Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

CWallace

macrumors G5
Aug 17, 2007
12,528
11,543
Seattle, WA
You have to explain why Apple would be reluctant to release an M3 before an A17.

The simple answer is the M3 will almost certainly use the A17 as it's foundation (CPU cores, GPU cores, neural engine, etc.) just as M1 used A14 as it's foundation and how M2 will almost certainly use A15. So A17 will be designed either before M3 or in parallel with it.

As you noted, Macs are not (as of yet) wedded to the annual upgrade cycle that the iPhone is. A-series SoC design is probably finalized in Spring (to support a Summer production run for a Fall product release) and presuming that the M-series SoC is designed in parallel with the A-series, it is possible the M3 could go into Summer production in parallel with the A17 (as some rumors claimed A15 and M2 did this year) so in theory Apple could announce the M3 MacBook Air and iMac before they announce the A17 iPhone 14. But I am inclined to believe they will wait until after the iPhone 14 is launched.
 

jdb8167

macrumors 601
Nov 17, 2008
4,859
4,599
The simple answer is the M3 will almost certainly use the A17 as it's foundation (CPU cores, GPU cores, neural engine, etc.) just as M1 used A14 as it's foundation and how M2 will almost certainly use A15. So A17 will be designed either before M3 or in parallel with it.
You haven't explained why you think the A17 has to be designed as the foundation of the M3. Why can't the M3 be designed as the foundation for the A17? Ultimately if the two SoCs are going to use mostly the same CPU and GPU design, it really doesn't matter which one is manufactured first. The design is the design.
 

cmaier

Suspended
Original poster
Jul 25, 2007
25,405
33,474
California
The A series is the most-important SoC as iPhone is the most-important product. So unless Apple significantly diverges the M series from the A series with the next generation (something I do not see happening), A will continue to precede M so the A17 would be the first 3nm SoC used by Apple with the "M3" following later.

And yet M2 taped out before A15.
 

CWallace

macrumors G5
Aug 17, 2007
12,528
11,543
Seattle, WA
You haven't explained why you think the A17 has to be designed as the foundation of the M3.

Because A17 must ship in Fall 2023. M3 does not. So Apple will focus on A17 first and put the most development resources on it.

Yes, M3 can be designed in parallel (and likely will be), but M3 should be more complicated than A17 and take longer to design and validate prior to full-scale production.

And yet M2 taped out before A15.

I have seen rumors to that effect, but I don't recall any hard evidence this is the case.
 

reallynotnick

macrumors 65816
Oct 21, 2005
1,257
1,296
Is Apple skipping the A16, or why does everyone keep saying the M3 is based on the A17? Are we all just quoting a typo? Or is the M chips skipping a generation?
 

jdb8167

macrumors 601
Nov 17, 2008
4,859
4,599
Is Apple skipping the A16, or why does everyone keep saying the M3 is based on the A17? Are we all just quoting a typo? Or is the M chips skipping a generation?
It's based on this rumor that Apple will not have time to create a 3nm A16 because TSMC will only start full-scale production in the 2nd half of 2022. That means the first iPhone SoC on 3nm will be the A17 in September 2023. The M3 talk is purely speculation. If Apple releases an M2 on 4nm late this year or early next year then we might see an M3 on 3nm in late 2022 or early 2023.
 

CWallace

macrumors G5
Aug 17, 2007
12,528
11,543
Seattle, WA
Apple doesn’t announce their tape outs. So you’d need to hear it from someone working on the chips. Wonder where to find such people…

Yes I understand neither Apple nor TSMC will issue a PR when they start A or M series SoC production. ;)

There was a report from Nikkei Asia back in April saying the M2 had entered mass production and could be ready by July though, IMO, it makes more sense that said SoC is "M1X" and not "M2".



Is Apple skipping the A16, or why does everyone keep saying the M3 is based on the A17? Are we all just quoting a typo? Or is the M chips skipping a generation?

The 2023 iPhone SoC will be the A16, not the A17, so just replace my comments about "A17" with "A16".

M1 is based on A14
M2 will be based on A15, IMO
M3 will be based on A16, IMO
 

theorist9

macrumors 68040
May 28, 2015
3,882
3,061
It doesn’t matter which is most important. The schedule matters. Apple isn’t likely to change from their September iPhone release but Macs release schedules are much more flexible. You have to explain why Apple would be reluctant to release an M3 before an A17. From a consumer point of view, there isn’t much chance of confusion. Everyone would see an iPhone and a Mac SoC as completely separate product lines and everyone already expects a Mac to be faster than an iPhone. I don’t see any conflict.
Good point.

