Future M1 Variants

[deconstruct60]

I'm thinking:
....

M1Z - 28P + 4E in Mac Pro (possibly less P cores, I just think they'd want to match the current top end core count. Maybe start from fewer P cores and up-sell to 28?)

Apple probably isn't going to match P cores toe-to-toe with the current Intel cores. The M1 is demonstrative that they are doing better with a smaller total count. That will probably continue in the top end. As long as Apple is incrementally better than the current Mac Pro I think they'll call that a win. ( As opposed to chasing the core counts of AMD and Intel in late 2022 in the server space. )

The workload that Apple can push into the Neural (ML) cores and onto future GPUs is where the "embarrassingly parallel" computational workload will probably what they increasingly lend on.

If Apple can scale their system cache to keep 16-20 cores fed that will probably be the top end.

 

[aeronatis]

GPU options like that would mean too many completed mobo SKUs to stock...

Apple will most likely segment the APUs by overall core counts, the more "powerful" the Mac is supposed to be, the higher the core counts on the SoC...

Updating my speculated APU line-up...!

M1 - 4 P / 4 E / 8 GPU / 16 Neural Engine / 16GB LPDDR5 / two USB4 ports
MacBook Air / 13" MacBook Pro / Mac mini

M1X - 8 P / 4 P / 16 GPU / 24 Neural Engine / 64GB LPDDR5 / four USB4 ports
14" MacBook Pro / 24" iMac / Mac mini

M1Z - 16 P / 4 E / 32 GPU / 32 Neural Engine / 128GB HBMnext / four USB4 ports
16" MacBook Pro / 27" iMac / 30" iMac Pro / Mac Cube

M1T - 32 P / 6 E / 64 GPU / 64 Neural Engine / 256GB HBMnext / eight USB4 ports
Mac Pro (Threadripper-sized package)

Apple discrete GPU (Lufika) - 64 GPU / 64 Neural Engine / ?? HBMnext ??

Mac Pro may have DDR5 DIMMs implemented as a secondary RAM array, up to 1TB...?

All desktops will have two USB-A ports & a HDMI port...?

I don't (or don't wanna) believe that there will be that much of a performance gap between 14" and 16" MacBook Pro (should be a similar situation with 11" and 12.9" iPad Pro). That ruins the whole purpose as Apple Silicon chip providing high performance without using too much power.

Also, those GPU core estimations I made earlier were just for the chip to be used in iMac Pro and Mac Pro (could be dedicated). Also, I don't think there will be different number of NE cores on the same architecture (see A12 vs A12X/A12Z). Like I say, iMac Pro and Mac Pro, I cannot imagine much; however, M1 on 3 different devices now actually states that there is another single chip for the likes of higher end MacBook Pro and iMac.

I also like the idea of HBM2E as a system memory; however, that, too, just seems like wishful thinking. LPDDR5 should be all we see for now (maybe a 4 GB HBM2E tile memory for the GPU at best). I would be extremely happy to be proven wrong, though.

 

[thingstoponder]

A14T and an Apple GPU (code name Lifuka) was recently leaked by Chinese Times. If so, Apple seems to go with two families of Apple Silicon for Macs.

That name is obviously wrong considering they didn’t call the Mac chips “A” series. If that chip is real it will be M series.

 

[aeronatis]

That name is obviously wrong considering they didn’t call the Mac chips “A” series. If that chip is real it will be M series.

Considering M1 is basically A14X with different naming (maybe higher clocks + thunderbolt controller), there is nothing to stop them from marketing a prototype named A14T as M1X or whatever they will call it.

 

[dogslobber]

Pro Desktop chips won't have 'efficiency' cores as they're useless and consume space on the chip. What you'll probably have is two division of chips for Pro and for power savings. We got the latter which is why those products were dumped on the market last week. The follow on Macs will be Pro and the laptops will be the systems they didn't have time to built yet. Remember, designing unique chips is a very expensive process so Apple will be trying to reuse designs across as many platforms as possible to reduce costs.

 

[deconstruct60]

I doubt the desktop chips will be called M1. I think M stands for mobile.

