Do you expect an M1X and M2 SoC to be released this year?

[Fomalhaut]

That is quite a valid point as well. I am curious what sort of a line-up they will come up with. Also agree with you on having more than four efficiency cores being a waste. What I wonder more is whether 16" MacBook Pro and 27" iMac will have different chips or not. There is also chance that both will get the same chip with the iMac one having a higher power envelope.

I think it quite possible that the MBP16 and iMac27 will have different chips to the rumored MBP14 and iMac24.

My guess (and the reason for this thread) is that the latter will have an "M1X" with more cores (CPU/GPU) - maybe around mid-year, and the larger machines will get an M2 (with a higher-performing GPU architecture), towards the end of the year.

IIRC, the Intel iMac 27 has about twice the CPU TDP as the Intel MBP16 (with the i9), so it's certainly possible that the M2 could be produced in different configurations with variations in CPU cores/frequency and GPU power, which would certainly make sense for the iMac. Maybe Apple will use the existing A-series prefixes to denote variations, so the iMac 27 gets and M2X or similar, and the MBP16 has a base level M2....

I don't have any idea whether Apple will offer higher frequency CPU options as per Intel CPUs. They could make money by binning the SoCs, and offering a 200MHz upgrade for $$ (as they do currently).

I would love to be a fly on the wall in the Apple design lab to see what they're really up to! We've already seen in the data center that ARM-based chips can scale to lots of CPU cores, but GPU performance is still unknown, particularly if trying to keep the SoC below about 50W TDP for the MBP16 (a wild-assed guess). GPU performance needs to be about 2-4x better than than the M1 to offer the expected performance increase over the current AMD dGPUs. AMD has done this with their Playstation and XBox SoCs but at 180-200W TDP....

Let's see what Apple comes up with!

 

[Jorbanead]

I see apple creating 3 different chips.

M1 - Low End
M1X - Mid Tier
M1z - Pro Tier

So my guess is if and when they do go to M2 it’ll still be the lower tier chip. It’ll just be modeled after the A15 architecture with faster single thread performance and better power consumption. I don’t think the M2 chip will go into a MBP 16” as I think it’s still the lowest level of chip design.

I think of it sort of like:

M1 - i3/i5 Intel equivalent category
M1X - i7/i9 Intel equivalent category
M1z - Xeon Equivalent category

So a theoretical M2, M3 or M4 are all still part of that i3/i5 market share. They’re aimed at low power, basic consumer macs like MacBook Air and Mac mini. Yes they’ll have better single core performance, but the higher core variants with more features and higher ram capacity will be the X and Z variants.

 

[JouniS]

M1 - i3/i5 Intel equivalent category
M1X - i7/i9 Intel equivalent category
M1z - Xeon Equivalent category

I think the M1 is equivalent to a low-power i5/i7. It's not a low-end chip by any means, but it's constrained by low power usage and limited memory capacity.

The next tier should be the equivalent of laptop i5/i7 and desktop i5 intended for the MBP 14" and the smaller (24"?) iMac. And maybe for the cheapest MBP 16" models. They need a cost-effective chip that could replace most consumer-grade Intel CPUs without any compromises.

The MBP 16" probably needs its own tier with a chip comparable to a laptop i7/i9. Something that would use too much power for the MBP 14" and would be too expensive for the smaller iMac.

I'm not sure what Apple should do with the larger iMac. They could take advantage of the much higher thermal capacity and design the equivalent of a desktop i7/i9. Or they could finally commit to the vision of the iMac as a non-portable laptop and use the same laptop-grade chip as in the MBP 16".

Finally, the Mac Pro needs its own Xeon-equivalent tier.

 

[jdb8167]

I think the M1 is equivalent to a low-power i5/i7. It's not a low-end chip by any means, but it's constrained by low power usage and limited memory capacity.

The next tier should be the equivalent of laptop i5/i7 and desktop i5 intended for the MBP 14" and the smaller (24"?) iMac. And maybe for the cheapest MBP 16" models. They need a cost-effective chip that could replace most consumer-grade Intel CPUs without any compromises.

The MBP 16" probably needs its own tier with a chip comparable to a laptop i7/i9. Something that would use too much power for the MBP 14" and would be too expensive for the smaller iMac.

