Apple's iPhone processor evolution hints at how powerful the 'M2' will be

Posted:
in General Discussion edited February 2022
We don't know how powerful or power-efficient Apple's rumored "M2" chip will be, but past Apple Silicon generational leaps could provide a hint, a new report claims.

M2 chip illustration
M2 chip illustration


Apple is largely expected to debut an "M2" Mac chip later in 2022. Though there are some rumors concerning the upgrades it'll bring to the Mac, a MacWorld writer believes it's possible to get a clearer idea at how much of an upgrade the "M2" will be by examining Apple's A-series chips.

The report starts with the premise that the M2 chip will be akin to what the M1 as the A15 Bionic was to the A14 Bionic. From there, he extrapolates how the "M2" could look based on a similar jump.

"The M1 is based on the basic architecture of the A14, scaled up with double the number of high-performance CPU cores (four instead of two) and double the GPU cores (eight instead of four)," he wrote. "It is, for lack of a better explanation, the 'A14X' that otherwise would've been in the iPad Pro."

Basically, the argument is that the next-generation chip will sport double the number of high-performance cores and double the GPU cores.

"In trying to predict the M2's features and performance, we will operate on the assumption that it will follow a similar pattern, only this time with the A15: the same architecture with twice the high-performance cores and GPU cores," Cross continued.

Using the A14 and A15 as a base, MacWorld then extrapolated how much of a performance jump the new "M2" could bring to the Mac. According to benchmark estimates, it could be a significant jump.

As the piece points out, a score over 14,000 for the "M2 Max" could put it well ahead of Intel's latest and greatest Alder Lake Core i9. Of course, the "M2 Max" will also likely be much more power efficient.

Additionally, the extrapolated data also shows significant gains in graphics performance, including a 10,409 score on 3DMark, up from 7,781 on the M1.

MacWorld notes that it didn't account for any performance gains from improved manufacturing processes, since there isn't a way to extrapolate or quantify what that could look like.

Of course, the estimates are just rough predictions. Because of that, the estimated scores and gains should be taken with a hefty grain of salt.

Whether or not the performance on the "M2" looks anything like the extrapolated data remains to be seen, as the architectures are not identical, nor are the intended use cases.

Read on AppleInsider

Comments

  • Reply 1 of 20
    Most are guessing M2 will be an enhanced M1 using ARMv9.
    killroywatto_cobra
  • Reply 2 of 20
    blastdoorblastdoor Posts: 2,918member
    mtanika said:
    Most are guessing M2 will be an enhanced M1 using ARMv9.
    "most"? I hadn't heard that, other than a few random speculators in comments around here. 

    I'm skeptical it will be ARMv9, and I think it's very possible Apple will never use ARMv9, simply because they might have no need to. 

    I think the MacWorld article is pretty reasonable and if it's wrong, the most likely way to be wrong is for the M2 to be based on the A16 rather than the A15. 
    Prometheutenthousandthingsmuthuk_vanalingamgeekmeewatto_cobra
  • Reply 3 of 20
    I don't know the particulars of Apple's ARM architecture license, but if royalty payments for older versions decline over time, or if payments for new versions increase over time, there's financial incentive for Apple to drift away from ARM into its own custom designs. Whether the ARM design teams can keep Apple interested is an open question.
    Beatswatto_cobra
  • Reply 4 of 20
    Those are decent gains considering the A15 is pretty long in the tooth for the 5nm process. I doubt they will produce an M2 Pro/Max chip at all. The M1 Pro/Max is still plenty impressive, and there are still machines yet to come that will feature these chips and variants of them.

    Probably Apple will not update the M1 Pro/Max until the 3nm process in late 2023.
    watto_cobra
  • Reply 5 of 20
    blastdoor said:

    I think the MacWorld article is pretty reasonable and if it's wrong, the most likely way to be wrong is for the M2 to be based on the A16 rather than the A15. 
    Yes, the author sidestepped that by saying the M2 will be on the N4P process and telling us how much of an improvement that is expected to be over the (N5) process of the A14 & M1, but he does it without mentioning that the A15 is on the N5P process, not the N4P, so his comparison between A14 and A15, while useful, only goes so far.

