Alleged Apple Chip Plans Suggest 'A16' Will Stick With 5nm, 'M2' to Make Jump to 3nm Instead

[CWallace]

Frankly, the rumor as it applies to A16 makes some sense to me considering the supply and production issues caused by China's rolling lockdowns in Shanghai and surrounding areas where most iPhone and Mac components are produced, sourced and assembled.

I don't believe Apple ever had am SoC road-map that was going to use TSMC's 4nm process. I believe the original plan was to go from the latest 5nm process (N5P) to the first 3nm process (N3). 5NP is the latest and greatest 5nm and a fair bit better in performance and efficiency compared to the first 5nm SoCs Apple was using. And TSMC has sufficient production scale on that process to meet Apple's needs whereas the first 3nm process (N3) is still ramping up and will be replaced by late 2022/early 2023 with an enhanced version (N3E).

So with TSMC able to provide plenty of A15 and A16 on N5P, I could see Apple staying with N5P for A16 and waiting until next year to fab A17 on N3E (or an even later 3nm process) for maximum benefit when TSMC is able to meet Apple's scale.

 

[Zdigital2015]

Am I the only one not that impressed with the M1. I have a 2017 Intel MBP and a 13 M1 MBP (from work) and though I agree on paper the M1 is the better chip, in real world usage I really don't see too much to be impressed with. It might be the restrictions imposed by the company limiting the laptop, but it just seems slow, not snappy.

It’s just you. Going back to an Intel model short of my 27” Core i9 iMac is just painful now.

 

[macsplusmacs]

So with TSMC able to provide plenty of A15 and A16 on N5P, I could see Apple staying with N5P for A16 and waiting until next year to fab A17 on N3E (or an even later 3nm process) for maximum benefit when TSMC is able to meet Apple's scale.

Esp. too since for the Mac Pro, power and heat is much less of an issue when you are piling on the architecture for. quad and octo. etc.

 

[jav6454]

I told you all the final variant would be M1 God.

Bold naming convention...

 

[jav6454]

Ordered a MacBook Pro and canceled. Might as well wait for the m2 at this point with delays.

Delays aren't going away anytime soon. Remember that they have almost 2.5 months worth of delays in MBPs.

 

[deconstruct60]

The A16 will reportedly be based on TSMC's 5nm process, just like the A14, A15, and M1 chips. Previous reports have been unclear about whether the A16 will be fabricated with TSMC's more advanced 4nm process, with an ambiguous report from DigiTimes claiming that Apple plans to use TSMC's 4nm N4P process – but N4P is in fact an enhanced, third-generation version of the 5nm process.

From the TSMC press release that is linked in above.

"... N4P designs will be well-supported by TSMC’s comprehensive design ecosystem for silicon IP and EDA. With TSMC and its Open Innovation Platform® partners helping to accelerate the product development cycle, the first products based on N4P technology are expected to tape out by the second half of 2022. ... "

Something that has just gotten to the 'tape out' process stage in the second half of 2022 is probably not shipping in very high volumes in Q3 2022. ( when iPhones would need to be made for September launch. Or M-series SoCs for October Launch. )

Apple might have their own slight tweak to N4P that was ready to go this quarter (and tope outs went Q4 2021 or very early Q1 2022 ) .

Apple could be using N4 for A16 and M2. It is a 6% reduction in logic area used. It that gives a extra transistor budget . On the M2, to throw in one or two GPU cores and do some tweaks to the other cores on the die without increasing the power consumption. On the A16 shrink it back down to the same size as a normal iphone dies ( 80-90mm^2 range ).

On the other hand, ShrimpApplePro said that the A16 will use TSMC's N5P process. This suggests that the A16 may be less of a substantial upgrade than previously thought.

