Apple's M2 chip - what to expect from the next Apple Silicon evolution

AppleInsider

Apple has yet to make the "M2" processor official, but details have already started to emerge about what to expect in Apple's second-generation Apple Silicon chip.

Apple's rumored M2 chipset

Likely arriving later this year, Apple's M2 chip will be the successor to Apple's first Mac-specific chip -- the M1. Apple did release two variants on the M1, dubbed the M1 Pro and M1 Max that made their way to the redesigned MacBook Pro, which will still be more capable than the M2.

While we await Apple's announcement, here's what we expect to see from the M2.

New chip, better graphics

The M2 will likely take after the M1, utilizing a 5nm process and equipped with an eight-core CPU design. Rumors have said that the M2 will have a slightly higher clock speed on its cores, though not a dramatic uptick.

Apple's M1 is already a capable CPU, so a marginal speed increase isn't surprising.

Despite the meager CPU improvements, Apple looks to be beefing up the graphics. The new M2 will likely move from the currently-available seven or eight cores to a nine-core GPU with a ten-core option.

M2 compared to the M1 Pro and M1 Max

With the M2 increasing in speed and graphics performance, many inquiring minds will be interested in comparing it to Apple's high-end chipsets.

Fortunately, the M1 Pro and M1 Max will be more capable than the rumored M2. Both M1 Pro and M1 Max still have ten-core CPUs, two more than the rumored M2. The M1 Pro can have either a 14 or 16-core GPU, while the M1 Max can have either a 24 or 32-core GPU.

The M1 Pro and M1 Max have other benefits compared to the M1, and some of these will likely carry over to the M2. The M1 Pro and M1 Max have higher memory bandwidths and the ability to support more memory. The M1 Max can handle up to 64GB memory, while the M1 is capped at 16GB.

The M1 Pro and M1 Max have dedicated video encoders for video production. There is the possibility Apple may bring these to the entry-level M2. However, it seems more likely the company will retain these for the high-end chips.

When will we see the M2?

We expect Apple to release the M2 in the second half of 2022. Apple will probably continue its rollout of machines with the M1 Pro and M1 Max. Namely the rumored 27-inch iMac and high-end Mac mini.

Then we predict Apple will release the newly-redesigned MacBook Air, entry-level 14-inch MacBook Pro, and possibly a new entry-level Mac mini.

It seems likely that the new iPad Pro will gain the M2 as well, just as the current generation has the M1.

Of course, all of this is subject to any number of delays. None of this is set and stone and purely based on rumors. While we believe this is what's most likely at the moment, only time will tell what Apple's plans are.

Read on AppleInsider

tommikele

This article says absolutely nothing.

Sunday filler of zero value.

algr

This is more about naming conventions then actually what is in the chips.  Customers are likely to assume that anything "M2" must be faster than any "M1" chip.  But Apple ultimately decides what the names actually mean, and might do things differently.  The M2 described in this article just sounds like something between the M1 and M1 pro.  

9secondkox2

M2 based on a15 or a16. 

Signifying next generation. 

A15 is marginally faster than a14, so m2 marginally faster than m1. Should add up to a nice multi ore and graphics boost. 

Looking for war to the m2 max. 

blastdoor

I had assumed M2 would use same core design as A15, but if it’s not coming until second half, maybe A16 cores?

mjtomlin

lmasanti said:

First of all, we need to know how frequently will Apple renew the M series-
With quite lower volume… will it make sense to a yearly release sequence? ¡
I think Apple will have like more than a 2-3 years cycle.

I think the base chips M1, M2, M3 will be a yearly upgrade (after the transition is finished) and the “pro” variants will be updated every 2 years or so.

Volume shouldn’t be as big an issue due to the fact that these chips share the same cores as the A-series.

bulk001

If this was said about Intel instead of Apple there would be howls of mockery here:

“Rumors have said that the M2 will have a slightly higher clock speed on its cores, though not a dramatic uptick.

Apple's M1 is already a capable CPU, so a marginal speed increase isn't surprising.

Despite the meager CPU improvements, Apple looks to be beefing up the graphics. The new M2 will likely move from the currently-available seven or eight cores to a nine-core GPU with a ten-core option.”