This is actually Apple’s second M series .

Apple silicon - Wikipedia

en.wikipedia.org

up until the recent intro M should for Motion , not mobile . That wiki page has been refactored . The motion stuff moved off to

Apple motion coprocessors - Wikipedia

en.wikipedia.org

since Apple is reusing the name it highly likely means Mac ( Macintosh ) . Apple has had ‘mobile’ processors all along .


First these are suppose to built with Macs in mind . The ‘mobile aspect has been subsumed into the SoC to point not a highlight . That ‘mobile’ aspect is likely also secondary . The M1 Pretty likely is just a rebadged A14x with some features turn off (for some and on for a few others )
Finally the notion that the Mac is going to get two distinct processors lines is huge leap . The M1 is likely shared . There probably will be at least one that isn’t shared but Apple can just throw a letter suffix on it .

The folks trying to map the naming system back into Intel ... that is just whacked . core i3 , i5 , i7 is only part of the names . full names are. i n NNNN. . The 3,5,7,9 just group into classes Anne there are model numbers inside that. Intel makes an order of magnitude ( at least ) wider variety of process than Apple is going to make .

The Mac line going to 3-5 variant is a radical change from the one per product line Apple usually does . There won’t be dozens that need to be split into subgroups.

 

[Boil]

Pro Desktop chips won't have 'efficiency' cores as they're useless and consume space on the chip.

All Apple Silicon chips for the Mac platform will have Efficiency cores; just because a Mac is part of the Pro line-up does not mean the OS does not still have mundane tasks that are best processed on Efficiency cores...

 

[Erehy Dobon]

The logical thing would be to put 2 M1 on the Mac Mini Pro, the iMac and 4-6-8 on the Mac Pro.

This makes zero sense. Each SoC has its own UMA. Having several SoCs would complicate memory management.

Also, there's duplication in certain functions like disk controller, Secure Enclave, etc.

For higher performance, it make more sense having a different balance of performance and efficiency cores. Let's say you are running some multimedia application plus an e-mail client on a MacBook Air. On a Mac Pro, it's not like you are running four e-mail clients or eight weather widgets and the Secure Enclave needs to be 4x or 8x larger.

It may make more sense having more performance cores than efficiency cores on a high-end SoC than the entry-level M1 especially if the unit is a desktop system that is getting AC power from the wall rather than a battery-dependent device.

My guess is that Apple will refresh the M1 next year with something like an M1X while concurrently releasing a completely new SoC with a different TDP; let's call it the M10.

In 2022, it would seem like Apple might release yet another SoC, let's call it the M20 or M100. Again, the new SoC would have a different balance of cores yet maintain certain features (like Secure Enclave) that don't need to scale linearly.

Apple explicitly said that Apple Silicon would be a family of processors, not just multiples of the same chip. If it were the latter, they could have refreshed all product families at once.

 

[thenewperson]

Apple probably isn't going to match P cores toe-to-toe with the current Intel cores. The M1 is demonstrative that they are doing better with a smaller total count. That will probably continue in the top end. As long as Apple is incrementally better than the current Mac Pro I think they'll call that a win. ( As opposed to chasing the core counts of AMD and Intel in late 2022 in the server space. )

That's very true, considering with the preliminary benchmarks we've seen from the M1 so far.

If Apple can scale their system cache to keep 16-20 cores fed that will probably be the top end.

Maybe 16 for the 1st SKU and 20 for the second? At least that'll appease my sense of the consistency with each variant going from 4 to 8 to 12 to 16 to 20. I like it!

 

[thenewperson]

Pro Desktop chips won't have 'efficiency' cores as they're useless and consume space on the chip.

This doesn't track. Why waste power on things the efficiency cores will be used for? Just because desktops don't have batteries doesn't mean they'll forgo the power savings that'll come from using the efficiency cores where they can.

Remember, designing unique chips is a very expensive process so Apple will be trying to reuse designs across as many platforms as possible to reduce costs.

For this reason as well I feel like we'll see the efficiency cores present.

