My Thoughts on Future Mac ARM Chips

[Mr. Awesome]

I’ve seen a lot of posts talking about a higher-end M-series chip used in the iMac, 16” MBP, etc. The problem is that most people are calling this the M2 chip, which I think is incredibly unlikely.

The M1, despite its incredible performance, is Apple’s entry level, power-efficient chip. They’ll have other variants for more powerful machines like the iMac, where power draw is less of a concern.

The “M2” chip, as most people are calling it, is probably what will be used in a new iMac and 16” MacBook Pro. I think, however, that it is way more likely that it will be called the M1X. This makes the naming consistent with the A-series processors, and allows them to use the M2 name for the next generation of their entry-level power-efficient chip, which will be the direct successor to the M1.

Think of it like this: the A12 chip was used in the iPhone, and the A12X chip, which was more powerful, was used in the iPad Pro. They didn’t call the 12X the A13, because it wasn’t the direct successor to the A12. It was a more powerful variant of the A12. Thus, the chip used in the 16” MBP and the iMac will almost certainly be the M1X.

—

Here’s how I expect these chips will be used.

Mac mini
M1 for base models, but configurable with an M1X (with a limited number) once it’s released.

MacBook Air
M1 chip only

13” MacBook Pro
M1 chip for entry-level two-port model.
Higher-end configurations can use the M1X chip, but it can’t be configured with the maximum number of cores.

16” MacBook Pro
M1X chip, configurable with up to the maximum number of cores.

iMac
M1X chip. Smaller (24”?) models can’t be configured to the maximum number of cores, but larger (30”?) models can.

iMac Pro
I expect the iMac Pro to be discontinued after the update to the iMacs. Apple may keep it around for a while for those who need an Intel iMac, but other than that I see no place for it in the lineup. It was a stopgap until Apple had the Mac Pro ready, and now that the Mac Pro is here, it’s not needed.

Mac Pro
M1Z chip. This could even be an M2Z, depending on if Apple already has the M2 chips ready by the time they update the Mac Pro with Apple Silicon.

“Mac Pro mini” / “Mac mini Pro”
The rumors about a smaller Mac Pro are intriguing, and if it was released, I would expect it to use an M1Z, potentially with a limit on core count.

—

I expect the M1X will be configurable up to around maybe 10 high performance cores and 6 high efficiency cores in the CPU, and probably around a 12 core GPU at max? That depends on if Apple will use dGPUs in these Macs.

I definitely think that the M1X/M1Z chips will have higher limits on connected displays, RAM, and Thunderbolt ports.

Anyway, those are my thoughts. I’m excited to hear yours.

 

[cmaier]

I’ve seen a lot of posts talking about a higher-end M-series chip used in the iMac, 16” MBP, etc. The problem is that most people are calling this the M2 chip, which I think is incredibly unlikely.

The M1, despite its incredible performance, is Apple’s entry level, power-efficient chip. They’ll have other variants for more powerful machines like the iMac, where power draw is less of a concern.

The “M2” chip, as most people are calling it, is probably what will be used in a new iMac and 16” MacBook Pro. I think, however, that it is way more likely that it will be called the M1X. This makes the naming consistent with the A-series processors, and allows them to use the M2 name for the next generation of their entry-level power-efficient chip, which will be the direct successor to the M1.

Think of it like this: the A12 chip was used in the iPhone, and the A12X chip, which was more powerful, was used in the iPad Pro. They didn’t call the 12X the A13, because it wasn’t the direct successor to the A12. It was a more powerful variant of the A12. Thus, the chip used in the 16” MBP and the iMac will almost certainly be the M1X.

—

Here’s how I expect these chips will be used.

Mac mini
M1 for base models, but configurable with an M1X (with a limited number) once it’s released.

MacBook Air
M1 chip only

13” MacBook Pro
M1 chip for entry-level two-port model.
Higher-end configurations can use the M1X chip, but it can’t be configured with the maximum number of cores.