But there's something else besides schedule that also factors into this (and that, it would seem, could actually favor Macs getting chips made with a new process before the iPhone): Availability.

Apple needs a lot more iPhone chips than Mac chips. And new processes tend to have lower production rates to start. Thus, even if the manufacturing start dates were the same for, say, the next-gen iPhone and next-gen MacBook Pro, the MBP could get the new process before the iPhone.

This of course applies to other components as well. E.g., it's possible the expected sales volume of the higher-end AS MBP's will be sufficiently low (relatively speaking) that it will enable Apple to outfit them with LPDDR5 RAM, even if the availability of the latter is low.
 
Last edited:
  • Like
Reactions: the future

theorist9

macrumors 68040
May 28, 2015
3,882
3,061

So much Qu-ntel hurt in this post.
Don't see why. When Intel said it would catch up by 2025, they didn't mean they would catch up in 2025 with what TSMC is doing today, they meant with what they'll be doing in 2025, which of course assumes that TSMC will be much further along than they are now, and thus accounts for such advances. So there's nothing inconsistent between Intel's statement that they'll catch up and that TSMC will continue to advance further in the meantime.

What really raises questions about Intel's statement is not whether TSMC will continue to advance. That's expected. What raises questions is whether Intel *can* catch up.

One of the key techs that enables TSMC's advances is the extreme UV machines TSMC purchases from AMSL (in which Intel, TSMC, and Samsung co-invested heavily around 2012, to promote the development of smaller processes). For some reason, recently TSMC has been using these machines extensively but Intel has not. Now that Intel says it's going to start using them heavily, perhaps they can catch up.
 
Last edited:

deconstruct60

macrumors G5
Mar 10, 2009
12,493
4,053
I have difficulty reconciling this with Apple dropping 32 bit support and OpenGL with Catalina. Surely Apple would have special plugins for these types of customers if they were insistent on backwards compatibility.

Apple deprecated ( no new fixes, updates ) OpenGL ( and OpenCL). They didn't drop (de-support) them.
Apple is going to be quite hard pressed to completely drop OpenGL in the short term. Probably moderate intermediate term also. it will be in a comatose state, but since Apple put substantive effort into porting it over to M-series GPU, they probably aren't going to throw that away quickly as long as it is working ( and relatively very low cost to zombie walk forward). Apple's OpenGL layer calls Metal which does the work. As long as Metal doesn't change to impede that handoff of work.. then Apple is probably not in a hurry ( presuming there are no major bugs and/or work to do to keep that handoff working.)


32-bit is far more crictial since that completely dumped that from the M-series. That is a show stopper at the hardware level; not the software one. The side-effect of making Rosetta 2 easier to do (lowered complexity) helped also. (thinned out apple libraries also when going to 'fat' libraries of x86 and M-series. Just makes those builds much more homogenous in bit width support also. )






Then that sort of defeats the purpose of a socketable cpu if it can’t be upgraded to successive gens. Not that that’s entirely Apple’s fault, Intel is notorious for that.

Likewise, over in the dark realm of the Mac Pro forums, many decried the 7,1 for it’s price to performance ratio. I don’t think they would clamor for an ice lake upgrade.

errr. probably would go over OK in the Mac Pro forum. Lots and lots of folks talking about how Mac Pro 5,1 can be cheaply upgraded to 12 core with used CPU packages that originally cost lots of money. A substantially large portion of the modularity is king crowd is all about how to buy the parts cheaper later in time ( and not paying bleeding edge release time costs ).

For that crowd, it doesn't help that the W-3200 isn't pragmatically the same as the Xeon SP CPU packages of that iteration. The PCI-e lane provisioning is different ( W-3200 has 64 lanes , the SP ones don't. And Apple uses all 64 lanes in the Mac Pro 2019. ). There is no gap on W-3300 and Xeon SP Gen 3 on PCI-e lane provisioning. Long term there are more e-wasted/retired Gen 3 processors that could be used.