 

[Erehy Dobon]

Pro Desktop chips won't have 'efficiency' cores as they're useless and consume space on the chip.

My guess is that Johnny Srouji will make some sort of public appearance in a keynote presentation sometime next year and prove that you are wrong.

Apple has been heavily focused on green computing. If you watched the M1 Mac keynote, one of the presenters said that the greenest power is the power that isn't being used.

Mac Pro will definitely have efficiency cores. You don't need some performance core running at 3.6GHz to check e-mail in the background while you walk away from your Mac Pro to get a cup of coffee or take a bio break.

Apple will use their prowess to minimize power consumption by the Mac Pro in the situations when peak performance isn't necessary. There is nothing new about this.

However, I doubt if Apple will have a 1:1 ratio of performance:efficiency cores on whatever M-series SoC is on the Mac Pro. More likely Apple will have more performance cores on the Mac Pro M-series SoC. It's not like the Mac Pro will run 8 instances of Mac Mail or need 8x space on the Secure Enclave.

This is a great example of how many techologists simply can't see the forest for the trees.

 

[thingstoponder]

Considering M1 is basically A14X with different naming (maybe higher clocks + thunderbolt controller), there is nothing to stop them from marketing a prototype named A14T as M1X or whatever they will call it.

True but that chip would never be in an iOS device so the A14T name is probably not true at all. Or just an internal name. Either way it it’s real it will only ever be an M chip for consumers.

 

[thingstoponder]

My guess is that Apple will refresh the M1 next year with something like an M1X while concurrently releasing a completely new SoC with a different TDP; let's call it the M10.

No way. The number is the generation. The successor to the M1 will be the M2. The higher TDP chip will probably be M1X and is probably months away.

As for the discussion above of whether the “desktop” chips will have efficiency cores. Yes they will. The iMac is not getting its own dedicated chips, they will almost certainly share a chip with the 16” Pro. However I think 4 efficiency cores is going to be on all of them. The rumor in Bloomberg was 8+4. You only need so many efficiency cores and I think Apple wants a common amount across all chips. I also don’t think that the neural engine will be any bigger than 16 cores either. They’ll keep it the same per generation of chip.

Size is not a concern really btw. The efficiency cores are absolutely tiny.

 

[leman]

A contributing problem here is that Apple SoC's gimp the output ports down to what Apple feels is sufficient for thin, " minimal port area" laptops. That's a substantive disconnect for a desktop system solution which is going to require a different mindset when it comes to setting the SoC provisioning requirements.

Apple can kept the chopped down mobile variant around like the non-Retina iMac to hit price points. It can also be the same treatment where it drifts in revisions to use cheaper components over longer lifespans.

I expect the I/O to improve considerably on the low-end in the coming years. Let's face it, this is the first time Apple releases desktop hardware, so the smart thing is not to over-extend.

 

[bernhard]

I expect Aple to move to a NUMA architecture with their own SoCs. Increasing the SoC-size will prove difficult, adding 64GB RAM as modules will take up considerable space, but adding multiple SoCs with local RAM will scale almost linear. Most likely not with the M1, which is probably a Single-SoC but with a Desktop-class M1X.

The Desktop-class SoC will probably:

• remove the iGPU and either
  • add an Apple-designed one on the PCIe-bus
  • add an nVidia/ATI card on the PCIe-bus
• use the freed up space for additional CPUs
  • adding Firestorm Cores
  • removing a couple Icestorm Cores, probably leaving 2 or so
  • adding NE cores

With the power budget of a Mac Pro you are more restrained by the complexity of the cooling system required for running ten 15-20W SoCs than the power draw of the SoCs themselves.

 

[dogslobber]

This doesn't track. Why waste power on things the efficiency cores will be used for? Just because desktops don't have batteries doesn't mean they'll forgo the power savings that'll come from using the efficiency cores where they can.


For this reason as well I feel like we'll see the efficiency cores present.

When you buy a desktop, the last thing you’re concerned about is saving a few watts when checking email. These machines are bought for speed and if these so-called efficiency cores are there then you immediately ask why your computer is being neutered for slowness when you buy it for speed.

