Apple Testing at Least Nine New Macs With Four Different M2 Chip Variants

[deconstruct60]

If the M2 is based on the A15 (which as the months pass, is becoming less likely — the A16 is already knocking on the door), and if they can't increase its clock much, and if Raptor Lake is more or less on schedule, yeah.

Gen 13 (Raptor Lake's ) schedule shouldn't be much in doubt. It is also on Intel 7 ( with making a narrow fab tweak that between 14nm++ and 14nm+++ ). The E cores ( Gracemount ) are the same (just more of them ... i.,e., incrementally bigger dies. ). The iGPU is the same. The memory controller appears to be the same . All of that is probably over 50% of the chip. If not changing much then going to be pretty hard to miss the schedule... as it is mostly the same thing. Converting from one die to another is a tasks that Intel really has not failed at. Iterating on the same fab node is something they have done on schedule over last 5 years.

Intel has already run a mostly-public demo

https://videocardz.com/newz/intel-s...-raptor-lake-cpu-with-24-cores-and-32-threads


The P cores are different. That really isn't going to be hard either. The Alder Lake P cores are mostly unmodified Xeon SP Gen 4 ( Sapphire Lake) cores. ( some L2/L3 adjustments but primarily the same thing; AVX-512 et al at the individual core subcomponent level. uncore memory controller is different. ). If Intel made some deliberate adjustments for a desktop/laptop tuned core that could be a substantive contributor to the bump and really wouldn't be a major schedule killer in terms of difficulty.

Intel isn't going to ship the whole Gen 13 line up. They didn't ship the whole Gen 12 line up in Fall 2022 either. Probably going to get some cherry picked top end benchmark 'targeted' SKUs to start to get some tech porn press buzz. And mobiles in 2023.

Intel is getting back on track by not trying to make huge jumps on a single iteration.

[ Rocket Lake was probably an unnecessary distraction. desktop/laptop Golden Cove probably didn't get the work done on it that it should have. Raptor Lake is just catching up. That isn't something that would blow a schedule.

 

[CWallace]

None of which answer the question I posed...

Fair enough.

My opinion is that the 12 core Pro and Max SoCs will be 10 performance and 2 efficiency cores. So the 24 core Ultra will be 20 performance and 4 efficiency cores. And the 48 core "Extreme" will be 40 performance and 8 efficiency cores.

 

[KurtMann]

I am on a Macbook Pro from 2013. My battery has had 905 cycles to date. She has worked flawlessly, I can still watch a movie and more on battery alone. I am waiting for MbP 14 M2 something, and will definitely buy one if gas, electricity and food prices don't ruin me!

 

[Nermal]

Not in the G4, G5 days. That was annual.

The G4 timeline in particular was often a bit quicker than annual.

Original G4: August 1999
Gigabit Ethernet: July 2000 (11 months)
Digital Audio: January 2001 (6 months)
Quicksilver: July 2001 (6 months)
Quicksilver 2002: January 2002 (6 months)
Mirrored Drive Doors: August 2002 (7 months)
MDD 2003: January 2003 (5 months)

The G5 models were a little more spaced apart, but again you often didn't have to wait a year between them (e.g. the Late 2004 model was only four months after the previous one).

 

[jerryk]

I hope the Mini is first. I made a mistake ordering mine with just 8 GB. I will immediately correct this if a M2 Mini comes out.

 

[EugW]

I hope the Mini is first. I made a mistake ordering mine with just 8 GB. I will immediately correct this if a M2 Mini comes out.

Makes sense. Im just curious though. What kind of stuff do you do with it?

 

[vanylapep]

Ordered the M1 MBP 14. Will wait for the M10.

 

[Mr. Dee]

I they don't ship a new Mac Pro by this year, then they will have missed their announced goal of completing the Apple Silicon transition within 2 years. Of course that is a self-imposed deadline but they will likely feel that they need to meet that goal unless there is some strong reason they cannot. It will look like a miss if they don't. It doesn't matter that most Mac customers will never even see, let alone buy a Mac Pro. It is a halo product that shows off their expertise and to have it come out late will be a black mark on it.

At the same time, its not impossible for them to launch it at WWDC. There are two scenarios here, the Mac Studio is the new stop gap until the Mac Pro is ready (just like the iMac Pro was) or Apple was concurrently developing both the Studio and the Pro. It just make sense that they would since the Pro is gonna be based on a variant of the M1 Ultra.