    The other thing that I’d quibble with is the assumptions he makes about Apple’s ability to compete with high-end desktop GPUs — I mean, that remains to be seen. I don’t think it’s coincidental that Apple announced a ray-tracing Metal API at WWDC 2020, with development active and ongoing. You might want to wait until Apple releases an actual Pro desktop M1 graphics configuration before you start making assumptions about how far behind the competition they are… 
    williamlondonwatto_cobra
  • Reply 6 of 20
    melgrossmelgross Posts: 33,334member
    blastdoor said:
    mtanika said:
    Most are guessing M2 will be an enhanced M1 using ARMv9.
    "most"? I hadn't heard that, other than a few random speculators in comments around here. 

    I'm skeptical it will be ARMv9, and I think it's very possible Apple will never use ARMv9, simply because they might have no need to. 

    I think the MacWorld article is pretty reasonable and if it's wrong, the most likely way to be wrong is for the M2 to be based on the A16 rather than the A15. 
    The belief as to why Apple didn’t use arm v 9 is because much of that is from Apple that got rolled into v 9, so Apple didn’t need it.

    if the A16 isn’t out until September, it will be difficult for the M2 to be based on it as we’re expecting at least one Mac before that. I’d love to see that though, but I don’t think the timing will allow it. Apple needs to work on these chips long before it’s ready. And then there’s the n 4 process step.
    watto_cobra
  • Reply 7 of 20
    melgrossmelgross Posts: 33,334member
    bsimpsen said:
    I don't know the particulars of Apple's ARM architecture license, but if royalty payments for older versions decline over time, or if payments for new versions increase over time, there's financial incentive for Apple to drift away from ARM into its own custom designs. Whether the ARM design teams can keep Apple interested is an open question.
    Apple doesn’t use ARM designs, so they don’t pay for them. Apple’s designs are entirely home grown. They have an architectural license which means they use the instruction set, basically.
    jroywatto_cobra
  • Reply 8 of 20
    melgrossmelgross Posts: 33,334member
    blastdoor said:

    I think the MacWorld article is pretty reasonable and if it's wrong, the most likely way to be wrong is for the M2 to be based on the A16 rather than the A15. 
    Yes, the author sidestepped that by saying the M2 will be on the N4P process and telling us how much of an improvement that is expected to be over the (N5) process of the A14 & M1, but he does it without mentioning that the A15 is on the N5P process, not the N4P, so his comparison between A14 and A15, while useful, only goes so far.

    The other thing that I’d quibble with is the assumptions he makes about Apple’s ability to compete with high-end desktop GPUs — I mean, that remains to be seen. I don’t think it’s coincidental that Apple announced a ray-tracing Metal API at WWDC 2020, with development active and ongoing. You might want to wait until Apple releases an actual Pro desktop M1 graphics configuration before you start making assumptions about how far behind the competition they are… 
    Apple competes evenly in single core speeds. They just haven’t, as yet, produced chips with the large numbers of cores AMD and Intel have. Supposedly they will for high end computers shortly. Apple is relying on shared memory for a number of their significant performance gains, as well as the neural engine, machine learning cores, video and graphics encoders and decoders, etc. The chips pull more than their weight because of that. But when we do see 20 core chips then they’ll compete easily against the other high core count chips from AMD and Intel.
    patchythepiratewatto_cobraDetnator
  • Reply 9 of 20
    lkrupplkrupp Posts: 10,341member
    mtanika said:
    Most are guessing M2 will be an enhanced M1 using ARMv9.
    Rene Ritchie’s youtube channel points out that Samsung’s new Galaxy S22 Ultra uses the latest Snapdragon SOC with ARMv9 and a 4nm process, yet the A15 bionic ARMv8 5nm mops the floor with the Qualcomm S8.