16” MacBook Pro
M1X chip, configurable with up to the maximum number of cores.

iMac
M1X chip. Smaller (24”?) models can’t be configured to the maximum number of cores, but larger (30”?) models can.

iMac Pro
I expect the iMac Pro to be discontinued after the update to the iMacs. Apple may keep it around for a while for those who need an Intel iMac, but other than that I see no place for it in the lineup. It was a stopgap until Apple had the Mac Pro ready, and now that the Mac Pro is here, it’s not needed.

Mac Pro
M1Z chip. This could even be an M2Z, depending on if Apple already has the M2 chips ready by the time they update the Mac Pro with Apple Silicon.

“Mac Pro mini” / “Mac mini Pro”
The rumors about a smaller Mac Pro are intriguing, and if it was released, I would expect it to use an M1Z, potentially with a limit on core count.

—

I expect the M1X will be configurable up to around maybe 10 high performance cores and 6 high efficiency cores in the CPU, and probably around a 12 core GPU at max? That depends on if Apple will use dGPUs in these Macs.

I definitely think that the M1X/M1Z chips will have higher limits on connected displays, RAM, and Thunderbolt ports.

Anyway, those are my thoughts. I’m excited to hear yours.

In various interviews, Apple executives have already said they will have variants of M1 that will simply be called “M1.” So, what I’ve been calling the M1X, for example, (more address bits, bigger cache, more CPU cores, higher external bandwidth for connection to discrete apple-designed graphics chip) will likely also be called “M1.” The idea being that consumers don’t really care.

 

[Mr. Awesome]

In various interviews, Apple executives have already said they will have variants of M1 that will simply be called “M1.” So, what I’ve been calling the M1X, for example, (more address bits, bigger cache, more CPU cores, higher external bandwidth for connection to discrete apple-designed graphics chip) will likely also be called “M1.” The idea being that consumers don’t really care.

Ah, interesting. I wonder how that’ll play out.

 

[thenewperson]

In various interviews, Apple executives have already said they will have variants of M1 that will simply be called “M1.” So, what I’ve been calling the M1X, for example, (more address bits, bigger cache, more CPU cores, higher external bandwidth for connection to discrete apple-designed graphics chip) will likely also be called “M1.” The idea being that consumers don’t really care.

That would be interesting. I wonder why they didn't do this with the iPads...

 

[KShopper]

I expect the M1X will be configurable up to around maybe 10 high performance cores and 6 high efficiency cores in the CPU, and probably around a 12 core GPU at max? That depends on if Apple will use dGPUs in these Macs.

I definitely think that the M1X/M1Z chips will have higher limits on connected displays, RAM, and Thunderbolt ports.

So generally agree with your post, however, I would expect the core counts on the M1X to substantially higher than what you have above (pure conjecture below).

4 x efficiency cores
12 x performance cores
16 or even 32 GPU cores
16GB or 32GB RAM options
4 x thunderbolt ports, support for 2 x external displays up to 6K each.

In addition to all that, the base clock speeds will be increased at least 10%, so 3.5Ghz vs 3.2Ghz on the base M1.

For the GPU cores, since they won't be supporting 3rd party GPUs, they will have to increase performance substantially. Here's my back-of-the-napkin analysis...

If you look at the benchmarks the M1 GPU with 8 cores is roughly equivalent to the Ti 1050 GPU, which has 3.3 million transistors. The new high end 1080 Ti has 12 million transistors. If we guesstimate that the M1 8x GPU cores take the same 3.3 million transistors as the 1080 ti, that represents about 21% of the total 16 milion transistor count on the M1 SoC. So the M1X would need to dedicate 3 to 4 x times transistors to GPU in order to get close to the 1080's performance, that means 24-32 GPU cores. In addition, the current 16" MBP higher end model uses an AMD Radeon 5500M GPU, which has 6.4m transistors is roughly 2 x faster than the 1080 Ti GPU. If Apple wants to keep the same 3x GPU performance improvement in the new M1X based MBP vs. the current Intel models that they achieved with the base M1 macbooks, they will need 32 GPU cores (16 cores to match the current Radeon 5500M, and double that to provide the 3x improvement). This should put them in very comparable terms with the Ti 1080 as well.