Even non long timers, but still folks wanting to duck the 20+ % Apple tax applied to Intel CPU prices. Buy 8 cores and then avoid the Apple tax to go to 28 cores. [ All the more egregious if don't need > 1 TB RAM since don't have to pay the approximately $3-4K "> 1 TB" tax that Apple and Intel apply to those. ]

Intel is way, way behind on core count so they are playing catch up and differentiate on I/O. W-3200 is PCI-e v3, W-3300 is PCI-e v4 , and W-3400 is very likely to be PCI-e v5. That is not conducive to using the same socket for all three.


And Apple Silicon is poised to beat the pants off Intel in 2022. It seems silly to have your lower end laptops and desktops walk circles around your highest end machine.

Single threaded. Max possible Multithreaded with no system thermal limits? No, the M1 or M2 probably aren't going to win there. Even Apple larger die probably isn't going to win either. First, Apple is deeply committed to iGPU. The die space that the iGPU consumes is just that more CPU cores AMD/Intel can lay down on a substantially larger package. Second, macOS is capped at 64 threads/cores anyway. They don't have software foundation to engage in "core count wars". Third, how are they going to feed the cores.. still likely stuck on PCI-e v3 and sub RAM capacity in 2022 also. ( Apple Silicon Mac Pro model is all about have few slots and less volume. That seriously doesn't scream top end aggregate bandwidth capabilities. The iMac 24" is thinned out so much the Ethernet jack is in the power cable because it can't fit. )

I can’t honestly justify such a product in my mind unless Apple Silicon isn’t as scalable as we thought (which is possible). To me it reads like Apple is not confident in their new architecture. Which definitely puts a damper on my enthusiasm for the new architecture.

It probably isn't about being confident. It is more likely about where the anchor point of the design is ... around mobile. Apple sells 3x as many laptops as desktops. That is where the focus is. Throw iPads into the mix it is about double that ratio ( just iPad Pros is probably pushing on 4x multiple )

If Apple nukes DIMMs in the rest of the desktop line up then the Mac Pro SoC is going to be in an oddball space with relatively very small volume attached to it. Apple probably won't do something just for that space by itself without passing along much of the same limitations as the rest of the product line. Security and SSD controller is built into the SoC so it highly likely isn't going to be socketable in the conventional sense.
 
  • Like
Reactions: michalm

macsplusmacs

macrumors 68030
Nov 23, 2014
2,763
13,275
Don't see why. When Intel said it would catch up by 2025, they didn't mean they would catch up in 2025 with what TSMC is doing today, they meant with what they'll be doing in 2025, which of course assumes that TSMC will be much further along than they are now, and thus accounts for such advances. So there's nothing inconsistent between Intel's statement that they'll catch up and that TMSC will continue to advance further in the meantime.

What really raises questions about Intel's statement is not whether TMSC will continue to advance. That's expected. What raises questions is whether Intel *can* catch up.

One of the key techs that enables TMSC's advances is the extreme UV machines TMSC purchases from AMSL (in which Intel, TMSC, and Samsung co-invested heavily around 2012, to promote the development of smaller processes). For some reason, recently TMSC has been using these machines extensively but Intel has not. Now that Intel says it's going to start using them heavily, perhaps they can catch up.

I was mostly joking. If intel gets ARM like fabs up soon they will be ok.
 

deconstruct60

macrumors G5
Mar 10, 2009
12,493
4,053
I believe the Apple Silicon Mac Pro model will have PCIe slots, just fewer. My guess would be four (so half of what MP 7,1 can hold) so that it can handle up to two MPX cards or an MPX card and two single-slot cards.

That is basically a re-defintion of a MPX bay. Full sized MPX modules are 4 slots wide. And there are two sockets per bay and one MPX connector. on the first slot of the bay.

You're pointing at half size MPX modules as being the "norm". That is a huge stretch. Of the five MPX modules Apple has offered so far only two have been half sized ones ( 580X and 5500). One could hand wave and say that they'll be Apple only GPUs and only will need two "height" cooling and still be top end market competitive. Thown in the stoarge MPX modules and the ratio of half : full height modules shrinks even more lopsided.

The MPX connector in the bay is doing bandwidth sharing with the second slot in the bay. Unless doing major upgrades to the M-series SoC can't both share the bandwidth on the second MPX bay slot and use the slot for top end GPU communication bandwidth. That is why "half sized" MPX modules don't come with Thunderbolt ports. They don't pull Thunderbolt provisioning data on the MPX connector. (i.e., if MPX modules 'uncovers' the second slot, it doesn't pull data from the second slot indirectly). You proposal is that Apple completely change that role.


