As if the iPhone needed another performance jump. Nothing in iOS or the App Store will likely challenge the A14, never mind the A15/16 for a couple more years.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
iPhone 14 Pro and Redesigned MacBook Air Reportedly Stuck Using Technology Behind A15 Chip
- Thread starter MacRumors
- Start date
- Sort by reaction score
I have an Xs Max which is still running strong. I am curious about the pill shape because I hate the notch. If the reports are accurate the all glass screen is around the corner in a year or two so I think I can wait it out.
Not sure what the negative skew on this article is intended for.
We always assumed the “next” Apple Silicon Mac chip would either be based on an A15 or A16. Both of which are faster than the A14/M1. And if the A16 is just a slightly improved A15 then it doesn’t matter what the M(next) chip is for macs.
If I were to speculate, I would assume the new M chip WILL be based on the A16 as it is now just going to be an enhanced A15 with better RAM etc anyway.
Doesn’t really matter what they call it but my general thinking is it WILL be called M2. The performance gains from faster clocks, more powerful E cores (and more efficient), and extra GPU cores should equate to quite decent total performance gains. Especially when you compare typical industry wide generational gains from the likes of Intel and AMD etc. 10-20% gains are better than nothing. And this is 10-20% on top of the already stellar M1 performance.
I guess we will find out next week!
We always assumed the “next” Apple Silicon Mac chip would either be based on an A15 or A16. Both of which are faster than the A14/M1. And if the A16 is just a slightly improved A15 then it doesn’t matter what the M(next) chip is for macs.
If I were to speculate, I would assume the new M chip WILL be based on the A16 as it is now just going to be an enhanced A15 with better RAM etc anyway.
Doesn’t really matter what they call it but my general thinking is it WILL be called M2. The performance gains from faster clocks, more powerful E cores (and more efficient), and extra GPU cores should equate to quite decent total performance gains. Especially when you compare typical industry wide generational gains from the likes of Intel and AMD etc. 10-20% gains are better than nothing. And this is 10-20% on top of the already stellar M1 performance.
I guess we will find out next week!
We’re not at the same stage of development. This is more like the first 20 or so years of cars, where there were still big technology improvements every year. Cars are relatively mature, the core technology anyway. Give it another hundred years.
Yeah my thoughts as well. “Stuck” on technology their competitors have still not caught up to. Apple set the bar really high, and a lot of manufacturers can’t get anything made at all.
Try being one of the suckers that bought the very last Intel MacBook Air, in the i5 configuration no less. It’s not exactly slow, but boy did I feel stupid when 8 months later here comes the M1 MBA everyone is still raving about 18 months later, still with no clear successor in sight.
So yeah, enjoy that M1 MBA. When you next upgrade, just bite the bullet and get a MacBook Pro. Got an M1 Pro base model to replace that Intel MBA and the difference is night and day. It’s not gonna be that dramatic for a new iteration of MBA.
A16 has to offer something over A15, otherwise Apple would have just kept A15 in the iPhone 14 Pro and Pro Max, as well, and saved A16 for the iPhone 15 family.
My assumption is that there will be improvements to both CPU cores (performance and efficiency) as well as the GPU cores.
I remain confident M2 will be based on A15 cores and M3 will be the one that incorporates A16 cores, but we shall know soon enough.
My assumption is that there will be improvements to both CPU cores (performance and efficiency) as well as the GPU cores.
I remain confident M2 will be based on A15 cores and M3 will be the one that incorporates A16 cores, but we shall know soon enough.
Apple does seem like they are in a bind here. It's true that the average Macbook Air buyer doesn't care what the chip is called. But lots of eyes are on them generally about this, and if they market this chip as an M2 and it doesn't carry a significant improvement over the M1, people are going to worry about what that means for the presumably upcoming M2 Pro, Max, and Ultra. Or, if it does carry a significant improvement, then they are going to hold off on buying any M1 machines. Personally I could see them calling this M1x, or M1.5, or really anything that avoids making is seem like this is a whole new generation of chip, because it's clearly a stepping stone to the next generation.
iPads and iPhones are a completely different thing. Most consumers have no idea what chip powers them. Laptops and desktops are very different, tho, and that is what I am referring to when i say computer consumers DO know what chip is in the product (they may not know the specs, but they know the chip name, e.g. "M1")
Exactly, meaning even if the M2 is only moderately faster than the M1, customers will still think that it’s newer and better, so either way it’s a win for Apple.
Customers don’t care about the nanometer count.
Keep in mind even though this new chip is rumored to still be 5 nm, it’s also rumourd to get a huge GPU improvement (up to 9/10 cores instead of 7/8), faster ram and the marginal but still important performance and efficiency gains that the A15 got over the A14.