I'm not sure what Apple should do with the larger iMac. They could take advantage of the much higher thermal capacity and design the equivalent of a desktop i7/i9. Or they could finally commit to the vision of the iMac as a non-portable laptop and use the same laptop-grade chip as in the MBP 16".

Finally, the Mac Pro needs its own Xeon-equivalent tier.

The M1 is pretty low end when looking at RAM addressing and the number of external displays supported. From a performance perspective it is pretty amazing. It will be very interesting to see what gets added to upcoming SoCs. More displays, more RAM, more CPU cores, and more GPU cores and corresponding cache increases. Then there are the less known cores like the neural engine. How about basic IO/PCIe lanes. Will we get enough to handle 4 40Gb/s ports or will it be 2 40Gb/s buses over 4 ports like current Intel Macs. Lots of fascinating details that are upcoming.

 

[Jorbanead]

I think the M1 is equivalent to a low-power i5/i7. It's not a low-end chip by any means, but it's constrained by low power usage and limited memory

My comparison is not comparing the raw power of the M1 chip. It’s comparing which market those chips are designed for. The M1 does outperform even some mobile i9 chips as we all know, but its market is geared towards the same market the i3/i5 is geared towards - as those are the chips that are geared towards the base level consumer. That’s why I made those comparisons.

Compared to Intel’s raw power, of course the M1 is much more powerful than intels i3/i5 counterparts, but as for apples chips going forward, I believe these will be considered low power chips in comparison to apples higher end chips. That’s why it’s so exciting because apples bottom tier chip geared towards the average consumer is already so powerful. If that’s their “low end” I can’t imagine what their high end will be.

 

[leman]

I see apple creating 3 different chips.

M1 - Low End
M1X - Mid Tier
M1z - Pro Tier

This is enough to cover the mobile use case, but probably not the desktop one. There is still a significant difference in cooling capacity between a MBP (which is around 80watts) and a large iMac or a Mac Pro. And while I have little doubt that Apple can build a ridiculously fast laptop with the thermal budget of 80Watts, that chip won’t be enough to replace the current desktop offers. Desktop chips have to be larger, otherwise Mac desktop is essentially dead.

 

[JouniS]

My comparison is not comparing the raw power of the M1 chip. It’s comparing which market those chips are designed for. The M1 does outperform even some mobile i9 chips as we all know, but its market is geared towards the same market the i3/i5 is geared towards - as those are the chips that are geared towards the base level consumer. That’s why I made those comparisons.

The cheapest Intel Macs were usually based on i5, with often an option to upgrade to i7. There were a few i3 models, but Apple was never interested in the market segment the i3 was targeted for.

 

[spiderman0616]

Going back to the OP's original question: I'm not sure it makes sense anymore to even try to put together a pattern or predict any of this based on past hardware releases from Intel or Apple or anyone else. I think that in a few years we're going to look back at this specific moment......the release of the first three Apple Silicon Macs......as a major milestone for Apple and computing in general. Even if Apple has made the iPad AND Mac obsolete with some kind of new device that acts as both by then, we'll still be able to trace that product development back to Apple Silicon Macs, which can then of course be traced back to iPhones, iPads, and Apple Watches having Apple Silicon even earlier.

All that is to say that in my opinion, Apple has done the equivalent of opening the gate for only their own race horse while Intel, AMD, and everyone else are still trying to WD40 theirs loose. Instead of waiting for the starting gun, they shot it themselves and are at least a half lap ahead already.

EDIT: Sorry--also meant to comment on someone else's point about most people not caring about Apple Silicon--that's definitely true but does not mean anything at all bad. The most significant reason people don't know is because they don't care to know. They just know whenever they buy a new Mac it works the way their old one did (more or less) and if it doesn't, Apple will help them out. That's all the average user needs to know. If average users were all over message boards talking about M1 right now, it would most likely be because Apple fumbled this transition, which they most certainly did not.

 

[Jorbanead]

This is enough to cover the mobile use case, but probably not the desktop one. There is still a significant difference in cooling capacity between a MBP (which is around 80watts) and a large iMac or a Mac Pro. And while I have little doubt that Apple can build a ridiculously fast laptop with the thermal budget of 80Watts, that chip won’t be enough to replace the current desktop offers. Desktop chips have to be larger, otherwise Mac desktop is essentially dead.