If true then one generation in and they are essentially where they were with Intel plus AMD for better graphics. Incremental updates each year presented as an “amazing new graphics chip” giving the minimum they can get away with to keep updates ticking. Hopefully the rumors are wrong and M2 blows off the doors in every way. 

tenthousandthings

While technically correct, it doesn’t seem quite fair to say the M2 will use the same 5 nm process as the M1. The A14 and the M1 family use the N5 process. The A15 uses the N5P process, which is the first enhancement of the original 5 nm process (N5). Since then TSMC has announced two additional enhancements of that process, N4 and N4P. The A16 and the M2 will likely use one or both of these.

So the A16 will tell us more about the difference between the M1 cores versus the M2 cores than the A14 versus A15 comparison, which is what is driving the current tendency to say there won’t be a significant bump. 

tht

I think the M2 will be faster than the M1 by about +15% in GB5 single, +25% in GB5 multi and +40% in GB5 Metal, and maybe +50% on neural unit oerformance. So, about 2000 points in GB5 single, 10k GB5 multi, and 30k in GB5 Metal. About as fast as the 6+2+14 M1 Pro, faster in single core, slower in graphics. And with a move to LPDDR5, 32 GB RAM capacity using 16 GB packages is possible, as well as 7 GB/s NAND. And as Tenthousandthings says, fabbed on TSMC N4.

It really don't take much to bump single core performance by 15%. Faster RAM, larger SLC, 5% faster IPC, 5% faster clock. For multi-core, they can do 6+2 p-core to e-core config, declock the p-cores, and call it a day. If it is 4+4, maybe 9000 points in GB5 multi which is pretty good too. For GPU, they already have 5 GPU cores in the A15. All it takes is 15% g-core improvement and 10 g-cores to get a +40% improvement over the M1 GPU, plus the 50% bandwidth improvement from LPDDR5 and larger SLC.

danox

The iMac and the desktop Mac need to move on to the M series chips as soon as possible, the last cpu switch took 13 months…..

hucom2000

Graphics is the only real differentiator between the M series and Pro and Max variants. They are going to keep it that way. There’s not much else they can do to differentiate.

godofbiscuits

9secondkox2 said:

M2 based on a15 or a16. 

Signifying next generation. 

A15 is marginally faster than a14, so m2 marginally faster than m1. Should add up to a nice multi ore and graphics boost. 

Looking for war to the m2 max. 

M-series aren't "based on" A-series or vice versa.  They each share IP from a given generation.   Saying otherwise just causes confusion.

mattinoz

tht said:

I think the M2 will be faster than the M1 by about +15% in GB5 single, +25% in GB5 multi and +40% in GB5 Metal, and maybe +50% on neural unit oerformance. So, about 2000 points in GB5 single, 10k GB5 multi, and 30k in GB5 Metal. About as fast as the 6+2+14 M1 Pro, faster in single core, slower in graphics. And with a move to LPDDR5, 32 GB RAM capacity using 16 GB packages is possible, as well as 7 GB/s NAND. And as Tenthousandthings says, fabbed on TSMC N4.

It really don't take much to bump single core performance by 15%. Faster RAM, larger SLC, 5% faster IPC, 5% faster clock. For multi-core, they can do 6+2 p-core to e-core config, declock the p-cores, and call it a day. If it is 4+4, maybe 9000 points in GB5 multi which is pretty good too. For GPU, they already have 5 GPU cores in the A15. All it takes is 15% g-core improvement and 10 g-cores to get a +40% improvement over the M1 GPU, plus the 50% bandwidth improvement from LPDDR5 and larger SLC.

Could they get there with 4p+ 2e and not declocking the p-cores?
Use the space for better bandwidth generally, more RAM support and more GPU Cores on this "base" SOC.
The larger chips are then on-die modular starting with the same base and adding 4p + 16g each step. Slicing all the variations out of the wafer based on testing?

doctorq

What to expect from Apple’s M2 chip?
More useless speculation.

bageljoey

Did the rumors say that the M2 will support two external monitors?

georgebmac

As always, it's the software and ecosystem that do the work.