I guess Apple is planning to make the entry level of the Pro just like the Intel Mac Pro. Which is why I go back to the point this is a niche product and Apple is in no rush to bring it to market since its not high volume.

 

[jerryk]

Makes sense. Im just curious though. What kind of stuff do you do with it?

Software development, browsing, email, and monitoring services for my home solar/battery system.

 

[420ravens]

We do have one more Mac desktop and one more Mac laptop expected in the near term based on the filings with the Eurasian Economic Commission.

The most logical desktop to me would be an M2 Mac mini. The most logical laptop to me would be an M2 MacBook Air or M2 MacBook Pro 13.3". Both could launch at WWDC or soon after.

I do expect Apple to announce the Apple Silicon Mac Pro at WWDC, but not give a release date because I could see it slip into 2023.

I do not believe we will see an M1 Pro Mac mini configuration. I'm not sure we'll see an M2 Pro Mac mini configuration, but if we do, my thinking is 2023. Same with an M2 Pro and M2 Max MacBook Pro. And the M2 iMac 24" (with possibly an M2 Pro configuration if Apple does not release an Apple Silicon version of the iMac 5K 27").

This. M1 Pro chip is very similar to M1 Max chip. Can't put M1 Pro chip in a small chassis Mac Mini and charge a lot less than Mac Studio, when it'll benchmark 90% the same. Mac Studio is clear delineated platform at much higher price, hoping for a mini with similar performance for much lower price- don't think so. On logic, I agree 100% with all this CWallace. But not everything is logical... who knows. Also think M2 will have smaller gains because M1 was already such a step change.

 

[mabhatter]

Apple updates their iOS devices every year. They obviously are going to update their Mac hardware every year as well now that they're on their own chips.

Chip design gets more expensive for the higher end chips. I think we'll see 18-24 month cycles.

The Air was about 16 months. Upgrading to M2 for Air and iPad Pro in the fall would be two years of M1. There are basically 4 groups of products now. iPhones & iPads (A-series), M1 macs and iPads, MacBook Pro M1Pro/Max, and M1 Ultra Studio. That will fall into a nice six month cadence of something new.

 

[R3k]

I’d say wait for the M3 ultra. I’m told it will be a huge performance gain over the Celeron.

 

[mabhatter]

It makes perfect sense for the Mac Mini to have M2 and M2 Pro, and the Mac Studio to have M2 Max and Ultra.

The other models make sense as well.

The LARGE remaining question is what the Mac Pro will bring. Do we start with 1 M2 Ultra and scale up to 8? M2U x1, 2, 4, 6, or 8?

Do they include expandability with Modules containing CPU’s, GPU’s, RAM, all the other stuff like Neural Engine and on board codec stuff?

Very interested to se the precedent set with the Mac Pro.

I think we can stop worrying about the Mac Pro. That's on like a 2-3 year cycle and it's never been a consumer or even prosumer product.

The main problem with the Mini is that it skews the SKUs too much. The M1 mini holds the fort at $600. An M1 Pro mini would have to be $1000-$1200 and if you put any upgrades on it you might as well buy a Studio. If anything the 24" iMac should get the M1 Pro in its current shape then it would be a clear difference between lineups of Mini, iMac, and Studio.

 

[code-m]

I am not sure if it’s worth upgrading from my M1 MBA, it runs like a champ and is dead silent and cool to the touch most of the time. Depending on support this maybe my last Mac laptop for quite sometime.

 

[EugW]

I think we can stop worrying about the Mac Pro. That's on like a 2-3 year cycle and it's never been a consumer or even prosumer product.

The Mac Pro is very interesting to me even though I'm not a target customer and I'm not going to buy one. It's interesting because it will utilize a high end 3D and scientific workstation SoC, designed by Apple. What we see in this machine is going to set the stage of what we will be expecting from Apple's engineering teams in the future.

The main problem with the Mini is that it skews the SKUs too much. The M1 mini holds the fort at $600. An M1 Pro mini would have to be $1000-$1200 and if you put any upgrades on it you might as well buy a Studio. If anything the 24" iMac should get the M1 Pro in its current shape then it would be a clear difference between lineups of Mini, iMac, and Studio.