IMHO, I think it is a stretch to think that the M-series SoC used in the Mac Pro is even going to provision more than PCI-e v3. Even if Apple did upgrade to PCI-e v4 in the second generation , likely not going to get that many lanes. The slot cut is more likely driven because the aggregate I/O bandwdith just isn't there. Something like:

One MPX Bay ( two slot, quad wide , one aux power connector. )
One x16 electrical slot single width ( one Afterburner card only bus )
On x4 semicustom I/O card. (single width )

If lucky maybe one more x8 card single with ( current slot 6 ) , but decent chance this bandwidth get consumed by dual 10GbE , some discrete SATA controller for at least same two ports as have now (maybe give out 3 (or 4 ) and allow for incrementally taller bracket). WiFi/Bluetooth and misc.


Apple spent a mint to bring Mac Pro 7,1 to market. Customers then spent a mint to purchase them. Neither Apple nor the customers are going to want MP 7,1 to be a "one and done" model like MP 6,1 was and Apple clearly engineered MP 7,1 with sufficient overhead to take successor generations of CPUs and GPUs.

Apple probably hit break even on Mac Pro 2019 development costs by June of this year (if not sooner). There is a 20+ % markup on every screw, wire, washer, NAND chip , etc. in that system with a "low volume " tax on the overall system. Apple picked the most expensive > 16 core options they could and taxed those high markups also. The longer it takes an W-3300 Ice Lake to come out ( e.g., slides to 2022 ) the more likely Apple has hit breakeven.

There are no other substantive "loss leaders" in the rest of the Mac line up . So not like DEll/HP/Lenovo where have to use workstation profits to cover losses in other parts of the business. If don't have to "rob Peter to pay Paul" the path to breakeven is much quicker because can use the full profits to pay off the upfront costs.

It is more like it is relative cheap for Apple to do an W-3300 update to a W-3200 system. Keep the same already paid for chassis design. Keep the same fans and cooling set up that is already paid for. Cheap. Same DIMMS slots with a slightly different traces back to the CPU. relatively cheap. Same 68 PCI-e lanes coming out of the socket. Relatively cheap ( need some re-timers for PCI-e v4 run over a distance , but not a moon shot project). If they only care about increasing slot 1 and slot 3 to PCI-e v4, then they could keep same PLEX PCI-e swtich (and rest of the slots all stay the same). Cheap, pratically already paid for R&D. Same power supply. Paid for.

Apple is probably doing this not because they are deep in the hole in R&D cost recovering. It is likely because it is vastly cheaper than coming up with a relative super low volume SoC that can actually compete with W-3300 level aggregate bandwidth. Apple can skip that since they don't have it (and probably don't want it.) . More likely Apple wants to keep the iGPU feed with data and trading off more higher aggregate memory I/O bandwidth for lower more generic PCI-e aggregate bandwidth (and power consumption) out of the SoC.

Also, decent chance Apple started on a W-3300 iMac Pro and pulled the plug in it when final W-3300 thermals blew past limits and slid out from 2020 to 2022. Again this is chance to re-use initial work already paid for ( or in this case already invested in).

Releasing a half sized , M-series Mac Pro will probably cut into profits that the full sized Mac Pro generates. Only deep slot count lovers and higher core count with high RAM working set footprint folks will be drawn to it. But if mostly already paid for foundation then breakeven doesn't need as many even higher priced units to sell . ( plus redoing the system allows Apple to do a face saving price cut so that the systems are more competitive. W-3300 will have incrementally lower prices. And substantively new lower prices at the top end where Intel dropped the extremely heavy "> 1TB " RAM tax . )



The only drawback for existing MP 7,1 owners is that W-3300 will use a new socket and therefore will require a new systemboard so a "drop-in" CPU upgrade will not be possible.

If the MP 8,1 can easily take the $5K sunk cost Vega MPX modules that's just fine. Some folks spend far more money on what goes in the PCI-e slots than what goes in the CPU socket. that is somewhat reflexive also. New 6000 series MPX that costs $2.5K is probably bigger draw for many 7.1 owners . The CPU socket is secondary as it isn't the computational chokepoint.