Apple needs to let them catch-up to the M1 before making M2, M3, M4, etc... super chips.  Otherwise, with the exception of a few power users, their speed and power will be wasted.

dewme

In my opinion it seems backward thinking to assume that Apple will simply follow the same kind of incremental spec bumping on a predefined schedule that Intel followed with their line of CPUs. There is no reason why Apple would arbitrarily release updated SoCs every 12 months, 24 months, or whatever time increment you want to speculate about for no reason other than the calendar flipping to a new month or year. This is "Intel's Way of Looking at the World," i.e., milking the cash cow, not Apple's way of looking at the world. 

Rather, I expect that Apple would release new custom SoCs on a timeline that matches their desire to deliver new functionality and unique functionality into a product or product line. They control the whole stack and thus can build problem-domain-specific, purpose built, SoCs that bring the optimal set of capabilities to bear against the problems that each product or product line is going after, all while maintaining a common core that delivers a base level of cross product capabilities. They've already done this with their A-series SoCs when you look across iPhone, iPad, Apple TV, and Apple Watch.

All I'm saying is that Apple can fine tune their SoC designs, and their release schedules, to their hearts content because they are not building commodity components intended for general consumption. I don't see them following the same release model that commodity chip makers have followed where they have to release newer, faster, and marginally better chips on a calendar based increment just to keep everyone buying their latest stuff.

The Apple Silicon team only has to meet the demands of Apple's product teams, which in many cases will be far more strenuous than anything that Apple ever expected from Intel, Qualcomm, or whatever commodity chip supplier they were dependent on in the past. Apple is a product company, not a chip company. Apple's product designs will push Apple's chip designs, not the other way around.

muthuk_vanalingam

dewme said:

In my opinion it seems backward thinking to assume that Apple will simply follow the same kind of incremental spec bumping on a predefined schedule that Intel followed with their line of CPUs. There is no reason why Apple would arbitrarily release updated SoCs every 12 months, 24 months, or whatever time increment you want to speculate about for no reason other than the calendar flipping to a new month or year. This is "Intel's Way of Looking at the World," i.e., milking the cash cow, not Apple's way of looking at the world. 

Rather, I expect that Apple would release new custom SoCs on a timeline that matches their desire to deliver new functionality and unique functionality into a product or product line. They control the whole stack and thus can build problem-domain-specific, purpose built, SoCs that bring the optimal set of capabilities to bear against the problems that each product or product line is going after, all while maintaining a common core that delivers a base level of cross product capabilities. They've already done this with their A-series SoCs when you look across iPhone, iPad, Apple TV, and Apple Watch.

All I'm saying is that Apple can fine tune their SoC designs, and their release schedules, to their hearts content because they are not building commodity components intended for general consumption. I don't see them following the same release model that commodity chip makers have followed where they have to release newer, faster, and marginally better chips on a calendar based increment just to keep everyone buying their latest stuff.

The Apple Silicon team only has to meet the demands of Apple's product teams, which in many cases will be far more strenuous than anything that Apple ever expected from Intel, Qualcomm, or whatever commodity chip supplier they were dependent on in the past. Apple is a product company, not a chip company. Apple's product designs will push Apple's chip designs, not the other way around.

And your point is? Is it that Apple won't stick to a 18 month schedule or 24 month schedule for the M series SoCs? If you think so, it is not based on reality. Apple is designing the M series SoCs for matured product lines and they cannot afford to have random schedules for those matured product lines. The customers would like to upgrade at regular intervals. And Apple would be better off providing clarity to people on when they can expect the next product in the lineup instead of keeping them clueless about the future of the lineup (be it MacBooks or Mac Mini or iMacs).

crowley

dewme said:

In my opinion it seems backward thinking to assume that Apple will simply follow the same kind of incremental spec bumping on a predefined schedule that Intel followed with their line of CPUs. There is no reason why Apple would arbitrarily release updated SoCs every 12 months, 24 months, or whatever time increment you want to speculate about for no reason other than the calendar flipping to a new month or year. This is "Intel's Way of Looking at the World," i.e., milking the cash cow, not Apple's way of looking at the world. 