Because of all of those improvements I personally think that Apple will be more than comfortable calling this an M2.
At the end of the day, the name is nothing but marketing, as someone else said it doesn’t actually mean anything in relation to the processor itself.
Apple could use 3 nm for the next MacBook Pro and still call it M2pro, that’s just marketing not actual specifications All of that plus a new design and any other improvements they choose to add make this upgrade look quite nice
Kuo's argument is this cuts the immense credibility Apple has built with M1 by calling a minor upgrade M2.
Sure, Apple could call it an M3 or M16 if they wanted to, but branding itself has value.
In addition, calling something an M2 if it performs like an M1.2 reduces the perceived performance of M1 Pro series as well as the real 3nm M2 Pro series.
Last edited:
But since the M1 is the first of a family, there is no precedent for how Apple will name/number the chips beyond this first generation. Or even what will constitute a "generation" for Apple marketing. I don't think it hurts their "credibility" at all - how could it? If we were on the 15th generation, and they had a history of incrementing the number with a particular attribute (e.g., with every 1nm die shrink), and then suddenly incremented chip name while using the same die size, that would hurt credibility. But there is no history here, so maybe they will increment the number with every small improvement. Who cares?
I believe if we have an M2, it will be revised version of the M1. Apple can barely make laptops, why anger people waiting for M1 devices?
If the M2 doesn’t carry a significant improvement over the M1, they’ll still sell roughly 20 or so million Macs, 10 million to folks that have never owned a Mac before and 10 million that have owned a prior Mac.
Keep in mind we're 18 months since M1 launch, so any kind of update is expected to a bang rather than whimper. Look at the way M1 Pro, Max, and Ultra perform compared to M1. I would say there's certainly a level of expectation of how M2 needs to perform to earn that name.
I'm more concerned of battery life at this point of the phone tech....there is enough performance for most on those chips.
N5P only offers -10% power reduction or +5% performance over baseline N5. Zero density improvements. It is not a huge leap.
Basically about the same tweak that the A15 was over the A14. Decent chance the A15 already is at N5P.
N4 is similar only it gets some density improvements also.
N4P is -22% in power reduction or +11% performance
TSMC Unveils N4X Node: Extreme High-Performance at High Voltages
Lots of folks on the forums seem to want to skip over N4. Mediatek isn't. Qualcomm isn't. They are placing large orders for it. But somehow Apple "has to" skip it. ( for some reason it can only pick a 'P' version. Either has to be 5P or 4P. ).
"... N4P will deliver an 11% performance boost over the original N5 technology and a 6% boost over N4 ... "
TSMC Expands Advanced Technology Leadership with N4P Process|Taiwan Semiconductor Manufacturing Company Limited
Hsinchu, Taiwan, R.O.C., Oct. 26, 2021 - TSMC (TWSE: 2330, NYSE: TSM) today introduced its N4P process, a performance-focused enhancement of the 5-nanometer technology platform.
11% - 6% = 5% performance for the N4 over N5. Just like N5P only get a bigger transistor budget to boot.
Power savings probably in same "N5P"-like ball park.
If 3-4 years ago Apple had targeted Spring or June 2022 for a M2 launch then N4 would be a better choice. N5P would have been a better target for the A15 a 3-4 years before its Fall '21 launch date (as it was a 1H 21 target).
Probably, most not all that different. At least CPU cores. Only getting a 6% logic density improvement. The A14 -> A15 move was already a die area bloat. 88mm^2 to 107mm^2. So a 22% increase. Pretty good chance Apple is elther looking to 'claw' some of that back. So a 16% increase ( 102mm^2) or spend that 6% or on fixing things like :
a. more work on 'weak' video out. (heading toward normal modern USB-C utility coverage. )
a. Either further flush out the ProRes en/decoder more or not falling even further behind locking a AV1 decoder. ( intel , Nvidia, and AMD are going to cover it)
c. better GPU parallel coverage ( like A14-> A15 small bump to GPU core count. GPU cores are smaller than P cores). [ if regular iPhone get the non-binned-down GPU core count than can bump the A16 up. ]. May make sense to do more NPU instead. But allocation to smaller sized cores.
Or a bit of both ( a narrow set of un-core (CPU core) fixes and overall die shrink. Take 3% shrink and a more narrow performance bump with higher synergy with camera/radio/etc improvements. ).
For the M2 also.
a. fix relatively weak video out ( M1 can't make TB4 muster because short on video out).
b. M1 doesn't even have any ProRes en/decode at all. ( versus A15 , Pro , Max , and up in M-series)
c. 1-2 GPU core increase.