This paragraph doesn't make much sense.

Not everyone wants 32 GB, but some people need CPU speed. Not everyone wants CPU speed, but wants more memory. In this middle ground there is a big price gap and the claim that one can just jump up to Mac Studio is to completely ignore that most of the world actually has a budget they need to work with.

The absolute cheapest Mac Studio is $1999. That is a lot of scratch. That means that even if you upgraded a high end Mac mini to $1499, the Mac Studio would still cost 30% more. Meanwhile there are others that would be satisfied with say a 16 GB $1299 Mac mini with M1 Pro, meaning the Mac Studio would be a whopping 54% more. The Mac Studio is simply not viable for a good chunk of budgets.

As for the iMac, not everyone wants an iMac either, especially when it's limited to a 24" screen. Or people may already have a screen for their previous Mac mini or whatever, so the integrated screen of the iMac is superfluous. I know in my case I already have a 5K iMac, and I have no desire to downgrade to a 24" iMac. However, I'm not replacing that machine anyway. The machine I'm replacing is a Mac mini paired to a 30" screen. The 24" iMac has no place in this setup.

 

[Bandaman]

Would be ridiculous to see new 14" and 16" M2 MacBook Pros this year. I just got the 16" M1 Max.

Yes, products should only be released around the timeframe that you purchase them.

 

[deconstruct60]

No conflict. Mac Pro announcement or sneak peek in June, but no M2 Ultra in 2022.

Why? Cuz Mac Pro likely won't have M anything. My prediction is that it will have an SoC that uses a different naming convention, since the SoC design will be quite different.

The looks more like a "move the goal posts" and wishful thinking than something likely to get matched to a "half sized" Mac Pro.

A Mac Pro , Mac Mini , Mac Studio. ... probably are going to have the same "prefix" on their SoC name; because they are all Macs. Apple is likely not going to strip the Mac Pro of the now common Mac boot protocols , built in security enclave , built in SSD controller , built in Thunderbolt controllers , etc.
A more coherent path for them is to perhaps add another suffix, but share the same prefix

M2 Ultra
M2 Ultra XL ( add a PCI-e controller to the package)

two easy paths to add a PCI-c controller(s). First, Don't use identical pairs. ( one die drop the redundant stuff ( extra secure enclave , extra SSD controller , unused TB port(s) , ). Second, leverage a subset of the UltraFusion bus to connect a PCI-e controller to the package.

If the UltraFusion bus can handle keeping 64 GPU cores as a unified whole then two x16 PCI-e v4 bus traffic probably won't change much. There is no huge deep seated technical need for a "quite different" implementation. Apple probably spent 10's of millions developing UltraFusion and now going to abandoned it now for the Mac Pro? Probably not.

Doing a substantial fork from the M-series for a product with relatively very, very low market share makes little economic sense. In 2017, Apple said the Mac Pro units sales was down in the single digits percentage of Mac market. Apple then in 2019 increased the entry price for a Mac Pro 100%. Extremely likely that did not increase the overall Mac market share. They also said in the recent Studio introduction the Ultra version covered the performance of the most popular CPU configuration for Mac Pro and more popular GPU configuration. Again going forward that market share is highly likely smaller than 2017.


If there were 2-3M Mac Pro sales per year then maybe Apple would do a forked SoC. However, that is probably at least an order of magnitude higher than what they actually are. So they probably won't.


You can go back to the lead post of this whole thread.

"... reports Bloomberg, citing developer logs. There are "at least" nine new Macs in development that use four different M2 chips that are successors to the current M1 chips. ....
...

• A Mac Pro that will include a successor to the M1 Ultra used in the Mac Studio.

... "

If there is already developer telemetry sending back "successor to M1 Ultra" data then probably has the same Mn prefix to the SoC ( something that is a successor to 'M1' ... like M and some number bigger than one . )

Apple could get to a substantively different looking "Ultra" successor by jumping to M3 (TSMC N3 ) and making the Ultra a monolithic die. ( then they would only have to pair up two to make the biggest variant. ). But that wouldn't be a major fork off of the baseline SoC design approach. However, it would be a desktop only die (shared only between Studio and Mac Pro .... and maybe an large chassis iMac if Apple jumps back there later on. )

But again if the aggregate Ultra die rates aren't high enough. ... going to find "Ultras" made out of Maxes to get the overall Max die run rate high enough for Apple to bother with it.