The W-3300 is yet another "dead ender" CPU socket. The "I gotta get more than one CPU generation in my socket" folks aren't going to be entirely happy with either one.

Both the W-3200 and W-3300 will get cheaper over an extended period of time to "drop in" higher core counts later. The W-3300 can probalby take old Xeon SP gen 3 also as they fill up the used sales in several years time.
 

CWallace

macrumors G5
Aug 17, 2007
12,528
11,543
Seattle, WA
That is basically a re-defintion of a MPX bay. Full sized MPX modules are 4 slots wide. And there are two sockets per bay and one MPX connector. on the first slot of the bay.

I'm not familiar with MPX. I was just going off the Apple support site that said an MPX card required two paired slots.

So okay, four PCIe slots for one MPX card. :)
 

deconstruct60

macrumors G5
Mar 10, 2009
12,493
4,053
The simple answer is the M3 will almost certainly use the A17 as it's foundation (CPU cores, GPU cores, neural engine, etc.) just as M1 used A14 as it's foundation and how M2 will almost certainly use A15. So A17 will be designed either before M3 or in parallel with it.

There is a small exception there in that Moore's law is that fab process densities increase every 18 months ( well 18-24 months ). The "problem" is that in the vast majority of times it is not a clean multiple of 12 months. The Mac isn't pragmatically, rigidly bound to exactly 12 month release cycle. So it is going to be periodically easier to adapt the Mac release schedule to TSMC roll out schedule. The iPhones solidly non-technical driven adherence to September is going to be a harder fit. Two different period widths are going to generate mismatches.

There is some substantive pressure for TSMC to sync up their rollouts to facility volume ramping in May-July time frame so can prime-the-pump peak in July-September , but they won't be able to line that up cleanly over a 3-5 year period. Some years they are likely to drift.

What is probably more true is that Apple will not sacrifice the June-September run up on new iPhone SoC volume to product Mac SoCs. That is highly unlikely to happen. Where Apple is constrained on wafer starts the Mac will very likely loose. And yes periodically Apple will probably be constrained on wafers. "Apple is super rich so they can get whatever they want" .... rich doesn't make the Sun come up and go down every day. There is some stuff just can't buy. As long as TSMC churning out at 100% there are periodically going to be no excess slack wafer starts to buy. Not an "outbid" , "got more money" contest. Just not going to be there.

Apple might sometimes need some Summer build up for a October release and start a relatively low rate build up earlier ( e.g. May -June ) and run a limited subset along side the A-series volume. Anything production glitches though are probably going to the A-series though. So very low probability of major Mac launces in August , September. substantively low probability in July. Low in June and most of October.
 

deconstruct60

macrumors G5
Mar 10, 2009
12,493
4,053
I'm not familiar with MPX. I was just going off the Apple support site that said an MPX card required two paired slots.

The MPX connect is line-up up behind the standard PCI-e socket relative to the outer edge of the motherboard. The MPX modules have two connectors on the bottom. One slides into the standard PCI-e slot. That is the data and partial power connection for the GPU inputs. The MPX connector handles three things. Primary power delivery for the board. DisplayPort output to the internal host thunderbolt controllers. PCI-e data for the thunderbolt controolers (if present on the card). The first two are always used on all the GPU MPX modules.


macos-catalina-system-info-pci-cards.png



The MPX modules also have "height" ( the up/down dimension in a vertically placed Mac Pro). The above diagram is only a "logical placement" diagram. Slots 1, 2, 3, and 4 are all double width wide. They are double wide but there is no physical slot in that second half of that "double". They are meant to take tall cards.

The issue is that full sized MPX modules are relatively very tall because there cooling is effectively integrated with the case fans. A full size MPX module covers either slots 1&2 or 3&4. Slots 1&2 are in MPX bay 1 and 3&4 are in MPX bay 2. Each bay also has an AUX power connector.