That would have very little to do with goosing Geekbench 5 single threaded drag racing scores. There is more than just a CPU P cores in a SoC. Improving those others matters in the non-hyper CPU benchmarked focused user community.
There are several things about the M1 that is weak. Number of TB ports backslid from Intel Mini. The max memory backslide from Intel Mini. The video out is weaker. That is a contributing factor to why the Intel Mini is still being sold TWO YEARS after Apple announced this transition. Apple dropped a mini case with a A14X in it as a transition kit and the Intel mini is still around. The M1 is generally good but it also has issues that need to be fixed.
Apple doesn't need a hefty fab process upgrade to fix it. For the Pro/Max M2's they can take the clock speed bumps to improve the performance. Not huge leaps forward , but better than it is now. For folks who are on a upgrade timeclock ( every 3 , 4 , or 5 years), they'll buy an incremental improvement because it will be better than what they got now.
Last edited:
All PC’s are the same whether it’s a Windows or macOS. The vast majority doesn’t care and wouldn’t understand if you tried to explain what CPU a system has. It doesn’t matter to them because whatever they buy is going to be more performance than they’ll ever use. And, it’s no surprise that the iMacs come in colors and the MacBook Air comes in gold as those are the systems Apple sells the most of.
Transistor density delivers all of the 30% to 50% boost? No. However, that transistor density is delivering no significant fraction of that 30% ro 50% boost ? It is no for that also.
Every time Apple 'sat' on a same process the die size got bigger.
A9 -> A10 TSMC 16nm 105 -> 125 ( +19% ) [ i.e., Adding transistors made a difference. ]
A12 -> A13 TSMC 7nm 83 -> 98 ( +18% ) [ i.e., Adding transistors made a difference. ]
A14 -> A15 TSMC 5nm 88 -> 107 ( +22% ) [ i.e., Adding transistors made a difference. ]
Apple's largeer than the competitive caches make a huge contribution to their performance numbers. If the A14 and A14 were still stuck with the same size caches on the A9 they'd be in a world of 'hurt'. Apple's 'phone' SoCs don't have a celluar modem in them. Most of there competition does. Yet Appel's SoC dies are not smaller than their competition. All that extra budget to cache and GPU cores , etc is a significant contributor to their performance leads.
Do the other architectural adjustments help ( e.g., going wider on CPU core function units , better branch prediction and even deeper prefetch , going wider on memory lanes . )? Yes. But if take the transistor budget back to the A10 era (and just make much smaller dies ) then big chunks of the performance goes also. Apple has put a floor under the A-series die size. It doesn't go below 80mm^2 and it generally doesn't climb as high as the A10 did (into A-X series , iPad Pro , 'big die' size range that the M1 inherited. )
The Qualcomm 8cx gen 3 goes with four X1 cores and 710 "little" cores and it jumps into being competitive with the Gen 11 Intel and AMD 4000/5000 mobile offerings.
Qualcomm’s 8cx Gen 3 for Notebooks, Nuvia Core in 2022/2023
A hefty chunk of that is just a bigger transistor budget than what Qualcomm has been using in their smartphone offerings. [ still has a celluar modem soaking up space though, so not surprising not beating newer AMD/Intel offering. ]
If Apple was not using bigger transistor budgets to goose perfmance improvements then the A-series die would have shrunk below 80mm^2 along the way over the last 10 years.
P.S. the very first die that Apple did by themselves, A4, was 53mm^2 but since they have been on a long term plan cadence they haven't been anywhere near that small. 80mm^2 is pragmatically the 'floor' that they won't fall below. So over 2-3 year period of time they need a process shrink to put more transistors into that area to go grab more performance. if the shrinks stop then the general overall performance will stop also. ( the size of the package has limits for the containers that Apple puts them into. Plus the amount of money Apple is willing to spend. )
Last edited:
Not at the same size die level ( efficient die layout )
A9 and A10 were same process node, but the A10 got bigger.
A9
".... Apple A9 chips are fabricated by two companies: Samsung and TSMC. The Samsung version is called APL0898, which is manufactured on a 14 nm FinFET process and is 96 mm2 large, while the TSMC version is called APL1022, which is manufactured on a 16 nm FinFET process and is 104.5 mm2 large. ..."
Apple A9 - Wikipedia
en.wikipedia.org
A10
"... he A10 (internally, T8010) is built on TSMC's 16 nm FinFET process[1][9] and contains 3.28 billion transistors (including the GPU and caches) on a die size of 125 mm2.[10] .."