 

[EugW]

@deconstruct60,

My statement remains unchanged. No M2 Ultra in 2022, but the Mac Pro gets announced in 2022, likely within 2 months, at WWDC. And I don’t think its chip will be called M anything.

I could very well be wrong, but no goal posts have been moved by me. If I’m wrong then I’m wrong, but if anything, it would be moving the goal posts on your part to declare an M2 Ultra release just because a different chip that shares a significant part of the architecture and design gets released.

 

[deconstruct60]

What if the entire M2 family is:

• M2
• M2 Pro

And the entire M2 Mac family released is:

• M2 13" MacBook
• M2 24" iMac
• M2 Mac mini
• M2 Pro Mac mini [Pro]

and iPad Pro M2

All released before the end of 2022, then Apple pivots to the M3 line-up for 2023/24, which would be a complete line-up; meaning the aforementioned models in M3 form, along with the MacBook Pro laptops & the Mac Studio desktops...?

There .. something like that. (Maybe a M2 Max (no Ultrafusion ) thrown in so can do MBP 14"/16" updates in early 2023 ). Skip all of the 2C ( and 4C) package validations because this is a "short iteration" generation. Primarily trying to get the M1 systems to retire by 2H 2023 (and wipe the initial batch in 2022).

Apple does the M3 in the reverse order. Early 2023. M3 Extreme (Mac Pro). , Mid 2023 M3 Ultra ( Mac Pro , Studio ) , early Q3 2023 M3 Pro ( Studio ). .... etc.

The very high margin M3 Extreme would 'pipe clean' the TSMC N3 process. As opposed to using a super high volume, relatively low die size like the A17. As get closer to the die sizes and volumes in the 'middle' .... those are not good candidates.

The relatively low volume, high end desktop stuff only 'leads' when the timing fits ( full node shrink available and also due for an update). The smallest die can lead more opportunistically (incremental/'half' or die shrinks and also due for upgrade ).

Longer term Apple is likely going to be weaving in their celluar modems into their package up mix process and that will periodically put the primarily laptop/'iPad Pro' dies on a different tempo also.

 

[MayaUser]

• A MacBook Air with an M2 chip that features an 8-core CPU and 10-core GPU.
• I thought people trusted that twitter account of " KUO " that said M1 for the macbook air redesign ))

 

[norwaypianoman]

c") The naughty planners in the planning department is planning 27 iterations of new macs in the coming 3 years.

Naughty, naughty...

 

[Random_Matt]

Wow, people think M2 Max will be significantly faster than the M1 Max, please. Don't buy anything if the fear is too great of whatever gets released afterwards.

 

[Mega ST]

Just think about the M3 ultimo. Or the M4 omega and then the M5 excelsior.

 

[theluggage]

My statement remains unchanged. No M2 Ultra in 2022, but the Mac Pro gets announced in 2022, likely within 2 months, at WWDC. And I don’t think its chip will be called M anything.

I think Apple pretty much have to announce the Mac Pro at WWDC, even if it is just a preview. They’re not obliged to “complete the transition in (exactly) 2 years” but it would start to look like carelessness otherwise, and the target users really need some sort of roadmap - £20,000 workstations are not “impulse purchases”.

There are several ways the Mac Pro could go - one is some sort of scalable system using multiple Mx SoCs using something like NUMA, which could conceivably launch with M1 Ultra, without Apple technically making another M1 variant. My money is still on this idea because of the economies of scale vs. making a whole new SoC -or even die- for Apple’s smallest selling system.

Then there’s the 4xM1 Max idea, which was a pretty strong rumour for a while, although it seems to have fallen out of favour now people have actually got a look at the M1 Ultra. Apart from anything else, chip design technicalities aside, I’m getting a strong “diminishing returns” vibe from the M1 Ultra reviews, so a 4x variant may be a solution looking for a problem.