    Watch the video and then come back  with your response.
    netroxwatto_cobra
  • Reply 10 of 20
    melgross said:
    bsimpsen said:
    I don't know the particulars of Apple's ARM architecture license, but if royalty payments for older versions decline over time, or if payments for new versions increase over time, there's financial incentive for Apple to drift away from ARM into its own custom designs. Whether the ARM design teams can keep Apple interested is an open question.
    Apple doesn’t use ARM designs, so they don’t pay for them. Apple’s designs are entirely home grown. They have an architectural license which means they use the instruction set, basically.
    Apple pays for the Architectural license. As instructions are added to the instruction set, I expect ARM will want compensation for them. To the extent those instructions support ancillary functions like AI, I expect Apple might not adopt them.
    watto_cobra
  • Reply 11 of 20
    crowleycrowley Posts: 10,453member
    lkrupp said:
    mtanika said:
    Most are guessing M2 will be an enhanced M1 using ARMv9.
    Rene Ritchie’s youtube channel points out that Samsung’s new Galaxy S22 Ultra uses the latest Snapdragon SOC with ARMv9 and a 4nm process, yet the A15 bionic ARMv8 5nm mops the floor with the Qualcomm S8.



    Watch the video and then come back  with your response.
    Doesn’t answer the question at all, so I don’t know why they’d respond.

    Ugh, that video was painful. Rene Ritchie is so annoying.
    muthuk_vanalingam
  • Reply 12 of 20
    netroxnetrox Posts: 1,286member
    lkrupp said:
    mtanika said:
    Most are guessing M2 will be an enhanced M1 using ARMv9.
    Rene Ritchie’s youtube channel points out that Samsung’s new Galaxy S22 Ultra uses the latest Snapdragon SOC with ARMv9 and a 4nm process, yet the A15 bionic ARMv8 5nm mops the floor with the Qualcomm S8.



    Watch the video and then come back  with your response.
    That video is good but I am tired of people - like him - asking for "pen" on iPhones... it's NOT ideal. And if you insist on having a pen, you CAN get a stylus to work with iPhone on Amazon.com.  
    patchythepiratewatto_cobra
  • Reply 13 of 20
    mjtomlinmjtomlin Posts: 2,586member
    bsimpsen said:
    melgross said:
    bsimpsen said:
    I don't know the particulars of Apple's ARM architecture license, but if royalty payments for older versions decline over time, or if payments for new versions increase over time, there's financial incentive for Apple to drift away from ARM into its own custom designs. Whether the ARM design teams can keep Apple interested is an open question.
    Apple doesn’t use ARM designs, so they don’t pay for them. Apple’s designs are entirely home grown. They have an architectural license which means they use the instruction set, basically.
    Apple pays for the Architectural license. As instructions are added to the instruction set, I expect ARM will want compensation for them. To the extent those instructions support ancillary functions like AI, I expect Apple might not adopt them.

    Huh? Apple licensed ARMv8 ISA and extended it as they needed. They don’t own ARM anything for that. Companies with an architectural license are completely free to extended the ISA as long as it remains backwards compatible - with the originally licensed version. Most of what ARMv9 offers, Apple has already implemented in their own designs, so there’s really no need to “upgrade”.
    watto_cobraweirdsmith
  • Reply 14 of 20
    mjtomlinmjtomlin Posts: 2,586member
    If M2 is based off A15 generation cores then that means Macs will end up being a couple of generations behind the iOS devices, and I don’t see that happening. The M1 was released with the A14 and they shared the same cores.. The M2 will more than likely be released in the Fall (regardless of what rumors say) alongside the A16 and they will also share the same cores.