Apple could get to a "half size " ( Half tall) Mac Pro enclosure by getting rid of MPX bay 1 (and its associated fan) and shortening the feet (since would be a desktop only tower) and lowering the handles. The additional problem is that the Mac Pro motherboard is dual sided. There is important stuff on the back. So if take away about 1/3 of the back some of that stuff has to move. PLex PCI controller, power converters, speaker , SSD data blades, RAM DIMMs, and backside cooler (cooler covers power supply needs also. So can't dump it completely because Mac SoC is "cooler"). Decent chance Apple is going to solder the RAM to the SoC so that would take care of the DIMMS and put that functionality on the front of the board . Move the rest up still on the back side. But that is much bigger SoC package if they max out on cores ( e.g., 32 + 'big' iGPU ). Even if don't solder RAM would still need to bring something backside to the front. So probably need to move slot 8 ( I/O card ) down. ( that would wipe out slot 7). Finally , as I outlined, they probably need to add some stuff not on the standard SoC like SATA and multiple 10GbE outputs so.... probably loose slot 6 also to fixed, soldered on, discrete controllers.

Because it will probably have a more than decent iGPU folks who want to things like use 4 audio cards could still get by because no GPUs would fill any slot. 1-2 Afterburner cards with the iGPU. A GPU MPX with a M.2 4 drive carrier card . etc. etc. Lots of permutations available.
 

deconstruct60

macrumors G5
Mar 10, 2009
12,493
4,053
Both?

New core design.

If the M2 and A15 both has the same basic core design there is really no material difference signified in the tape out order. the core design has A15 design parameter aspects heavily influcing the design.

The 'uncore' parts being different and differing amounts of core counts would still have the baseline foundation anchored firmly on the A-series evolution.

As alluded to in other post. A-series has a "problem" in that fab rollouts don't always go to high volume production in May-July timeframes. A Mac SoC getting an occasional early , relatively low volume start isn't not a sufficiently clear indicator that priortities have wildly shifted off of iPhone Soc (i.e. what drives more than 65% of Apple's profits. Service revenue off of iPhone clients is driven by iPhone SoCs. ).
 

deconstruct60

macrumors G5
Mar 10, 2009
12,493
4,053
Unless Apple changes the Darwin kernel to support more than 64 logical CPUs, it's unlikely Apple will use any Hyperthreading Intel CPU with more than 32 cores/64 threads.

Or ..... just turn the hyper threading off on the > 32 core chips. Or just stop at 32 core ( so only really pasing on 38 core chip that is likely super-duper high priced anyway. ). However, yeah could be a blocker for the folks who later try to jam a cheaper because it is old 40 core Xeon SP 3rd Gen in there.

For the hyper sensitive real time lime latency folks , some of them may just want 'real cores' (and not competing for function units and time slices from a "virtual" thread).

But yeah, another one of those things that won't impact M-series because doesn't have SMT at all. But also indiciative that M-series won't be as good at workloads that get heavy traction from SMT. So folks believing that Apple is going to fork the kernel scheduler just be benifit a Intel CPU SoC over a Mac M-series SoC are probably smoking something. ( probably not going to happen. Likely nobody at Apple is jumping out of bed in the morning to make that their top priority.... more advantage to x86 in the kernel )


If the M-series caps out at 32 P cores then maxing on would simplier Apple's marketing message too. Need 32 "old school' x86 cores and orders of magnitude better I/O aggregate bandwidth.... buy taller, deskside , noiser , heavier , more expensive Mac Pro. Want desktop with 32 cores of Apple slickon and have far more modest I/O aggregate bandwidth... buy this quieter , more affordable , desktop "half sized" Mac Pro.

One model for the "hardcore" modularity holdouts and new model so folks who are primraily just CPU core count bound , but periodically need top end single threaded.

Both won't need any deep kernel changes.
 

thingstoponder

macrumors 6502a
Oct 23, 2014
916
1,100
Both?

New core design.
By successor I mean something that will go in M1 devices. If its a direct M1 successor I don't think that timeline makes any sense. I don't expect those products until next year at the earliest. New MacBook Air is rumored for March I think and the iMac and iPad Pro are still new devices. a15 iPhone is right around the corner and will come before any M1 device is refreshed. If it's the chip we heard about entering production for the new MacBook Pros that makes sense.
 

cmaier

Suspended
Original poster
Jul 25, 2007
25,405
33,474
California
By successor I mean something that will go in M1 devices. If its a direct M1 successor I don't think that timeline makes any sense. I don't expect those products until next year at the earliest. New MacBook Air is rumored for March I think and the iMac and iPad Pro are still new devices. a15 iPhone is right around the corner and will come before any M1 device is refreshed. If it's the chip we heard about entering production for the new MacBook Pros that makes sense.

The new MacBook Pro coming in a couple months is not an M1 product.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.