This is, of course, completely mind-blowing to consider vs where they were at with the Intel situation.

 

[leman]

In various interviews, Apple executives have already said they will have variants of M1 that will simply be called “M1.” So, what I’ve been calling the M1X, for example, (more address bits, bigger cache, more CPU cores, higher external bandwidth for connection to discrete apple-designed graphics chip) will likely also be called “M1.” The idea being that consumers don’t really care.

I am quite sure there will be some sort of suffix distinguishing the higher-performance M-chips. I don't think it will work out otherwise, since users might get the impression that the $3000 16" has the same CPU as the $999 MBA.

 

[Mr. Awesome]

So generally agree with your post, however, I would expect the core counts on the M1X to substantially higher than what you have above (pure conjecture below).

4 x efficiency cores
12 x performance cores
16 or even 32 GPU cores
16GB or 32GB RAM options
4 x thunderbolt ports, support for 2 x external displays up to 6K each.

In addition to all that, the base clock speeds will be increased at least 10%, so 3.5Ghz vs 3.2Ghz on the base M1.

For the GPU cores, since they won't be supporting 3rd party GPUs, they will have to increase performance substantially. Here's my back-of-the-napkin analysis...

If you look at the benchmarks the M1 GPU with 8 cores is roughly equivalent to the Ti 1050 GPU, which has 3.3 million transistors. The new high end 1080 Ti has 12 million transistors. If we guesstimate that the M1 8x GPU cores take the same 3.3 million transistors as the 1080 ti, that represents about 21% of the total 16 milion transistor count on the M1 SoC. So the M1X would need to dedicate 3 to 4 x times transistors to GPU in order to get close to the 1080's performance, that means 24-32 GPU cores. In addition, the current 16" MBP higher end model uses an AMD Radeon 5500M GPU, which has 6.4m transistors is roughly 2 x faster than the 1080 Ti GPU. If Apple wants to keep the same 3x GPU performance improvement in the new M1X based MBP vs. the current Intel models that they achieved with the base M1 macbooks, they will need 32 GPU cores (16 cores to match the current Radeon 5500M, and double that to provide the 3x improvement). This should put them in very comparable terms with the Ti 1080 as well.

This is, of course, completely mind-blowing to consider vs where they were at with the Intel situation.

Yeah, I was being kinda conservative with the GPU. 16 cores definitely sounds like a good estimate. 32 might happen, but it’ll probably be an upgrade, and might not be available for all devices with the M1X. For instance, if the 13” MBP gains an M1X configuration, I doubt you’ll be able to upgrade it all the way to 32 cores.

 

[UltimateSyn]

In various interviews, Apple executives have already said they will have variants of M1 that will simply be called “M1.” So, what I’ve been calling the M1X, for example, (more address bits, bigger cache, more CPU cores, higher external bandwidth for connection to discrete apple-designed graphics chip) will likely also be called “M1.” The idea being that consumers don’t really care.

I haven’t read or heard that anywhere and I’ve been keeping up with all of their interviews. Where did you hear them explicitly say that?

Pretty sure they will stick to the X/Z suffixes that the A-Series chips get.

 

[KShopper]

Yeah, I was being kinda conservative with the GPU. 16 cores definitely sounds like a good estimate. 32 might happen, but it’ll probably be an upgrade, and might not be available for all devices with the M1X. For instance, if the 13” MBP gains an M1X configuration, I doubt you’ll be able to upgrade it all the way to 32 cores.

Yeah, I'm probably being too optimistic with the GPU stuff. The leap to the 5nm M1 SoC enabled the massive increases in performance and efficiency we see over the previous Intel models, but things will get tougher moving up from here, in relative terms. There has to be a limit to the number of transistors they can embed into a single SoC module and still manufacture enough of them to make business sense.