Decent chance A15 is already on N5P. It does not off any logic shrink at all , just power saving. The A14 -> A15 transition saw the die get bigger ( more stuff took up more space), but there was little to no power hit. ( which likely means Apple took some power saving to roll out a bigger (incrementally more expensive ) die and move iPhone battery life forward. ). There is lillte sense to 'crow' at the process change since it really doesn't bring "much horsepower" due to logic density gains because there are none. ( can slightly increase clocks without burning more power, but not much ).

ShrimpApplePro was unsure about the final naming of the "A16," "M2," and the final M1 chip variant and said that the rumor should be taken with a pinch of salt.

This is the quality of rumors deserving of the front page of Macrumors. Someone who doesn't even know if they got the names of the chips correct. Or just trolling for ad clicks.

If have followed TSMC announcements it is relativley clear that N3 is largely a 2023 product for them. And as highlighted above N4P is a stretch. (maybe Apple's "inside track" gets them product for Q4. )

If the A14 and A15 were on plain TSMC N5 then another A-series chip on that is a stretch.
N3 is also a stretch because it is delivers wafers in 2023 product line. ( production starts in 2022 but wafers coming out of the end of the production pipelines isn't until substantively later.)

Article Link: Alleged Apple Chip Plans Suggest 'A16' Will Stick With 5nm, 'M2' to Make Jump to 3nm Instead

thought they said the Ultra was the final M1 variant.

 

[playtech1]

I was struggling to parse this article, but I think what's being suggested is:

- A16 on N5P, architecture tweaks and LPDDR 5 for speed gains
- M1X - M1 but with A15 based Blizzed and Avalanche cores
- M2 - 3nm and ARMv9

Sounds... sort of plausible?

 

[AltecX]

M1 Uber

I think M1X is most logical given their naming conventions.

I'm going to wait for the new chips to hit the 14in before I move to one. Hopefully they sort out the USB performance issues also.

 

[rme]

28W xeon chip? We can dream of that efficacy!

That also caught my eye ;-)
TDP of the 28 core W-3275M is 205W, not 28W.

Intel® Xeon® W-3275M Processor (38.5M Cache, 2.50 GHz) - Product Specifications | Intel

Intel® Xeon® W-3275M Processor (38.5M Cache, 2.50 GHz) quick reference with specifications, features, and technologies.

ark.intel.com

 

[jdb8167]

So ShrimpApplePro knows more about Apple's future plans than the Senior Vice President of Hardware Engineering at Apple? Sounds plausible.

John Ternus, Senior Vice President of Hardware Engineering at Apple said of the M1 Ultra,
"We’re adding one last chip to the M1 family..."

 

[greenbreadmmm]

Agreed apple needs to start leading off with the pro models and trickle down. No reason entry level should be getting newest tech.

Also if this article is to believed m1x is the naming move. Why rush through the M naming conventions? At same time, yes apple named a14 and a15 even tho same nm. So the precedent is there.

 

[deconstruct60]

So apple would say at WWDC, End of Year, for our new Mac Pro, which means late Dec. Which means time for them to get the chip they want out?

Or Apple gets on stage and says the Pandemic and supply chain drama have been really, really tough and Mac Pro is a 2023 product and we'll be doing a "sneak peak" later in the Fall (i.e., Fall Mac dog and pony show). Throw out a "M3' teaser after doing a "M2" demo on a higher volume mac that will ship in Q3-Q4. Basically a theme of "wait a bit extra longer , it will be insanely great (and insanely expensive) ".

If they are shipping at end of December then it is more likely that the not using some variant of N4 for the Mac Pro (and just a pretty huge package. Different packaging tech than the Ultra used and probably even higher costs. )

 

[cb3]

Final variant: M1 Mordecai

 

[macsplusmacs]

Final variant: M1 Mordecai

He's right... apple is taking to long to name it. Let's do it ourselves:

M1 MOARPOWER

 

[Havoc035]

The M2 chip will apparently be the first Apple chip to make the jump to TSMC's 3nm process, skipping 4nm entirely. The M2 is believed to be Apple's first custom ARMv9 processor.