Remember ‘efficiency cores’ is just marketing speak for slow and small cores. That’s fine But not is a real computer.

 

[dogslobber]

My guess is that Johnny Srouji will make some sort of public appearance in a keynote presentation sometime next year and prove that you are wrong.

Apple has been heavily focused on green computing. If you watched the M1 Mac keynote, one of the presenters said that the greenest power is the power that isn't being used.

Mac Pro will definitely have efficiency cores. You don't need some performance core running at 3.6GHz to check e-mail in the background while you walk away from your Mac Pro to get a cup of coffee or take a bio break.

Apple will use their prowess to minimize power consumption by the Mac Pro in the situations when peak performance isn't necessary. There is nothing new about this.

However, I doubt if Apple will have a 1:1 ratio of performance:efficiency cores on whatever M-series SoC is on the Mac Pro. More likely Apple will have more performance cores on the Mac Pro M-series SoC. It's not like the Mac Pro will run 8 instances of Mac Mail or need 8x space on the Secure Enclave.

This is a great example of how many techologists simply can't see the forest for the trees.

Green computing is a fabricated marketing message, not a design goal. I don’t swallow this rubbish about trying to save the planet as it’s pure BS to me. You turn the engineering output into some type of tree hugging marketing message to win plaudits. But the original engineering req was still how to make the battery last longer.

As I point out in another post, people who buy expensive desktops don’t give a damn about efficiency cores. They want speed. And more of it.

 

[thenewperson]

When you buy a desktop, the last thing you’re concerned about is saving a few watts when checking email. These machines are bought for speed and if these so-called efficiency cores are there then you immediately ask why your computer is being neutered for slowness when you buy it for speed.

Remember ‘efficiency cores’ is just marketing speak for slow and small cores. That’s fine But not is a real computer.

I think this is all nonsense. So what if you bought your computer for speed? The efficiency cores aren't somehow going to be taking over from the performance cores for intensive tasks, they'll be there for background tasks which speed isn't necessarily an important factor. You mentioned yourself their need to reuse designs across each of their SoCs, and the 4 efficiency cores being present would likely be a part of that. Also the Mac mini has them.

 

[ophh1]

https://www.macrumors.com/2020/04/23/12-core-arm-macs-2021-report/

Back in April Bloomberg reported that Apple was developing 3 SoCs based on A14 and one of them has a 12-core (w/ 8 Performance + 4 Eifficient) CPU.

Given that the M1 only has an 8-core CPU, my guess is that 12-core CPU variant would be the M1X for the upcoming 14 & 16 inch model, or even the low end 24 inch iMac.

And the last one (probably called M1Z) will be the desktop class SoC for the larger iMac.

Btw, I hope Apple will bring back target display to the iMac. As far as I recall, 5K iMac doesn’t support target display is because of Intel/bandwidth.

 

[dogslobber]

I think this is all nonsense. So what if you bought your computer for speed? The efficiency cores aren't somehow going to be taking over from the performance cores for intensive tasks, they'll be there for background tasks which speed isn't necessarily an important factor. You mentioned yourself their need to reuse designs across each of their SoCs, and the 4 efficiency cores being present would likely be a part of that. Also the Mac mini has them.

Let me rephrase it. If you have 4 e-cores that could be gutted and fit 2 normal cores then you would go with the latter. Otherwise people who buy these systems will be wondering why they're netted with slow cores when they want speed. That's the trade of you are willing to pay for real computing power. The Mac mini is simply reclaiming the crown of the lowest powered desktop on the planet. The 2018 Mini was too powerful which is why it has been castrated. Think of what they're doing here. There is classic market segmentation planned with the ARM M1 and the ARM P1 chips.

