The new 14 and 16” models should jump to A16 architecture next year… right?

[collin_]

I am really hoping that we will be getting annual SoC upgrades in these laptops like we do in iPhones. After all, being able to consistently upgrade Macs in a meaningful way was a primary goal of the transition to Apple Silicon, right?

Assuming they do get upgraded next October, they should jump from A14 architecture to A16 architecture. I’m not exactly sure what TMSC node that would be, but given that we’re now dealing with Apple and TSMC instead of Intel, 2 years of SoC upgrades is likely to be quite significant.

This is one of the main reasons I’m holding out until at least next gen.

Thoughts?

 

[bill-p]

Thoughts?

I guess you didn't get the memo:

Apple May Be Forced to Delay Its Custom 3nm Chips Thanks to TSMC

Due to the setback in 3nm chip production, Apple might be forced to delay the launch of its advanced custom silicon for various products

wccftech.com

Being that TSMC is still unable to sustain enough operation to provide 5nm chips for other customers (Apple is still the only one as far as I know), we're looking at probably a 2-year upgrade period for the Mac. Also note that the original M1 MacBooks were not updated at all this year. If they're updated in summer next year, it'll be 1.5 years since their release.

 

[collin_]

I guess you didn't get the memo:

Apple May Be Forced to Delay Its Custom 3nm Chips Thanks to TSMC

Due to the setback in 3nm chip production, Apple might be forced to delay the launch of its advanced custom silicon for various products

wccftech.com

Being that TSMC is still unable to sustain enough operation to provide 5nm chips for other customers (Apple is still the only one as far as I know), we're looking at probably a 2-year upgrade period for the Mac. Also note that the original M1 MacBooks were not updated at all this year. If they're updated in summer next year, it'll be 1.5 years since their release.

The 3nm delay is irrelevant. The chips could use N5+ (already in iPhones) or (more likely) N4, which has looked more realistic than N3 for 2022 for a while now.

Apple can make architectural improvements (see “A16 architecture“ in the OP) regardless of the process node, which I see as equally important at this point (at least).

So the question still stands. Will Apple refresh these MacBooks a year from now with the A16 architecture (shortly after its debut in the iPhone 14 models)? Haven’t seen anyone weigh in on this yet.

 

[bill-p]

But what are you hoping to see in this theoretical A16 architecture? Even TSMC has said that their 4nm process is just an optical shrink that will yield 6% die shrink but... it's not a revolutionary change like 3nm.

At most, it'll be a 20% improvement, just like how going from A13 (7nm) to A14 (5nm) yielded. That's not nearly enough for the A16/M2 to match the performance of M1 Pro, let alone M1 Max.

Also Apple has slowed down their hardware upgrade cycle significantly in recent time. The first M1 Macs were introduced a whole year ago. The MacBook Pro 14" and 16" should have come out with these new M1 chips much earlier, but instead, it took Apple a whole year. Clearly, TSMC has difficulty scaling up production for the M1 Pro/Max.

And now we just barely got these M1 Pro/Max chips and you're already asking whether we'll see new improved chips next year, when we have no confirmation from TSMC that a better process (either 4nm or 3nm) will even be ready by next year. Sounds like wishful thinking to me.

I'm just saying. We just barely got new devices and you're already looking for an upgrade. That's kind of... rushing it.

 

[Kazgarth]

If you have been following the tech news, the A16 are going to be on 4nm (Which is basically rebrand for 5nm++).

So far only the new iPad pro is guaranteed to get 3nm chip (M2) due to production constraints.

 

[nicho]

I guess you didn't get the memo:

Apple May Be Forced to Delay Its Custom 3nm Chips Thanks to TSMC

Due to the setback in 3nm chip production, Apple might be forced to delay the launch of its advanced custom silicon for various products

wccftech.com

Being that TSMC is still unable to sustain enough operation to provide 5nm chips for other customers (Apple is still the only one as far as I know), we're looking at probably a 2-year upgrade period for the Mac. Also note that the original M1 MacBooks were not updated at all this year. If they're updated in summer next year, it'll be 1.5 years since their release.