Apple A10 - Wikipedia
The A10 'bloated' up to "iPad Pro" A--X / M1 size and hasn't been that big since. [ The A10X had to jump to TSMC 10nm to stay in the 120-140mm^2 zone for a 'bigger A series' offering. ]
The A10 probably got some die space layout uplift because it was not dual sourced. but did Apple "effiecenly" save space to add substantive new features is on thin ice for evidence ( it is a bigger die). Did they make a bigger die that usually showed no increase in power levels. Yes. Did they save die space? No. ( got better at turning stuff not being used off. )
Apple iPhone 7 Teardown | Chipworks
I of course acknowledge that the TSMC 16nm A10 is bigger than the TSMC 16nm A9. I'm pointing out that the A10 had a more efficient die layout as per Chipworks' analysis. Essentially Apple put more effort laying out the A10 such as placing the blocks closer together to reduce dead space. Apple also made better use of more efficient track libraries when laying out the chip to save space. Chipworks estimates that if the A10 were laid out in the looser style of the A9, it would have been ~150mm^2 instead of the 125mm^2 it ended up being. I don't have the expertise nor access to the full paid report for the details to verify Chipworks' analysis, so I'm relying on their expert opinion provided in their summary unless you can provide evidence to contradict Chipworks.
The problem with A16/M2 staying with TSMC N5 generation is that they have already done that. A14 (TSMC N5) and A15 ( pretty good chance was already on N5P. ). Apple took a die 'bloat' here also. A14 ( 88mm^2 ) and A15 ( 107mm^2 )
Apple A14 Die Annotation and Analysis – Terrifying Implications For The Industry
Despite TSMC’s claims of a 1.35x shrink on SRAM, Apple’s system cache has only shrunk 1.22x. This has far reaching implications for the industry. Designers cannot use increased LLC sizes as a crutc…semianalysis.com
Apple A15 Die Shot and Annotation - IP Block Area Analysis
TechInsights released a die shot of the A15 and with the help of SkyJuice’s, we are analyzing it today.semianalysis.substack.com
The notion that would keep N5 density and go to a > 107mm^2 for A16 and likewise grow the M2 past the normal "iPad Pro" die size is dubious when there is a N4 process with a 6% shrink readily available. ( MediaTek and Qualcomm are using it this year. Why would Apple skip it for A16 when had already bloated up on A15? That would mean more wafers to make the same amount of die. )
Apple hasn't stuck around for three iterations on the same process general node before. The notion that "well they have done two generation in the past"... yeah. But this wouldn't be two. They have already done two. The real illustrative past example would be looking for is 3. ( N5P gets no density improvements to "add more stuff". It is a power utlization update where can either incremental bump the clock speeds (and use same power) or use less power ( and keep same clock. ). Apple did a little of both of those with A15 )
As for whether Apple has ever stuck with the same general process node for 3 generations before, according to the Chipworks analysis of the A10, they actually have and the A10 is the shipping product. Reportedly, the 20nm planar process used in the A8 is technically the same general process generation as the 16nm FinFET process used in the A9 and A10. The 16nm FinFET process has similar gate densities as a 20nm planar process while the addition of FinFET improved performance. "16nm" is primarily a marketing term to indicate you get the performance improvement of a shrink vs 20nm planar without actual physical shrinkage, very much like the difference between N5 and N5P. The SemiWiki article below also provides an explanation.
16nm FinFET versus 20nm Planar! - Semiwiki
The common theme amongst semiconductor ecosystem conferences this year is FinFETS, probably the most exciting technology we will see this decade. A lot has been written on SemiWiki about FinFETS, it is one of the top trending search terms, but there is some confusion about the process naming so...
semiwiki.com
I do agree Apple should make use of the N4 process since even minor density improvements are valuable even if there isn't much performance or power improvement compared to N5P. I was just trying to point out that it's possible to improve performance/watt without relying on a process shrink and not trying to diminish the importance of smaller processes. The N4 process probably costs more, but that doesn't seem to be an impediment for Apple. As for why Kuo thinks Apple will stick with N5P vs switching to N4, I guess he believes Apple's greater experience with a mature N5P process provides more opportunities for optimization that can overcome N4's minor density improvement and result in a better overall chip. I don't have the knowledge to judge that.
The biggest critique I have heard of is about the graphics, but hard-core gamers aren't that many
An expectation from who exactly?
Better performance, better graphics, faster ram, better battery life are all more significant than it looking better in geekbench.
Not sure why *anyone* expected a major leap from M1 to M2.
Noticeable major performance jumps between every generation of chip is just not something that anyone should expect.