 

[UltimateSyn]

Here’s my guess, based on basically nothing:

June 2021 - M1X: 14” MBP, 24” iMac, Upper-Tier Mac Mini
8P + 4E + 12 GPU + 16 Neural

June 2021 - M1Z: 16” MBP, 30” iMac
12P + 4E + 16 GPU + 16 Neural

November 2022 - M3T: iMac Pro, Mac Pro
28P + 4E + 32 GPU + 32 Neural

 

[Traverse]

I don't think Apple will go with that scheme like Intel (i3, i5, i7, i9). I think it will M1 for medium performance, H1 for high performance, and X1 for extreme performance. Something along those lines....

I agree. The other scenarios just don't make sense to me, although neither did the iPhone XS Max.

Some people think they'll announce the M2 for the higher end laptops, but then what happens to next year's MacBook Air? Jump from M1 to M3? Or M1.2 or M1(letter)? That seems messy. I see either two scenarios:

• The M-line remains for lower end and the other lines get a different letter (e.g. M1, D1 for iMac, X1 for Pros, etc.). Or more likely:
• The M1 line is the family of chips for this generation so it's:
  • M1 - lower end devices
  • M1X - mid-tier and higher laptop
  • M1Z (or other letter) for high-end iMac
  • Mac Pro may get it's own thing.

With option 2, Apple can just iterate with M1X, M2X, etc. But knowing Apple the next device will come with a M3X PRO or something random.

 

[iPadified]

Here’s my guess, based on basically nothing:

June 2021 - M1X: 14” MBP, 24” iMac, Upper-Tier Mac Mini
8P + 4E + 12 GPU + 16 Neural

June 2021 - M1Z: 16” MBP, 30” iMac
12P + 4E + 16 GPU + 16 Neural

November 2022 - M3T: iMac Pro, Mac Pro
28P + 4E + 32 GPU + 32 Neural

I almost agree. The interesting part is the "pro" models as these by necessity needs to have large configurability. One SoC seem very limited in choices. I favour a model with one decent SOC plus add on cards like the afterburner but for GPU and neural engine. However, another approach would be to make multi SoC solutions. Think 10 of your M1Z in the Mac Pro in a cluster. One of our suggested M3T has too little compute/graphics power.

 

[jamisonbaines]

Let me rephrase it. If you have 4 e-cores that could be gutted and fit 2 normal cores then you would go with the latter. Otherwise people who buy these systems will be wondering why they're netted with slow cores when they want speed. That's the trade of you are willing to pay for real computing power. The Mac mini is simply reclaiming the crown of the lowest powered desktop on the planet. The 2018 Mini was too powerful which is why it has been castrated. Think of what they're doing here. There is classic market segmentation planned with the ARM M1 and the ARM P1 chips.

Just because you're pushing a demanding workload it doesn't mean the mundane background tasks stop. There will likely always be a role for efficiency cores. Low power draw, leave it running 24/7 is my ideal use case for a Mac mini and this new version seems like a good candidate for that.

I used to have a Mac mini, in 2007. We all go through these phases. Some of us clung to the 2011/2012 quad core mini's, maybe you liked the 6 core. They were never too powerful, just modern incarnations. There were always more powerful desktop options or equally capable laptops with higher price tags.

 

[UltimateSyn]

I almost agree. The interesting part is the "pro" models as these by necessity needs to have large configurability. One SoC seem very limited in choices. I favour a model with one decent SOC plus add on cards like the afterburner but for GPU and neural engine. However, another approach would be to make multi SoC solutions. Think 10 of your M1Z in the Mac Pro in a cluster. One of our suggested M3T has too little compute/graphics power.

I like the thought of the Afterburner-like add-on cards to keep the Mac Pro modular, but wouldn’t that sort of ruin the whole Apple Silicon unified memory approach? I’m really not sure.

The multi-CPU approach is naturally where I first went with the Mac Pro, too. I thought “Oh, of course. They’ll have a high-end M1Z for the iMac and just put 8 of them in a Mac Pro” but I have seen a lot of people around here with much more processor knowledge than me saying that is inefficient and impractical because of the CPU to CPU communication lines. And there would apparently be redundancy because it would necessitate multiple Secure Enclaves.

I’m reallllllllyyyy interested to continue seeing how they structure the Apple Silicon family, even if I won’t fully comprehend the reasoning behind why they make the decisions do.