That would be pretty great. I am considering picking up an M2 Pro MBP and with ARMV9 at a 3 nm process it would be a chip that's ready for the future.

 

[jdb8167]

So ShrimpApplePro knows more about Apple's future plans than the Senior Vice President of Hardware Engineering at Apple? Sounds plausible.

Ah yes, Ray Tracing, that exclusive Nvidia feature...?

https://twitter.com/x/status/1530109888633765888

 

[deconstruct60]

Agreed apple needs to start leading off with the pro models and trickle down. No reason entry level should be getting newest tech.

It is substantively unlikely that the Mac Pro is going to iterate at the same rate as the lower end of the Mac line up. That hasn't happened at all over the last decade largely do to Apple self imposed effort prioritization. Being on M-series highly likely won't change that.

That doesn't mean the Mac Pro always will come in 'last' , but it also means that Mac Pro won't always come in 'first'.

For the shift to TSMC N3 the Mac Pro going first makes some sense. The timing of TSMC is way off the iPhone cadence ( high volume in May-June to get chips in July-August ) to ship in late September.). It is also off for October launch for replacements for the two year old initial M1 Macs.

Apple doesn't have to do a Mac Pro (and upper studio) SoC on every generation. Apple just did the Ultra in Q2 2022 ... doesn't need one for minimally a year or more. It too could skip a generation without lots of problems.

With the iPad Pro A_X series Apple skipped A11 , A13 , ( A10X -> A12X/Z -> M1 ) . Now that there is bigger than M1 stuff and more volume prouduct to put the "140mm^2" class die into, it can now iterate faster and the super large packages drift into the iterate at a slow pace and skip generations where no big fab process density uplift.

 

[Jumpthesnark]

So ... nobody outside of Apple really knows what's going on, we all just have to wait and see.

Put this on the home screen, it should be MR's motto.

 

[_Spinn_]

This doesn't really make sense to me. Apple specifically said that the M1 Ultra was the last M1 chip.

Since the M1 was based off the A14, I would expect the M2 to be based off the A16.

 

[Takeo]

What happens when we get to 0nm?

 

[0924487]

I think we are seeing the R&D speed limit of the Apple Silicon team.

 

[iamMacPerson]

So ShrimpApplePro knows more about Apple's future plans than the Senior Vice President of Hardware Engineering at Apple? Sounds plausible.

Yeah I think this ‘final’ M1 that is rumored will actually be the M2. Same process as the M1 with the updated cores from the A15. The M2 Pro, Max, and Ultra will be on the N3 process. Would make sense to let the lower end chip languish slightly, especially since it seems it will only be used in low-cost and/or lower-power applications (iPad Pro, MacBook Air, etc).

As far as A16, who knows. It’s so up in the air with China’s lockdowns and the chip shortages.

 

[dgdosen]

I would think this would be plausible because:

- Apple still hasn't released a Mac Pro - and the existing M1 architecture can't support > two dies (right?)
- Apple's M-based laptops evolve over time - and Apple can probably design systems that build in flexibility for environmental factors outside of their control (Covid delays/TSMC delays/LPDDR4-5 availability)
- We haven't seen any M chip based on A15, and that's been around for while, 3nm is around the corner... so why not? Why would Apple not take advantage of performance/efficiency gains vs having to sit and wait for M3 chips which would at least 12 months out from any M2 devices shipping.

I hope M2 is 3nm - and hope we find out at WWDC.

My thoughts - WWDC will show:
M2
M2 based Mac Pro
M2 based Macbook Pro
Maybe 'high end' Mac mini
(to officially close new Intel offerings on the two year anniversary)

 

[jdb8167]

What happens when we get to 0nm?

I can never tell if the person is serious when they ask this question. There is no 0 nm. The industry just starts measuring with a smaller scale. In this case Ångströms.

 

[Freeangel1]

I'll wait and buy an M2 Max Mac Studio. Should be a good time to jump to Mac studio.

I don't see Apple silicon getting more powerful by leaps and bounds after the M2 is released.