Time for Apple to buy TSMC and forget about the other customers...

 

[Jorbanead]

I think it’s extremely hard to predict this because the pandemic really changed everything. We don’t know what Apple had originally planned had everything been on schedule. We don’t know if the current release is plan B, plan C, whatever - so It’s really hard to know or predict any sort of annual, or bi-annual cycle.

What’s clear to me is Apple is going to have to skip a generation at some point for Mac chips - unless they accept that pro/max chips always one year behind the A-series (which I doubt). We know it takes them roughly ~3 years for these chips to come to fruition so my guess is we won’t see a jump for a few years, especially since supply constraints are still widely effecting the industry.

 

[leman]

Who knows how Apple is going to do it and who knows what the industry can deliver… The next gen could be either A15 or A16 based. It’s really impossible to tell…

 

[altaic]

My gut reaction is that A16 is highly unlikely, however, being first gen SoCs (yes, many iterations of the IP has led to these, but the goals and constraints of the previous ultra mobile SoCs are completely different), we could see faster iteration to adapt to the new goals... I recommend stockpiling patience for several months.

 

[Falhófnir]

2023 if the current pattern holds. Consumer chips in even years (2020,2022,2024) high end chips in odd years (2021,2023,2025) so the M2 Pro and Max won’t be refreshed for another 23 months or so, and any products based on them will also have to wait at least that long. Even the bigger iMac hasn’t yet received the M1 Max update yet.

 

[Boil]

Time for Apple to buy TSMC and forget about the other customers...

TSMC has a fab they are opening in the United States at some point; maybe it is actually all for Apple, and they can say their SoCs / SiPs are designed & manufactured in the USA...! ;^p

As for SoC progression; I see the A-series being first, allowing the node / yields / etc. to be "smoothed out" by the time the larger (and therefore more expensive & with lower yields) SoCs come online...

Then the same happens with the entry-level Mn-series SoCs, with the (even larger) Mn Pro / Mn Max series SoCs following those...

 

[Macintosh IIcx]

As for SoC progression; I see the A-series being first, allowing the node / yields / etc. to be "smoothed out" by the time the larger (and therefore more expensive & with lower yields) SoCs come online...

Then the same happens with the entry-level Mn-series SoCs, with the (even larger) Mn Pro / Mn Max series SoCs following those...

Yes, it is highly likely that the M Pro and M Max will trail the A SoC with one year. Part of the reason for this will be that the A SoC will use the latest cutting edge TSMC process to have best possible power efficiency but the yields will not be mature to the point where you want to make a huge M Max on just yet. The TSMC process progress will dictate things. Note however that the M1 Pro and M1 Max have the ProRes encoder and decode also in the A15, so I think that only the CPU will be a generation behind on the M Pro and M Max, not sure about the GPU. Probably also that.

TSMC won’t have 3nm process ready before 2023 so not sure next year will bring a huge leap of performance to the M2 Pro and M2 Max (based on A15?). Also, the next major jump in GPU might be to go from 32 to 48 GPU cores, and that would almost certainly require a mature 3nm.

 

[Boil]

M2 SoC

• 10-core CPU (6P/4E)
• 12-core GPU
• 16-core Neural Engine
• 32GB LPDDR5 RAM
• 200GB/s memory bandwidth

M2 Pro SoC

• 16-core CPU (14P/2E)
• 24-core GPU
• 24-core Neural Engine
• 64GB LPDDRX5 RAM
• 250GB/s memory bandwidth

M2 Max SoC

• 16-core CPU (14P/2E)
• 48-core GPU
• 24-core Neural Engine
• 128GB LPDDRX5 RAM
• 500GB/s memory bandwidth

M2 Max Duo SiP

• 32-core CPU (28P/4E)
• 96-core GPU
• 48-core Neural Engine
• 256GB LPDDRX5 RAM
• 1TB/s memory bandwidth

M2 Max Quadro SiP

• 64-core CPU (56P/8E)
• 192-core GPU
• 96-core Neural Engine
• 512GB LPDDRX5 RAM
• 2TB/s memory bandwidth

 

[leman]

Note however that the M1 Pro and M1 Max have the ProRes encoder and decode also in the A15, so I think that only the CPU will be a generation behind on the M Pro and M Max, not sure about the GPU. Probably also that.