    The base M-series SoCs are basically the same as the “X” variant of the A-series and these can be produced relatively cheap enough for a yearly update, especially since they’re used by the iPad Pros plus all consumer level Macs. The “pro” M-series SoCs are a different beast and are probably very expensive to produce, so Apple will want to stretch out their run a couple of years to bring down costs.
    watto_cobraVermelho
  • Reply 15 of 20
    Performance is important but features are important as well. The GPU is missing hardware ray tracing (that's even available on the latest Samsung phones). NVIDIA and AMD have had it for the past three years. Hoping Apple knocks our socks off!
  • Reply 16 of 20
    mcdavemcdave Posts: 1,919member
    This makes no sense. Why would M-series slide from parity to a year (or 6-months)  behind A-series architecture? If anything, M should lead A.
    October release on A16 equivalent.
  • Reply 17 of 20
    melgrossmelgross Posts: 33,334member
    bsimpsen said:
    melgross said:
    bsimpsen said:
    I don't know the particulars of Apple's ARM architecture license, but if royalty payments for older versions decline over time, or if payments for new versions increase over time, there's financial incentive for Apple to drift away from ARM into its own custom designs. Whether the ARM design teams can keep Apple interested is an open question.
    Apple doesn’t use ARM designs, so they don’t pay for them. Apple’s designs are entirely home grown. They have an architectural license which means they use the instruction set, basically.
    Apple pays for the Architectural license. As instructions are added to the instruction set, I expect ARM will want compensation for them. To the extent those instructions support ancillary functions like AI, I expect Apple might not adopt them.
    Of course they do. But that’s very different than paying for designs, as you said they were doing, they are not paying for designs. It’s also believed, but nobody knows from certain, that as a founder, Apple has their license in perpetuity.
    tenthousandthingswatto_cobra
  • Reply 18 of 20
    blastdoor said:
    mtanika said:
    Most are guessing M2 will be an enhanced M1 using ARMv9.
    "most"? I hadn't heard that, other than a few random speculators in comments around here. 

    I'm skeptical it will be ARMv9, and I think it's very possible Apple will never use ARMv9, simply because they might have no need to. 

    I think the MacWorld article is pretty reasonable and if it's wrong, the most likely way to be wrong is for the M2 to be based on the A16 rather than the A15. 
    Agreed…
    But it’s a nice opinion. Not a fact.
    Why don’t we wait for this to become a fact,
    before we blurry the line between entertainment and information.
    edited February 2022 williamlondonwatto_cobra
  • Reply 19 of 20
    geekmee said:
    blastdoor said:
    mtanika said:
    Most are guessing M2 will be an enhanced M1 using ARMv9.
    "most"? I hadn't heard that, other than a few random speculators in comments around here. 

    I'm skeptical it will be ARMv9, and I think it's very possible Apple will never use ARMv9, simply because they might have no need to. 

    I think the MacWorld article is pretty reasonable and if it's wrong, the most likely way to be wrong is for the M2 to be based on the A16 rather than the A15. 
    Agreed…
    But it’s a nice opinion. Not a fact.
    Why don’t we wait for this to become a fact,
    before we blurry the line between entertainment and information.
    To be fair, it’s not really an “opinion” — it’s an informative assessment of the only actual fact that exists, the A15 cores (versus the A14/M1 cores). What if Apple were to use the A15 cores in the M2? It’s a valid, fact-based question. 

    It doesn’t predict Apple will use the A15 cores (that would be an opinion), it just provides information about what it would mean if they did. The author states repeatedly that this is an “assumption” as he lays out this premise. 
  • Reply 20 of 20
    melgrossmelgross Posts: 33,334member
    netrox said:
    lkrupp said:
    mtanika said:
    Most are guessing M2 will be an enhanced M1 using ARMv9.
    Rene Ritchie’s youtube channel points out that Samsung’s new Galaxy S22 Ultra uses the latest Snapdragon SOC with ARMv9 and a 4nm process, yet the A15 bionic ARMv8 5nm mops the floor with the Qualcomm S8.



    Watch the video and then come back  with your response.
    That video is good but I am tired of people - like him - asking for "pen" on iPhones... it's NOT ideal. And if you insist on having a pen, you CAN get a stylus to work with iPhone on Amazon.com.  
    I don’t see why anyone would get annoyed at more features rather than less. Just like the Pencil for iPads, you don’t have to buy it, but it’s great to have when you need the precision and features. So, yes, I’d like to have a pen available for my iPhone. Whether I’d use it depends on how it felt, and how useful it would be. I know some who have, or have had the Note. Some liked the pen, and some didn’t. But they had the option. Options are always good. And I like seeing someone influential pushing for it.
    muthuk_vanalingam
Sign In or Register to comment.