If you accept Apple’s own claims about the Studio Ultra then it already has Mac Pro level processing grunt, then the distinguishing features of the Mac Pro come down to PCIe expansion and RAM capacity. That might require a custom SoC with more PCIe and maybe an external RAM bus, in which case, as you say, Apple could call it something other than M1/M2 (which lets them have a longer lifecycle without sounding obsolete). That said, either a quad Mx or M1 Ultra cluster would be packing a hell of a lot of Thunderbolt ports which could be turned back into PCIe (if they aren’t already configurable that way).

I think launching a M2 Ultra Mac Pro any time soon would really, really tick off some people who have bought Studio Ultras, as it would be seen as replacing the M1 Ultra. Plus, I suspect the higher-selling MacBook range would be first in line for M2 whatever chips.

 

[EugW]

I think Apple pretty much have to announce the Mac Pro at WWDC, even if it is just a preview. They’re not obliged to “complete the transition in (exactly) 2 years” but it would start to look like carelessness otherwise, and the target users really need some sort of roadmap - £20,000 workstations are not “impulse purchases”.

There are several ways the Mac Pro could go - one is some sort of scalable system using multiple Mx SoCs using something like NUMA, which could conceivably launch with M1 Ultra, without Apple technically making another M1 variant. My money is still on this idea because of the economies of scale vs. making a whole new SoC -or even die- for Apple’s smallest selling system.

I was under the impression that macOS does not support NUMA at all. If not, they could always add it, but while I'm not a coder, I am thinking it would not be simple to include it all of a sudden in one OS update. And even if they did, all Mac Pro owners would essentially be macOS beta testers for a core kernel feature that didn't exist before. Correct me if I'm wrong about this.

I suppose it could already be there, just not exposed, but now we're reaching.

Then there’s the 4xM1 Max idea, which was a pretty strong rumour for a while, although it seems to have fallen out of favour now people have actually got a look at the M1 Ultra. Apart from anything else, chip design technicalities aside, I’m getting a strong “diminishing returns” vibe from the M1 Ultra reviews, so a 4x variant may be a solution looking for a problem.

Yes I agree, both these factors are important, SoC design limitations and diminishing returns in actual practice with this approach.

If you accept Apple’s own claims about the Studio Ultra then it already has Mac Pro level processing grunt, then the distinguishing features of the Mac Pro come down to PCIe expansion and RAM capacity. That might require a custom SoC with more PCIe and maybe an external RAM bus, in which case, as you say, Apple could call it something other than M1/M2 (which lets them have a longer lifecycle without sounding obsolete). That said, either a quad Mx or M1 Ultra cluster would be packing a hell of a lot of Thunderbolt ports which could be turned back into PCIe (if they aren’t already configurable that way).

You hit the nail on the head. That's pretty much exactly what I was thinking, a custom SoC possibly with PCIe and external RAM bus, with a different naming convention. I was also thinking it wouldn't be on the same upgrade cycle as other Macs. Regular Macs get updated every 1-2 years, but Mac Pro generations would occur closer to every 3 years or so. Not 6 years like the trash can Mac Pro fiasco though.

Just a guess really, but aren't we all guessing?

I think launching a M2 Ultra Mac Pro any time soon would really, really tick off some people who have bought Studio Ultras, as it would be seen as replacing the M1 Ultra.

This is also a very important consideration. Releasing a true M2 Ultra just 3 to 7 months later, in the same calendar year, would just p!ss people off. We are not talking about an iPad here, but a pretty high end professional level computer setup (Mac Studio).

I know people always say something better is always coming and you should just suck it up and buy when you need, but that's only true to a point. Apple realizes that having updates too frequently won't actually help generate more sales, and will likely actually just annoy people.

Having a different naming convention for the Mac Pro SoC would mean they could update it any time without sounding obsolete as you say, but also would mean such Mac Pro SoC releases would not be obsoleting other recently released Macs from a marketing standpoint.

Plus, I suspect the higher-selling MacBook range would be first in line for M2 whatever chips.

I agree, but I think this is kind of a different discussion, since as you know MacBook Pros would top out at M2 Max. M2 Ultra is not even a consideration for MacBook Pros.

BTW, I'm not sure if we'll see M2 Pro and M2 Max in 2022, but it wouldn't be out of the question, esp. given Apple's habit in the past of updating MacBook Pros nearly every year. That was on Intel though, so I'm thinking maybe Apple could be changing to a 1.5+ year cycle instead for Apple Silicon MBPs.