Rumored pro chip

[thingstoponder]

The 2020 iPad Pro had a12 cores when the a13 was out. Even the Air had a14 while the Pro was still at a12 for months. And that’s not even considering existing Intel machines where any recent Apple chip beats in single core performance. Or even the Intel Mac Pro having slower single core than the 5k iMac.

I just don’t think it matters much. Some lower end products will be faster in some workloads sometimes. You can’t sync every product on the same release cycle. And the higher end models with more cores will be faster overall in multicore which is what you buy pro machines for anyways.

Gurman is saying “as soon as this summer” for the new MacBook Pros so I think they are not getting pushed to next year and I think that means M1X.

Big iMac is another story. There’s been basically no reporting of that machine. Screen size, release date, etc. there could be something to the “m2x” thing but Gurman could just be being safe And hedging his bets. Didn‘t he say that in a newsletter? I don’t think that should be taken with the same degree of certainly as the Bloomberg news articles.

 

[cmaier]

The 2020 iPad Pro had a12 cores when the a13 was out. Even the Air had a14 while the Pro was still at a12 for months. And that’s not even considering existing Intel machines where any recent Apple chip beats in single core performance. Or even the Intel Mac Pro having slower single core than the 5k iMac.

I just don’t think it matters much. Some lower end products will be faster in some workloads sometimes. You can’t sync every product on the same release cycle. And the higher end models with more cores will be faster overall in multicore which is what you buy pro machines for anyways.

The a12z/a12x/a13 thing was more of an exception than the rule. But, again, we’ll see.

 

[Jorbanead]

It would be tremendously more work to go to M2. But they come out with a new microarchitecture every year anyway. (I am assuming it takes them about 2 years from start to finish, so they are working on at least 2 microarchitectures at a time).

Going from M1 -> M1x (same microarchitecture but more cores, more cache, different i/o's, etc.) would be much faster. I frequently worked on multiple related chips like that simultaneously.

As for your second question, M1x would, by definition, be based on M1 (with more cores, different cache size, other comparatively minor changes). So the question is whether the bigger iMacs and MBPs coming in the next 6 months or so have M2x's or M1x's in them. Either way, we know that new core microarchitectures will be out in Septemberish for iPhone, so that work will have been done.

Awesome thanks for the info.

Follow-up question: I’ve seen some on this forum speculate that they could start with an M2X chip for the next generation of silicon and “trickle down” to an M2. What are your thoughts on this? Wouldn’t it be easier to start with the M2 and then add more complexity later?

 

[cmaier]

Awesome thanks for the info.

Follow-up question: I’ve seen some on this forum speculate that they could start with an M2X chip for the next generation of silicon and “trickle down” to an M2. What are your thoughts on this? Wouldn’t it be easier to start with the M2 and then add more complexity later?

It wouldn’t make a difference. At AMD I designed the “big” chip before the “little“ chip (sledgehammer before claw hammer), for example. (Though both were single-core, they were actually both designed with multicore in mind, and it would have been just as easy to design that one first).

The reason you do the big or little first has more to do with marketing, market demands, or yields. For example, one good reason to do the big one first (if you’re a company like, say, AMD), is that if it’s on a new fab process, yields may be low, so cost-per-die will be higher than normal, and it’s easier to hide those costs in the price of expensive chips rather than cheap ones.

 

[theMarble]

Awesome thanks for the info.

Follow-up question: I’ve seen some on this forum speculate that they could start with an M2X chip for the next generation of silicon and “trickle down” to an M2. What are your thoughts on this? Wouldn’t it be easier to start with the M2 and then add more complexity later?

I don't think that would happen as Apple would have to scale the still in early stages M2 up to a pro level when they could just improve on the already released M1 and then focus on the next-gen. I'd say that it would go M1, M1X, M2, M2X, M3, M3X... and counting in anymore chips they make such as a more-powerful M.X for the upcoming Mac Pro next year and any cut down versions of those.

 

[cmaier]

I don't think that would happen as Apple would have to scale the still in early stages M2 up to a pro level when they could just improve on the already released M1 and then focus on the next-gen. I'd say that it would go M1, M1X, M2, M2X, M3, M3X... and counting in anymore chips they make such as a more-powerful M.X for the upcoming Mac Pro next year and any cut down versions of those.

Why would they have to scale the M2 up when they designed the M2”x” first and just need to scale it down to make the M2?

 

[Joelist]

Why would they have to scale the M2 up when they designed the M2”x” first and just need to scale it down to make the M2?

Kinda my thinking. If they are going straight to the new microarchitecture then they would lead off with M2X because off where they are in the Apple Silicon transition - the next targets are the Pro level MacBooks and iMacs.

 

[theMarble]

Why would they have to scale the M2 up when they designed the M2”x” first and just need to scale it down to make the M2?

