower density and better performance than N3 (or N3 Apple), because TSMC found that N3 (and presumably Apple N3) was short of target performance and difficult to make -- which closed up the cost/performance/density gap with N5, and even more so with N4/N4P (which can just be a simple shrink of N5).

So N3E is delayed (too late for A16 chip?) compared to the N3 target timescale, and I guess that TSMC don't really want to make lots of chips in N3 for Apple or anyone else because it's difficult to make.

Was the difficulty EUV related? From what I hear N3 started at 30 EUV layers and is down closer to 20 layers. That may also be a result of not having enough EUV systems to scale up N5, N4, and N3 in the coming months.

In order for Apple to make the Fall iProduct ship dates the process is frozen by the end of Q4. I'm not sure what TSMC will classify the iPhone 14 SoC as in terms of revenue, N5 or N3, but I can assure you these chips are being made in the new N3 fabs. The N5 fabs are overbooked as it is. We will know soon enough.

 

[IanD]

Was the difficulty EUV related? From what I hear N3 started at 30 EUV layers and is down closer to 20 layers. That may also be a result of not having enough EUV systems to scale up N5, N4, and N3 in the coming months.

In order for Apple to make the Fall iProduct ship dates the process is frozen by the end of Q4. I'm not sure what TSMC will classify the iPhone 14 SoC as in terms of revenue, N5 or N3, but I can assure you these chips are being made in the new N3 fabs. The N5 fabs are overbooked as it is. We will know soon enough.

Depends what you mean by "EUV related" -- from looking at the N3 vs. N3E design rules some of the pitches have been increased by a few nm, I assume that TSMC had just pushed the pitch shrink from N5 a bit too far, hence their explanation of "improving manufacturing margin". They also needed to get some extra speed to hit the N3 power/performance targets, and bigger pitch lowers capacitance which improves speed and power at the cost of a density decrease.

As to exactly what Apple are using, all I can see is the delays in dates that we have seen -- especially for N3E vs. original N3 targets -- suggest that rumours that Apple had to stay in N5/N4/N4P for one extra chip instead of moving to N3 might well be true, but obviously I have no direct knowledge of this. Or maybe Apple will use N3 (whatever variety) for the A16 in the smaller-volume higher-price Pro/Pro Max?

Bear in mind that there's no such thing as an "N3 fab", there's the new EUV fabs that were built to cope with expected N3 demand including that from Apple and these were built on time; if N3 production (or now N3E) is delayed then they won't sit idle, they'll initially be used for products in N5/N5P/N4/N4P (1st/2nd/3rd/4th generation 5nm processes) that were originally planned to be 3nm but have been shifted to 5nm due to timescale pushout. I'm aware of several such (non-Apple) products...

 

[name99]

Apple has never done this before so I doubt it but would not rule it out completely. According to the other rumors the iPhone 14 will be 4nm silicon. Are we past that now? Either iPhone 14s will be half and half or all 3nm?

"Never done this before" is a strong claim!

On the one hand we have the iPad line where versions of this have been standard for a while. On the other hand we have the iPhone 5C which was released in parallel with the iPhone 5S yet had an A6 (iPhone 5S had A7).

So in a way it's a more an issue of how explicitly does Apple want to bifurcate the line?
iPads come as
- iPad cheap (old phone chip)
- iPad middle (used to current phone chip, now M1)
- iPad expensive (X or M1 chip)

iPhones come as
- iPhone cheap (ie SE, sometimes lagging chip)
- iPhone middle (the "current" chip)
- iPhone expensive (mainly better camera and screen)

Moving to differentiate iPhone middle/cheap by SoC (and or RAM) is not, per se, a crazy idea.

The real issue IMHO is that Apple also uses a SECOND line of differentiation, namely selling the iPhone of last year and two years ago...
(iPhone 11 is $499, iPhone 12 is $599, iPhone SE3 is $429)
Is this a sensible pricing scheme? Is it stable?
Sticking with last year's CPU seems like it will absolutely up-end that "social" equilibrium.

So I'm not sure how they plan to work around that. Maybe they've concluded that selling last year's model (a very good idea at the time, IMHO) no longer makes sense, that the SE strategy (including keeping it around for a few years, SE lasted 4 yrs, SE2 lasted 2 yrs) is a better low cost strategy?

I could imagine something like this year we get
- iPhone cheap remains SE3
- iPhone middle is the iPhone 14 (slightly better than iPhone13, slightly cheaper, same A15 SoC
- BUT no iPhone 13 as the $599 option, that goes away
- iPhone 12 still available as the $499 FaceID option
- iPhone expensive with A16

then next year
- iPhone cheap = SE4 gets FaceID, maybe using A15
- iPhone middle uses A16
- iPhone expensive uses A17

Now three clean tiers, no need for the weird cheap SE vs cheap iPhone 11 business (since the difference between them of FaceID or not no longer exists), and a nicely cleaned up line that maybe makes better business sense?

 

[M. Y. Zuo]

Depends what you mean by "EUV related" -- from looking at the N3 vs. N3E design rules some of the pitches have been increased by a few nm, I assume that TSMC had just pushed the pitch shrink from N5 a bit too far, hence their explanation of "improving manufacturing margin". They also needed to get some extra speed to hit the N3 power/performance targets, and bigger pitch lowers capacitance which improves speed and power at the cost of a density decrease.

As to exactly what Apple are using, all I can see is the delays in dates that we have seen -- especially for N3E vs. original N3 targets -- suggest that rumours that Apple had to stay in N5/N4/N4P for one extra chip instead of moving to N3 might well be true, but obviously I have no direct knowledge of this. Or maybe Apple will use N3 (whatever variety) for the A16 in the smaller-volume higher-price Pro/Pro Max?

Bear in mind that there's no such thing as an "N3 fab", there's the new EUV fabs that were built to cope with expected N3 demand including that from Apple and these were built on time; if N3 production (or now N3E) is delayed then they won't sit idle, they'll initially be used for products in N5/N5P/N4/N4P (1st/2nd/3rd/4th generation 5nm processes) that were originally planned to be 3nm but have been shifted to 5nm due to timescale pushout. I'm aware of several such (non-Apple) products...

How costly do you think changing the production line from an N4 process to N3 would be? I can imagine Apple would have large enough volume to book the entire fab starting Q1 or Q2 2023. Would TSMC really set a fab going for just 9 months, or less, then switch over? Even with some deep pocketed buyers it seems like it would be a lot of money spent on shuffling equipment around.

 

[IanD]

How costly do you think changing the production line from an N4 process to N3 would be? I can imagine Apple would have large enough volume to book the entire fab starting Q1 or Q2 2023. Would TSMC really set a fab going for just 9 months, or less, then switch over? Even with some deep pocketed buyers it seems like it would be a lot of money spent on shuffling equipment around.

There's not much direct cost to swapping a line over from N4 to N3, most or all of the process steps are similar, just N3 has more EUV masks -- no shuffling of equipment is needed. It's difficult running two different processes on the same line (e.g. dropping some N3 lots in amongst the N4 ones) because any changes in setup and process flow reduce yield and throughput, which is the last thing you want.

But nothing costs as much as having an N3-capable fab half-empty -- even for 9 months -- because N3 delays means big customers switched to N4, lowering short-term demand for N3 and increasing it for N4 so it exceeds supply -- in that case it makes a lot of sense to run one (or more) of the lines targeted at N3 on N4 instead until the N3 demand rises later, then switch over to N3.