I think the M1X chip will be scalable just like the M1 chip. So putting the M1X in a MacBook Pro 16” it may be something like a 45 watt chip, whereas the iMac they may push the chip closer to 65 or 75 watts. Those are just random guesses, but the point is you can use the same chip and just allow it more thermal headroom and wattage to bump up the performance.

They can also bin those chips, so there may be a 12-core M1X and then a 16-core M1X as well. Same chip, just the 12-core variant has 4 cores disabled and is meant for notebooks and the higher core count is desktop. Similar to how they have disabled a GPU core for the MacBook Air.

The M1z will be a 32 core variant but they may sell a 24 and 32 core version both called M1Z. You’ll just see 32-core M1Z in the processors specs.

 

[Jorbanead]

Going back to the OP's original question: I'm not sure it makes sense anymore to even try to put together a pattern or predict any of this based on past hardware releases from Intel or Apple or anyone else.

I think it is safe to make a few general guesses.

1) They’ll likely announce an Apple silicon MacBook Pro 14”, MacBook Pro 16”, and iMac 24”, and an iMac 27” this year. (Sizes TDB, but you get the point).

2) any new Apple silicon Mac is for sure being announced at an event. They won’t update thr MacBook Pro with AS as just a press release

2) they generally release new generation macs either in May, June, or October. That’s not necesssirly because of Intel. Those are just the best times to release products en mass.

Based off of recent rumors it looks like we’re getting the iMac updates at WWDC in June, and MacBook Pro updates in the fall. They could surprise us a split up the iMac update with the small one being announced in March, and big one announced at WWDC or in the fall.

The wildcard is less apple silicon and more virtual events. Before they were locked into specific times of year for the press to attend, but now they could do an event each month if they wanted to.

 

[leman]

I think the M1X chip will be scalable just like the M1 chip.

Maybe you are right, just looked at the iMac and it's using the same GPU chip (full 40 CU Navi 10) as the MacBook Pro... just with 50% higher clock.

The question is how overclockable Apple chips are. M1 is actually not that scalable — the chips in the MBA run the same frequencies as the chips in the Mini, its just about the thermal headroom. In other words, M1 scales down (from it's nominal 20-25W), but it doesn't really scale up. But if it can be clocked higher, I can certainly see the same say 16 CPU and 32 GPU part be used in both a 16" MBP (as a 70-80W part) and a 27" iMac (as a 120-160W part). TDP kind of checks out when we assume max 5-6 watts for a CPU core and 3 watts for a GPU core at it's peak.

Still, the Mac Pro would need a separate, much larger chip if Apple wants to trump the performance of the Intel model. My idea is actually that iMac chips could be heavily binned partially defect Mac Pro chips, with 50% or more of the chip area disabled in order to improve yields. It won't be cheap, but Apple could still retain healthy profit margins and build some really powerful chips this way.

 

[nothingtoseehere]

"M1X" has become the common placeholder for the upcoming chips but I would like to suggest a different wording with plusses, i.e. M1 - M1+ - M1++ etc.
I could imagine that Apple will have four types:

• M1: MBA, Mac mini, as known
• M1+: MBP 14, smaller iMac, "Mac mini Pro": more cores, I/O, RAM
• M1++: MBP 16, bigger iMac, something like "Mac Cube pro mini": even more of everything. Maybe with upgradeability, with slots to add RAM and Apple-made cards with more cores of either GPU, CPU or neural?
• M1+++: Mac Pro (probably only with the M2-architecture around the end of 2022)

 

[antwormcity]

"M1X" has become the common placeholder for the upcoming chips but I would like to suggest a different wording with plusses, i.e. M1 - M1+ - M1++ etc.
I could imagine that Apple will have four types:

• M1: MBA, Mac mini, as known
• M1+: MBP 14, smaller iMac, "Mac mini Pro": more cores, I/O, RAM
• M1++: MBP 16, bigger iMac, something like "Mac Cube pro mini": even more of everything. Maybe with upgradeability, with slots to add RAM and Apple-made cards with more cores of either GPU, CPU or neural?
• M1+++: Mac Pro (probably only with the M2-architecture around the end of 2022)

Interesting way to look at upgrade capability.