True, but I think that the M2 (next-gen silicon not pro version of M1) will be based on ARM v9 and/or a process lower than 5nm so that it could maybe be used in the MacBook Air, iPad Pro and a remake of the 12" MacBook?

Apple is probably working on the M1X or whatever the pro chip will be called right now (Probably in the EVT or DVT stage) while the M2 is just in concept. It wouldn't make sense to have the next-gen chip on the previous gen process if the plan is to use a newer process anyway unless there are manufacturing issues.

 

[Jorbanead]

True, but I think that the M2 (next-gen silicon not pro version of M1) will be based on ARM v9 and/or a process lower than 5nm so that it could maybe be used in the MacBook Air, iPad Pro and a remake of the 12" MacBook?

Apple is probably working on the M1X or whatever the pro chip will be called right now (Probably in the EVT or DVT stage) while the M2 is just in concept. It wouldn't make sense to have the next-gen chip on the previous gen process if the plan is to use a newer process anyway unless there are manufacturing issues.

Cmaier may correct me on this, but I believe the A15 design is likely already done and maybe even in production? Final assembly for iPhone starts about now AFAIK if they want to release in September.

I believe if there was an M1X chip it’s likely done and in production right now. I could see M1X revealed this fall, with an M2 chip early next year. My main reasoning for this is if Apple does decide to use the X-branding for the more powerful chips, from a marketing standpoint it makes sense to have their next chip be M1X (or maybe just M2?) That way there’s a clean transition: “we had great success this last year with our M1 products, but today we have something even more powerful that we think the pros are really going to love! We call it, the M1X”.

Not that they couldn’t do an M2X next, but from a marketing standpoint to me that feels potentially confusing as they haven’t established the M2 yet.

 

[theMarble]

Cmaier may correct me on this, but I believe the A15 design is likely already done and maybe even in production? Final assembly for iPhone starts about now AFAIK if they want to release in September.

I believe if there was an M1X chip it’s likely done and in production right now. I could see M1X revealed this fall, with an M2 chip early next year. My main reasoning for this is if Apple does decide to use the X-branding for the more powerful chips, from a marketing standpoint it makes sense to have their next chip be M1X (or maybe just M2?) That way there’s a clean transition: “we had great success this last year with our M1 products, but today we have something even more powerful that we think the pros are really going to love! We call it, the M1X”.

Not that they couldn’t do an M2X next, but from a marketing standpoint to me that feels potentially confusing as they haven’t established the M2 yet.

Thinking about it, I'd agree with you in that the M1X is probably done and ready to ship whenever Apple reveals it to the public. It would not make sense to go M1, M1X, M2X, M2. They would do the low-end first and then later give the treatment to the high-end products (as seen many times in the past like how when Apple ditched PPC, the iMac went first and the Mac Pro was last.)

Apple seems to care more about the average consumer more than the hardcore pro so it would make sense that the low-mid range chip and it's respective products would be announced first in the year/lifespan before the pro level products would.

 

[theMarble]

M1X would come out at the end of this year and the M2 / Next-gen low-mid end chip would come in Q1 or Q2 2022 so that the first M1 products (M1 MBA, MBP, Mini) can be updated except from the M1 MBP which will probably get an M1X 14" update.

 

[09872738]

while the M2 is just in concept.

Doubt that part. M2 is almost certainly in a far more advanced stage than just concept. Apple‘s got 10 more months only if they intend to complete the transition in time, so they need M2 at that latest

 

[JMacHack]

Cmaier may correct me on this, but I believe the A15 design is likely already done and maybe even in production? Final assembly for iPhone starts about now AFAIK if they want to release in September.

I believe if there was an M1X chip it’s likely done and in production right now. I could see M1X revealed this fall, with an M2 chip early next year. My main reasoning for this is if Apple does decide to use the X-branding for the more powerful chips, from a marketing standpoint it makes sense to have their next chip be M1X (or maybe just M2?) That way there’s a clean transition: “we had great success this last year with our M1 products, but today we have something even more powerful that we think the pros are really going to love! We call it, the M1X”.

Not that they couldn’t do an M2X next, but from a marketing standpoint to me that feels potentially confusing as they haven’t established the M2 yet.

Iirc the rumor mill said tsmc had the “pro” m1 chips in mass production a month ago. The thing that’s “held up” the 14 and 16 inch MacBook Pros is mini led yield issues.

 

[Homy]

True, but I think that the M2 (next-gen silicon not pro version of M1) will be based on ARM v9 and/or a process lower than 5nm so that it could maybe be used in the MacBook Air, iPad Pro and a remake of the 12" MacBook?

It's been already discussed and established many times that M1 already had many of ARM v9 features and that it is ARM v9 that is catching up to M1.

 

[AgentMcGeek]

True, but I think that the M2 (next-gen silicon not pro version of M1) will be based on ARM v9 and/or a process lower than 5nm so that it could maybe be used in the MacBook Air, iPad Pro and a remake of the 12" MacBook?