Quick comment on this: M1 Pro/Max seem to share the encoders and display engines with A15 (or maybe it’s the other way around). So there is some mixing and matching across generations at least. But the GPU is still A14.

 

[Hexley]

I think the performance per watt lead of Apple Silicon in the PC industry is such that we will see a ~24 month replacement cycle.

As many loved to point out my assumption of a ~12 month replacement cycle was wrong. Which is odd as my being wrong is a detriment to everyone's wanted outcome of annual meaningful performance per watt improvements.

What is fun to imagine is what will be the performance figures of Apple Silicon on the smartphone, tablet, desktop, watch & laptop will be by Nov 2031.

Will the M1 Max's performance be found in a future Apple Watch?

 

[Serban55]

after decades of A series on iphones/ipads...and M1 pro/Max cannot be placed even in ipad pro...there is no way in 10 years that M1 max perf to be found in an apple watch...of course if Apple still want to be a fashion wearable and not an ipad on your wrist

 

[Hexley]

after decades of A series on iphones/ipads...and M1 pro/Max cannot be placed even in ipad pro...there is no way in 10 years that M1 max perf to be found in an apple watch...of course if Apple still want to be a fashion wearable and not an ipad on your wrist

USB-C charger has more computing power than Apollo 11's computer. Back in the 1960s no one could imagine that ever occurring.

Die shrinks would be less than 1nm by 2030s from today's 5nm.

 

[leman]

As many loved to point out my assumption of a ~12 month replacement cycle was wrong. Which is odd as my being wrong is a detriment to everyone's wanted outcome of annual meaningful performance per watt improvements.

I think it’s too early to tell what the update cycle would be. Given the parts shortage and supply issues, it’s very possible that what we saw until now was not how Apple planned it. Personally, I think that 12 months is what they are going for: A-series snd M-series in autumn and pro chips in spring. Or something like that.

 

[neinjohn]

Was not the rumoured reason for the delayed release of the new 14'' and 16'' the lack of sufficient stock/production on the Mini-LED displays? Maybe Apple and TSMC could have perfectly released the models last year or during the WWDC and the bottleneck is on another part.

I'd also guess the bottleneck is way easier to have on parts that are new/from a redesign and there is not yet a large stock or mass production capacity. For example the Air and the Pro that are largely the same parts from previous years are on stock or 1-3 days away on Apple US site, the M1 iMac not so much.

 

[thenewperson]

Was not the rumoured reason for the delayed release of the new 14'' and 16'' the lack of sufficient stock/production on the Mini-LED displays?

Yep, that was what was blamed by the rumour reporters. Hopefully 2022/2023 things will stabilise; can't imagine the Mac Pro being on M1 cores for 2 years (I'm assuming it's getting M1 Max chips not some later release).

 

[collin_]

I think it’s too early to tell what the update cycle would be. Given the parts shortage and supply issues, it’s very possible that what we saw until now was not how Apple planned it. Personally, I think that 12 months is what they are going for: A-series snd M-series in autumn and pro chips in spring. Or something like that.

Yeah it looks like these Macs were originally meant to be released in the summer and I think that’s relevant to this discussion.

 

[collin_]

But what are you hoping to see in this theoretical A16 architecture? Even TSMC has said that their 4nm process is just an optical shrink that will yield 6% die shrink but... it's not a revolutionary change like 3nm.

At most, it'll be a 20% improvement, just like how going from A13 (7nm) to A14 (5nm) yielded. That's not nearly enough for the A16/M2 to match the performance of M1 Pro, let alone M1 Max.