I would be surprised if Apple went into the business of additional RAM expandability for notebooks at least, their unified architecture means everything is proprietary - those bandwidths for blistering performance are probably not possible with plug/play devices, its baked into the SoC design from the start - Storage, RAM, ML engines, TPUs and what not. Mac Cube is a far fetched thought unless driven purely by aesthetics - we don't need more space moving away from Intel era, what's it going to house... (empty space perhaps? thermals are solved with the M1 much better than anything in the market)

Apple would not go into computer peripherals "parts" business with selling their own RAM modules, or add on processing power with cores or secondary complete chips - its not their business direction for anything they make - its a use and throw and trade-in philosophy asking you to upgrade often. iPhone and Macbook prices are getting blurry with iPhone <> MAX approaching M1 MBP price. I am more or less prepared for a non upgradable config for all Macs except maybe the iMac pro or Mac Pro. Notebooks are out of question for upgrade options - BTO and that's it - true since my 2013 rMBP 15".

 

[leman]

I would be surprised if Apple went into the business of additional RAM expandability for notebooks at least, their unified architecture means everything is proprietary - those bandwidths for blistering performance are probably not possible with plug/play devices, its baked into the SoC design from the start - Storage, RAM, ML engines, TPUs and what not.

I agree that probability of RAM expandability on laptops is basically zero. With space and power consumption being extremely important factors, I don't see how replaceable RAM would be possible.

On desktops... DDR5 brings a decent increase in speed, so one could entertain a possibility of socketed RAM. At the same time, Apple would require a large number of parallel channels (6 to 8 for mid-range desktop), with RAM slots being placed as close as possible to the SoC itself, resulting in some rather awkward logic board design and some really awkward slot access. Not to mention that upgrading RAM on such a system would be non-trivial and also really expensive as you'd need a bunch of exactly matching memory modules... so yeah... I don't think we will see upgradeable RAM on desktops either. Maybe some sort of custom RAM module, but given the extreme pinout one would need, I very much doubt that they would be user-accessible.

 

[Jorbanead]

"M1X" has become the common placeholder for the upcoming chips but I would like to suggest a different wording with plusses, i.e. M1 - M1+ - M1++ etc.
I could imagine that Apple will have four types:

• M1: MBA, Mac mini, as known
• M1+: MBP 14, smaller iMac, "Mac mini Pro": more cores, I/O, RAM
• M1++: MBP 16, bigger iMac, something like "Mac Cube pro mini": even more of everything. Maybe with upgradeability, with slots to add RAM and Apple-made cards with more cores of either GPU, CPU or neural?
• M1+++: Mac Pro (probably only with the M2-architecture around the end of 2022)

This would deviate from Apples naming scheme from the A series. The reason why people assume the next chip is M1X is because Apple has consistently done an X variant of the A series line for years used in the iPad and iPad Pro. They also did a Z variant one year too which is why some think the third tier of chip would be M1Z.

So imo it would be odd to use the X variant on the A series of chip, but then switch to a plus variant on the M series of chips.

This is how I see the line (sizes TBD):

Notebooks:
MacBook Air 13” - M1
MacBook Pro 13” - M1
MacBook Pro 14” - M1X
MacBook Pro 16” - M1X

Desktops:
Mac mini silver - M1
Mac mini grey - M1X
iMac 24” - M1X
iMac 27” - M1Z
Mac Pro - M1Z

(Notice it says “up to # cores” as not every model will have that many cores depending on thermal envelope).

M1 - Entry Level Chips
Up to 8 CPU and 8 GPU cores
Up to 16gb of ram
Up to 2 TB controllers
Up to 2 displays

M1X - Prosumer Level Chips
Up to 16 CPU and 32 GPU cores
Up to 64gb of ram
Up to 4 TB controllers
Up to 4 displays

M1Z - Professional Level Chips
Up to 32 CPU and 128 GPU cores
Ram unknown
Up to 8 TB controllers
Displays unknown

I agree that probability of RAM expandability on laptops is basically zero. With space and power consumption being extremely important factors, I don't see how replaceable RAM would be possible.