Sure, because Apple release iPhones any time they damn well want, not on any kind of easily observable schedule.

dewme

muthuk_vanalingam said:

dewme said:

In my opinion it seems backward thinking to assume that Apple will simply follow the same kind of incremental spec bumping on a predefined schedule that Intel followed with their line of CPUs. There is no reason why Apple would arbitrarily release updated SoCs every 12 months, 24 months, or whatever time increment you want to speculate about for no reason other than the calendar flipping to a new month or year. This is "Intel's Way of Looking at the World," i.e., milking the cash cow, not Apple's way of looking at the world. 

Rather, I expect that Apple would release new custom SoCs on a timeline that matches their desire to deliver new functionality and unique functionality into a product or product line. They control the whole stack and thus can build problem-domain-specific, purpose built, SoCs that bring the optimal set of capabilities to bear against the problems that each product or product line is going after, all while maintaining a common core that delivers a base level of cross product capabilities. They've already done this with their A-series SoCs when you look across iPhone, iPad, Apple TV, and Apple Watch.

All I'm saying is that Apple can fine tune their SoC designs, and their release schedules, to their hearts content because they are not building commodity components intended for general consumption. I don't see them following the same release model that commodity chip makers have followed where they have to release newer, faster, and marginally better chips on a calendar based increment just to keep everyone buying their latest stuff.

The Apple Silicon team only has to meet the demands of Apple's product teams, which in many cases will be far more strenuous than anything that Apple ever expected from Intel, Qualcomm, or whatever commodity chip supplier they were dependent on in the past. Apple is a product company, not a chip company. Apple's product designs will push Apple's chip designs, not the other way around.

And your point is? Is it that Apple won't stick to a 18 month schedule or 24 month schedule for the M series SoCs? If you think so, it is not based on reality. Apple is designing the M series SoCs for matured product lines and they cannot afford to have random schedules for those matured product lines. The customers would like to upgrade at regular intervals. And Apple would be better off providing clarity to people on when they can expect the next product in the lineup instead of keeping them clueless about the future of the lineup (be it MacBooks or Mac Mini or iMacs).

My first point is that they don't have to stick to a predefined release cadence for the Apple Silicon components. This is not to say that they won't stick to a predefined release schedule for certain products, like the iPhone, even if it means releasing year over year incremental updates that some folks are starting to get a bit less than excited about. Yeah they can always add another camera lens. Woo hoo. So I'm not saying they won't stick to a predefined cadence, I'm saying they don't have to, at least not across their whole array of products.

My second point is that they can diversify the configurations of their SoCs, such that the M-series chip in a MacBook Pro is not outfitted exactly the same way (core types, memory, neural engines, etc.) as the same generation M-series SoC in a Mac Pro, iMac, iPad, a future AR headset. They can custom tailor their own chips to better match the specific requirements of their product groups.

I'm not trying to be contrary, I'm just pointing out that the Apple Silicon team isn't simply an "Intel replacement" for Apple. Apple having total control over their chip designs and their release schedules affords them a lot more flexibility over how they bring new and upgraded products to market. How they choose to utilize this additional flexibility is totally within Apple's control. Considering the investment required to give them this capability, one would hope that they use it to great advantage and not feel constrained to how they managed their relationship with Intel, which obviously wasn't working. 

tenthousandthings

crowley said:

dewme said:

In my opinion it seems backward thinking to assume that Apple will simply follow the same kind of incremental spec bumping on a predefined schedule that Intel followed with their line of CPUs. There is no reason why Apple would arbitrarily release updated SoCs every 12 months, 24 months, or whatever time increment you want to speculate about for no reason other than the calendar flipping to a new month or year. This is "Intel's Way of Looking at the World," i.e., milking the cash cow, not Apple's way of looking at the world. 

Sure, because Apple release iPhones any time they damn well want, not on any kind of easily observable schedule.

I don't think Dewme's basic observation can be quite so easily dismissed. There is nothing in Apple's extensive Macintosh history to suggest they will suddenly change their ways and start refreshing computers like clockwork, like they do for phones. The iPhone is the anomaly here, a product of competitive pressures and sales volumes that don't really translate into the PC realm.

I get the argument for it, but the idea that Apple is going to suddenly start adhering to a regular, predictable refresh and/or redesign schedule for Macs is something that has never been the case in the past. Expecting that to suddenly change just because Apple is designing its own silicon may or may not be a reasonable expectation. But it's definitely not yet "reality" (to quote Muthuk_vanalingam's response to Dewme above).