Also Apple has slowed down their hardware upgrade cycle significantly in recent time. The first M1 Macs were introduced a whole year ago. The MacBook Pro 14" and 16" should have come out with these new M1 chips much earlier, but instead, it took Apple a whole year. Clearly, TSMC has difficulty scaling up production for the M1 Pro/Max.

And now we just barely got these M1 Pro/Max chips and you're already asking whether we'll see new improved chips next year, when we have no confirmation from TSMC that a better process (either 4nm or 3nm) will even be ready by next year. Sounds like wishful thinking to me.

I'm just saying. We just barely got new devices and you're already looking for an upgrade. That's kind of... rushing it.

I am confused by this post. Nobody said anything about the A16 and M2 matching the performance of the M1 Pro and M1 Max. I’m talking about 2021 to 2022 MacBook Pro performance improvements here.

> At most, it'll be a 20% improvement, just like how going from A13 (7nm) to A14 (5nm) yielded.

Okay so a 1-generation jump was a 20% improvement. Extrapolating that, these computers going from A14 to A16 architecture would be a 40% improvement. Is that not significant??? That would be a big deal to me.

Yes I understand that the devices just came out. It’s an unusual situation in that it looked like they were meant to debut in the summer but got delayed because of the displays and that kind of threw the architectural timing off. I’m just trying to make an informed decision on whether I should buy now or wait.

 

[CarbonCycles]

Why...why do we care? How would this improve the product(s) and other business lines?

Apple is more than hardware...they have apply tv, music, application development. I suspect deep down Tim would be happy to kill the MBP line if it wasn't marginally profitable bc it sure has a lot of complainers.

If there were ever to be a group of users who have a valid complaint, that would be the Mac Pro users. The last upgrade, they got a fancy cheese grater.

 

[bill-p]

I am confused by this post. Nobody said anything about the A16 and M2 matching the performance of the M1 Pro and M1 Max. I’m talking about 2021 to 2022 MacBook Pro performance improvements here.

> At most, it'll be a 20% improvement, just like how going from A13 (7nm) to A14 (5nm) yielded.

Okay so a 1-generation jump was a 20% improvement. Extrapolating that, these computers going from A14 to A16 architecture would be a 40% improvement. Is that not significant??? That would be a big deal to me.

Yes I understand that the devices just came out. It’s an unusual situation in that it looked like they were meant to debut in the summer but got delayed because of the displays and that kind of threw the architectural timing off. I’m just trying to make an informed decision on whether I should buy now or wait.

But the only MacBook Pro released in 2021 come with M1 Pro and M1 Max. Are you perhaps thinking about the 2020 MacBook with M1 here?

Also, A14 to A15 is not a 1-generation jump, A15 is a small improvement over A14. But that's irrelevant because the A line only applies to iPhones and iPads. It's clear iPhone is on an annual release cycle. iPad Pro isn't necessarily on an annual release cycle anymore, and MacBook Pro has been on a 2-year release cycle. Note that the 2020 MacBooks with M1 came out a whole year ago.

Seeing as TSMC has not confirmed whether their 4nm process is ready or not by now. i doubt we will see M2 by summer. Heck, we still have not seen iMac 27" with any M1 Pro/Max chip, nor have we seen the Mac Pro with Apple Silicon. The transition has not covered all devices yet. Until we see those devices, I doubt Apple will upgrade any existing Mac with Apple Silicon to a next generation chip.

 

[Hexley]

I think it’s too early to tell what the update cycle would be. Given the parts shortage and supply issues, it’s very possible that what we saw until now was not how Apple planned it. Personally, I think that 12 months is what they are going for: A-series snd M-series in autumn and pro chips in spring. Or something like that.

I hope you are correct but the 2020 volume of Macs is 8.6% of iPhones & iPads.

With a much lower shipment I was thinking they're stretching out the replacement cycle to build up more units sold.

Having too many SKUs increases costs for Apple. That's the problem Intel & AMD had with PCs.