On desktops... DDR5 brings a decent increase in speed, so one could entertain a possibility of socketed RAM. At the same time, Apple would require a large number of parallel channels (6 to 8 for mid-range desktop), with RAM slots being placed as close as possible to the SoC itself, resulting in some rather awkward logic board design and some really awkward slot access. Not to mention that upgrading RAM on such a system would be non-trivial and also really expensive as you'd need a bunch of exactly matching memory modules... so yeah... I don't think we will see upgradeable RAM on desktops either. Maybe some sort of custom RAM module, but given the extreme pinout one would need, I very much doubt that they would be user-accessible.

Apples RAM is on-package though which bypasses any sort of PCIe bus. One of the major selling points of the M-series so far is unified memory between GPU and CPU. Apple has done this by incorporating the memory on the chip package itself rather than using slower PCIe busses on the motherboard like Intel macs. So I’m not sure how apple would handle this while still sticking to their unified memory on package. I don’t know if using a PCIe bus would work.

 

[Mac... nificent]

I don't see how replaceable RAM would be possible.

Or (for most people) needed. RAM in the Apple Silicon SoC works much differently than "intel" style RAM.

For example; I have a 2018 Mac mini with 32 Gigs RAM. My M1 with "8" Gigs outperforms it. This is because the SoC has system memory and is much more efficient.

As far as the original question of will the M1x be released this year or not, I feel that it will, probably Oct/Sept. if not sooner. The reason why I feel it might even be sooner is because they had no reason to discontinue the iMac Pro now, if they weren't going to replace it with a more powerful SoC version.

 

[Maconplasma]

Or (for most people) needed. RAM in the Apple Silicon SoC works much differently than "intel" style RAM.

For example; I have a 2018 Mac mini with 32 Gigs RAM. My M1 with "8" Gigs outperforms it. This is because the SoC has system memory and is much more efficient.

While I do not own an M1 Mac I have to say that many people here will not agree with you about this. Many here with M1's say it's nowhere close to the truth that 8GB Ram performs even as a good 16GB Intel, let alone 32GB.

 

[Mac... nificent]

While I do not own an M1 Mac I have to say that many people here will not agree with you about this. Many here with M1's say it's nowhere close to the truth that 8GB Ram performs even as a good 16GB Intel, let alone 32GB.

They are entitled to their own opinion. I have to go by what I see right in front of me, and what my clients have seen.

There are also videos online showing what I'm talking about so I know I'm not alone feeling this way

 

[JouniS]

For example; I have a 2018 Mac mini with 32 Gigs RAM. My M1 with "8" Gigs outperforms it. This is because the SoC has system memory and is much more efficient.

Memory is not as much about performance as about the ability to run software. If you don't have enough memory, software may refuse to start, or it may simply crash. Or even worse, the OS may choose to overcommit and provide memory it doesn't have. That may result in heavy swapping, which could mean writing up to tens of terabytes per day with current SSD speeds. Consumer-grade SSDs can't handle that for very long.

 

[Maconplasma]

They are entitled to their own opinion. I have to go by what I see right in front of me, and what my clients have seen.

There are also videos online showing what I'm talking about so I know I'm not alone feeling this way

So you're discounting what others here with M1's are saying? I mean, just like you they have an M1 Mac in front of them. If you're going to use the word "Opinion" then it's just as much your opinion as it is theirs. You can't say your experience is the only one that counts, sans for a few YouTube videos. That's a pretty condescending attitude.
What your clients have seen? Do you sell Mac computers to them?

 

[jerryk]

While I do not own an M1 Mac I have to say that many people here will not agree with you about this. Many here with M1's say it's nowhere close to the truth that 8GB Ram performs even as a good 16GB Intel, let alone 32GB.

Depends on the task. When comparing many browser and Office-based tasks, my M1 Air with 8GB is faster than my 16" MBP with 1TD SSD, and 32 GB RAM was.

 

[Jorbanead]

Or (for most people) needed. RAM in the Apple Silicon SoC works much differently than "intel" style RAM.