You are probably correct in that the next level up, M1X, will be all on one SoC, and maybe 16 GPU cores only, and then they'll allow an upgraded configuration (probably later in 2021?) with a separate GPU module with it's own dedicated RAM on-chip as well. So the system will use the embedded GPUs until it needs the external GPU, like it uses the efficiency cpu cores until it needs the performance ones as well.

They already have the "Afterburner" boards you can buy for the Mac Pro, so something similar but all on-board. Eventually a new Mac Pro line that will in fact allow you plug in more M1XX type chips as daughter boards makes sense.

Going to very fascinating to see what they've come up with for sure.

 

[UBS28]

So generally agree with your post, however, I would expect the core counts on the M1X to substantially higher than what you have above (pure conjecture below).

4 x efficiency cores
12 x performance cores
16 or even 32 GPU cores
16GB or 32GB RAM options
4 x thunderbolt ports, support for 2 x external displays up to 6K each.

In addition to all that, the base clock speeds will be increased at least 10%, so 3.5Ghz vs 3.2Ghz on the base M1.

For the GPU cores, since they won't be supporting 3rd party GPUs, they will have to increase performance substantially. Here's my back-of-the-napkin analysis...

If you look at the benchmarks the M1 GPU with 8 cores is roughly equivalent to the Ti 1050 GPU, which has 3.3 million transistors. The new high end 1080 Ti has 12 million transistors. If we guesstimate that the M1 8x GPU cores take the same 3.3 million transistors as the 1080 ti, that represents about 21% of the total 16 milion transistor count on the M1 SoC. So the M1X would need to dedicate 3 to 4 x times transistors to GPU in order to get close to the 1080's performance, that means 24-32 GPU cores. In addition, the current 16" MBP higher end model uses an AMD Radeon 5500M GPU, which has 6.4m transistors is roughly 2 x faster than the 1080 Ti GPU. If Apple wants to keep the same 3x GPU performance improvement in the new M1X based MBP vs. the current Intel models that they achieved with the base M1 macbooks, they will need 32 GPU cores (16 cores to match the current Radeon 5500M, and double that to provide the 3x improvement). This should put them in very comparable terms with the Ti 1080 as well.

This is, of course, completely mind-blowing to consider vs where they were at with the Intel situation.

Apple is apparently working on a descrete GPU solution. So their GPU performance will most likely not come from the M1.

Ofcourse, they are just rumours and I am not sure how reliable these sources are.

 

[MrEcted]

I'm a bit of a tech nerd so I have a gaming PC and have built many in the past. I can only imagine the absolute insanity that will come from Apple (insanity in a good way).

These days it's not uncommon to require 750+ Watt PSUs to drive power hungry CPUs, GPUs, etc. I'm imagining a Mac Pro in the next couple years that will destroy all other workstations in CPU performance, and will be on the level of something like an RTX 3080 with GPU performance, it will have insanely fast storage, insanely fast and plentiful RAM, it will have a bunch of hardware level encoding/decoding, a crazy hardware level neural AI, etc - and it will nip at the heels of the most power hungry X86 workstation that pull 1000+ watts - all in a package that has a stupidly crazy TDP and pulls something like 30 watts and is completely silent with not even an audible whisper coming from a fan.

 

[UBS28]

Probably the Mac Pro will share the same chip as the iMac Pro (I am guessing 32 high performance cores). So I don’t think it will compete with a 128-core AMD build. But we shall see in 2 years.

 

[Yebubbleman]

I’ve seen a lot of posts talking about a higher-end M-series chip used in the iMac, 16” MBP, etc. The problem is that most people are calling this the M2 chip, which I think is incredibly unlikely.

The M1, despite its incredible performance, is Apple’s entry level, power-efficient chip. They’ll have other variants for more powerful machines like the iMac, where power draw is less of a concern.