M2 should get N4 (4nm) and M1X N5P (5nm+) processes, from what we’ve heard.

 

[dgdosen]

Whether it be M1 (slightly disappointed) or M2 based -

I think something that looks a Jade C-die with 32GB of ram would be ludicrously adequate for my needs.

 

[cmaier]

M2 should get N4 (4nm) and M1X N5P (5nm+) processes, from what we’ve heard.

Perhaps.

 

[Colstan]

While the speculation is fun, we have a data point of exactly one, namely the M1. Just because Apple had a certain cadence with the iPhone and iPad processors, doesn't mean that they will follow that same pattern with the Mac, or even continue that pattern with other products.

It's like speculating on whether alien life exists outside of our solar system. We have a data point of one: Earth. We can plug in all the variables we want into the Fermi paradox, but anyone who claims to know the answer is just guessing. There can be educated guesses with Apple Silicon, but the Apple leakers are as reliable as UFO conspiracy theorists, at this point. Still, I want to believe that a second generation chip is in the works, not just a scaled up M1, because an M1X would be boring.

 

[AgentMcGeek]

Perhaps.

I guess it also depends on what process A15 has been produced with. Did they go for N5P or N4? We’ll soon find out.

 

[Jorbanead]

While the speculation is fun, we have a data point of exactly one, namely the M1. Just because Apple had a certain cadence with the iPhone and iPad processors, doesn't mean that they will follow that same pattern with the Mac, or even continue that pattern with other products.

It's like speculating on whether alien life exists outside of our solar system. We have a data point of one: Earth. We can plug in all the variables we want into the Fermi paradox, but anyone who claims to know the answer is just guessing. There can be educated guesses with Apple Silicon, but the Apple leakers are as reliable as UFO conspiracy theorists, at this point. Still, I want to believe that a second generation chip is in the works, not just a scaled up M1, because an M1X would be boring.

Yes, but we can look at the A-series of chips to extrapolate some data that may or may not be relevant. Additionally, we have information from Gurman, and there are people in this forum that may or may not have some insider information as well. So while the only 100% confirmed piece of data we have is M1, there's at least a bit more info we can discuss. Like you said, it's just fun to speculate.

 

[PsykX]

M2 should get N4 (4nm) and M1X N5P (5nm+) processes, from what we’ve heard.

Doubt it. Shouldn't it be :
M1X = 5nm
M2 = 5nm+
M3 = 4nm

Remember, M1 and M1X are based on A14 and M2 on A15.

 

[PsykX]

an M1X would be boring.

Yet it should bench at around 13,000 in Geekbench 5.
It's going to be like the difference Between A chips and AX chips.
M3 should bench the same thing as M1X. An X chip is always two generations ahead in terms of speed, I wonder if it's going to be true with the M chips.

 

[cmaier]

Doubt it. Shouldn't it be :
M1X = 5nm
M2 = 5nm+
M3 = 4nm

Remember, M1 and M1X are based on A14 and M2 on A15.

pretty sure that whatever the next chip is, it is not on the 5nm node. Pretty sure it's the plus node. I don't think anything else is coming on node 5.

 

[AgentMcGeek]

Doubt it. Shouldn't it be :
M1X = 5nm
M2 = 5nm+
M3 = 4nm

Remember, M1 and M1X are based on A14 and M2 on A15.

Well A14 and M1 are on N5, yes. But we've had contradicting rumours about whether M1X will be on N5 or N5P. N5P has been available for mass production for a while now, and the M1X has been rumoured to be mass produced only last month, which sounds like the same timing as for A15 chip production.

A15 should logically come out on N5P, and if both iPhones and new MBP come out roughly at the same time, it wouldn't be surprising if both A15 and M1X were on the same N5P node.

N4 is going on risk production in Q3 this year, so we could see M2 mass produced on N4 fairly early next year. Apple tends to get priority orders from TSMC. Both N5P and N4 are design compatible with N5. N5P is a process optimization and N4 is a fairly simple node shrink. So the node of the M chips are gonna be dictated mostly by TSMC availabilities' and Apple priorities vs A chips.

 

[jdb8167]

Apple could solve all the naming controversy by calling everything M1 in the current A14 generation and everything in the A15 generation M2. Then call out features that differentiate each specific tier. If you notice on the M1 they are already doing that. M1 4p/4e CPU cores and 7 or 8 GPU cores. The next M1 could be 8p/2e CPU cores and 16 or 32 GPU cores.

This also solves the problem if they are going to just move on to the M2 without any further M1 SoCs. Now it doesn’t matter if the MacBook Pro models are before the M2 MacBook Air models. It is simply M2 8p/2e CPU cores and 16 or 32 GPU cores. When the MacBook Air comes out next year it is something like 6p/2e CPU cores and 16 GPU cores.

I don’t see any real reason for Apple to start with a M#X naming scheme now that iPad Pros are no longer using the A#X scheme.