For example; I have a 2018 Mac mini with 32 Gigs RAM. My M1 with "8" Gigs outperforms it. This is because the SoC has system memory and is much more efficient.

As far as the original question of will the M1x be released this year or not, I feel that it will, probably Oct/Sept. if not sooner. The reason why I feel it might even be sooner is because they had no reason to discontinue the iMac Pro now, if they weren't going to replace it with a more powerful SoC version.

While I do not own an M1 Mac I have to say that many people here will not agree with you about this. Many here with M1's say it's nowhere close to the truth that 8GB Ram performs even as a good 16GB Intel, let alone 32GB.

The difference from Intel is 1) lower memory latency due to on package ram and 2) faster memory swapping from the much faster SSD

So 8gb for *some* apps can be as efficient as 16gb but it may also not be. Both your experiences are valid and it depends on how the program uses ram and what can be swapped.

The main thing is that 8gb on M1 does not equate to 16gb on Intel. Ram is still ram. How M1 manages that is what is different.

 

[Fomalhaut]

Or (for most people) needed. RAM in the Apple Silicon SoC works much differently than "intel" style RAM.

For example; I have a 2018 Mac mini with 32 Gigs RAM. My M1 with "8" Gigs outperforms it. This is because the SoC has system memory and is much more efficient.

As far as the original question of will the M1x be released this year or not, I feel that it will, probably Oct/Sept. if not sooner. The reason why I feel it might even be sooner is because they had no reason to discontinue the iMac Pro now, if they weren't going to replace it with a more powerful SoC version.

I'm not sure why you think M1 memory is "system" memory and Intel memory is not part of the system or vastly different in concept.

The M1 memory is fast (4266 MT/s)

It's located on the SoC package so has shorter / more efficient inter-connections to the memory controllers on the die.

It is fully shared with the integrated GPU - this is not that dissimilar to Intel iGPUs although I think the M1 memory sharing is more efficient (maybe uses the whole addressing space to prevent the need to copy?).

 

[Fomalhaut]

Memory is not as much about performance as about the ability to run software. If you don't have enough memory, software may refuse to start, or it may simply crash. Or even worse, the OS may choose to overcommit and provide memory it doesn't have. That may result in heavy swapping, which could mean writing up to tens of terabytes per day with current SSD speeds. Consumer-grade SSDs can't handle that for very long.

Exactly, and this is the topic of the (currently) 43-page thread on excessive SSD write volumes. Even with 16GB my M1 Mini is writing at least 250GB per day to the SSD when I don't treat it with kid-gloves, by closing apps & tabs throughout the day in order to keep swap usage low.

My 32GB MBP16 has already written 170TB to the SSD after 15 months, which implies it is probably using swap memory very regularly.

There is nothing magical about M1 RAM other than it being quite fast and efficient due to it being on the SoC package. It is shared with the GPU so if you run high-resolution screens or hammer your GPU you are going to run short of RAM if you have many (or demanding) apps running.

I think my "natural" usage to not use much swap is probably about 40GB+. I have a 64GB Windows workstation that always seems to have spare RAM, but my 32GB Mac laptop still swaps quite a bit.

 

[leman]

Or (for most people) needed. RAM in the Apple Silicon SoC works much differently than "intel" style RAM.

For example; I have a 2018 Mac mini with 32 Gigs RAM. My M1 with "8" Gigs outperforms it. This is because the SoC has system memory and is much more efficient.

As others mentioned, the rumors of Apple Silicon RAM having magical properties are greatly exaggerated. Apple chips are simply.... faster, and in some ways much faster (for example M1 can sustain almost 60GB/s memory bandwidth to a single core where Intel chips — even Xeons — are stuck with 20GB/s), and it's really efficient at memory compression and swapping.

At the end of the day, RAM is RAM. There is a certain amount of working memory that an application need to operate efficiently. This amount is usually much lower than what people claim (I shake my head at how many people claim that 16GB is necessary for basic home and office use), but it can be blown up by the number of apps people use simultaneously. However, if your system is efficient at juggling memory (like M1 is), you won't notice RAM being relocated between active application as you switch between your windows. That is basically the essence of M1's "magic" — it creates the illusion of needing less RAM because it's more agile in assigning RAM to where it's needed.