The “M2” chip, as most people are calling it, is probably what will be used in a new iMac and 16” MacBook Pro. I think, however, that it is way more likely that it will be called the M1X. This makes the naming consistent with the A-series processors, and allows them to use the M2 name for the next generation of their entry-level power-efficient chip, which will be the direct successor to the M1.

Think of it like this: the A12 chip was used in the iPhone, and the A12X chip, which was more powerful, was used in the iPad Pro. They didn’t call the 12X the A13, because it wasn’t the direct successor to the A12. It was a more powerful variant of the A12. Thus, the chip used in the 16” MBP and the iMac will almost certainly be the M1X.

—

Here’s how I expect these chips will be used.

Mac mini
M1 for base models, but configurable with an M1X (with a limited number) once it’s released.

MacBook Air
M1 chip only

13” MacBook Pro
M1 chip for entry-level two-port model.
Higher-end configurations can use the M1X chip, but it can’t be configured with the maximum number of cores.

16” MacBook Pro
M1X chip, configurable with up to the maximum number of cores.

iMac
M1X chip. Smaller (24”?) models can’t be configured to the maximum number of cores, but larger (30”?) models can.

iMac Pro
I expect the iMac Pro to be discontinued after the update to the iMacs. Apple may keep it around for a while for those who need an Intel iMac, but other than that I see no place for it in the lineup. It was a stopgap until Apple had the Mac Pro ready, and now that the Mac Pro is here, it’s not needed.

Mac Pro
M1Z chip. This could even be an M2Z, depending on if Apple already has the M2 chips ready by the time they update the Mac Pro with Apple Silicon.

“Mac Pro mini” / “Mac mini Pro”
The rumors about a smaller Mac Pro are intriguing, and if it was released, I would expect it to use an M1Z, potentially with a limit on core count.

—

I expect the M1X will be configurable up to around maybe 10 high performance cores and 6 high efficiency cores in the CPU, and probably around a 12 core GPU at max? That depends on if Apple will use dGPUs in these Macs.

I definitely think that the M1X/M1Z chips will have higher limits on connected displays, RAM, and Thunderbolt ports.

Anyway, those are my thoughts. I’m excited to hear yours.

For the most part, I'm in agreement with you. I'm wondering what will happen to the 4-port 13" MacBook Pro and/or if it will ever merge with the 2-port version upon making the jump to Apple Silicon. I agree that there won't both be a larger standard iMac and an iMac Pro at the same size on the other end of this transition. That being said, I could just as easily see the larger sized iMac simply being dubbed "iMac Pro" with the smaller size iMac simply being "iMac", especially since whatever gets thrown into the larger iMac will probably handily best the iMac Pro ANYWAY. It's all marketing semantics, but I think that could at least simplify things for people (hopefully they err on the side of pricing the would-be "iMac Pro" closer to the current 27" iMac pricepoints rather than the current iMac Pro's pricepoints). But I totally agree that both larger iMacs won't exist as we know them today and that they'll merge.

I do believe that the Mac Pro's SoC will have to be a cut above even that of the 16" MacBook Pro, let alone low-mid range iMac SoCs. But what that entails is beyond me. I would think that they'll want to do something crazy with expandability, despite their SoC architecture seeming to go against expanding graphics and compute (outside of Afterburner). I could see them expanding their family of GPUs to include custom dGPUs, but how that would interface with their unified memory would seem to almost be a contradictory mystery.

I don't know about the "Mac mini Pro", but I do know that Apple does need to complete the transition of the Mac mini line to Apple Silicon. It is technically the only Mac model for which the transition was only half completed.

 

[Mr. Awesome]

For the most part, I'm in agreement with you. I'm wondering what will happen to the 4-port 13" MacBook Pro and/or if it will ever merge with the 2-port version upon making the jump to Apple Silicon. I agree that there won't both be a larger standard iMac and an iMac Pro at the same size on the other end of this transition. That being said, I could just as easily see the larger sized iMac simply being dubbed "iMac Pro" with the smaller size iMac simply being "iMac", especially since whatever gets thrown into the larger iMac will probably handily best the iMac Pro ANYWAY. It's all marketing semantics, but I think that could at least simplify things for people (hopefully they err on the side of pricing the would-be "iMac Pro" closer to the current 27" iMac pricepoints rather than the current iMac Pro's pricepoints). But I totally agree that both larger iMacs won't exist as we know them today and that they'll merge.

I do believe that the Mac Pro's SoC will have to be a cut above even that of the 16" MacBook Pro, let alone low-mid range iMac SoCs. But what that entails is beyond me. I would think that they'll want to do something crazy with expandability, despite their SoC architecture seeming to go against expanding graphics and compute (outside of Afterburner). I could see them expanding their family of GPUs to include custom dGPUs, but how that would interface with their unified memory would seem to almost be a contradictory mystery.

I don't know about the "Mac mini Pro", but I do know that Apple does need to complete the transition of the Mac mini line to Apple Silicon. It is technically the only Mac model for which the transition was only half completed.

I agree with most of your post, but when you say that the Mac mini is the only line that hasn’t fully transitioned, what about the 13” MacBook Pro? Only the entry level base model has been updated, not the whole line. A similar story to the Mac mini.

 

[Yebubbleman]

I agree with most of your post, but when you say that the Mac mini is the only line that hasn’t fully transitioned, what about the 13” MacBook Pro? Only the entry level base model has been updated, not the whole line. A similar story to the Mac mini.

The 2-port 13" MacBook Pro and the 4-port 13" MacBook Pro are technically two distinct Macs with different Mac model identifiers, firmwares and everything.

It's not the same as the Mac mini which merely had three processor options (quad-core i3, hexa-core i5, hexa-core i7) but was otherwise the same Mac. The Core i3 model of Mac mini was a different shipping model from the Core i5 version, but it was still "Mac mini (2018)" and it otherwise had the same internal components otherwise.

On the other hand, the 2-port 13" MacBook Pro, especially with its final Intel release was running a completely different generation of Intel CPUs than the current 4-port 13" MacBook Pro, also with a completely different internal and thermal design (stemming from the fact that the 2-port 13" MacBook Pro, during its run with Intel, used the same class and caliber of Intel CPUs that went into the 2011-2017 MacBook Air and was effectively the continuation of that Mac, whereas the CPUs in the 4-port have traditionally been higher clocked [and I want to say higher wattage, but don't quote me on that]). All of that, more or less stems from the fact that, in 2016, Apple was moving towards getting rid of "Air" in their products and the 2-port 13" MacBook Pro was meant to entice would-be Air customers wanting a Retina Air. Though, they went back on that, converted the 12" MacBook into the 2018-2020 era of Intel MacBook Airs. But yeah, long rant short, the 2-port 13" MacBook Pro and the 4-port 13" MacBook Pro are effectively two different Macs that just happen to be under the umbrella of "MacBook Pro 13-inch".

All that to say that when M1 came to town, the 2020 Intel MacBook Air was wholesale replaced by the 2020 M1 MacBook Air while the 2020 Intel 2-port 13" MacBook Pro was also wholesale replaced by the 2020 M1 2-port 13" MacBook Pro (two shipping models in each, all replaced). The Mac mini, on the other hand, had its lower-end shipping model replaced, but with the higher-end hexa-core Coffee Lake CPUs still in tow.

 

[Sanpete]

I am quite sure there will be some sort of suffix distinguishing the higher-performance M-chips. I don't think it will work out otherwise, since users might get the impression that the $3000 16" has the same CPU as the $999 MBA.

There's no suffix distinguishing the 7- and 8-core M1 chips. Apple might just go with 16-core M1, etc.

 

[leman]

There's no suffix distinguishing the 7- and 8-core M1 chips. Apple might just go with 16-core M1, etc.

Because it’s the same chip, just with one core disabled. A 16-core would be a physically different chip.