M2 Pro, M2 Max MacBook Pro models could arrive by the fall

2»

Comments

  • Reply 21 of 29
    thttht Posts: 4,618member
    Fred257 said:
    I’m not getting near any M2 anything. The reviews I have seen show absolutely no consistency in both the CPU and GPU reviews. As a matter of fact I think it’s the worst Apple architecture we will see. Apple, fix your stupid mistakes!! Only one NAND chip on the lower end models is a complete design flaw and ultra cheap on your part. The GPU being slower on the 10 core then the 8 is inexcusable!! Stop and improve these mistakes please. And for Apple fanboys here who complain about me I’ve used Macs since 1989 making music on Cubase before many of you were in your mothers womb 
    I think this is a confluence of a couple of things.  Their low end machines are using a single NAND chip as a cost saving measure, and the higher end machines shouldn't do that.  The GPU being slower on 10 vs 8 cores is likely a software thing (in the driver or Metal) or a benchmarking artifact.

    As usual, you need to assess the machine you're interested in (when that becomes possible), not projections or theorizing about what it'll be.  I think the M2 pro/max/ultra/whatever are going to be pretty impressive chips and Apple will put them in either the same or improved systems compared to last year's models.  Hopefully the long awaited Mac Pro arrives too.

    It's also a confluence of your usual social media rage machine. For the vast majority of buyers, the base SKU is faster than then the prior generation. And by vast, I mean 99% of base model buyers. The storage may be slower than a prior Intel or M1 model, but net-net it will be faster for the vast majority of workflows. There really isn't a controversy here. For the few users who have workflows gated by storage speeds, they are now aware of it and can option up if they want to, and it's all relative to the machine they are upgrading from.

    I don't think there is a scenario where I would recommend getting the M1 MBA over the M2 MBA. If someone can afford $1000, they can afford $1200. I suppose bulk buyers will feel that +$200 difference if they are buying >1000 machines, but for an individual buyer, they really should get an M2 machine. They may choose the M1 MBA for reasons not related to performance, but the M2 MBA is a better machine in terms of display, performance, ports, and probably reliability too.

    The rumor is that the M2 Pro/Max will be TSMC 3nm products. That will buy about +15% clock speed increase or about a -15% decrease in power consumption. So, net-net, it's going to be pretty darn good. We will know in about 2 months when the iPhones come out. If the A16 is fabbed on 3nm, odds are pretty good that the M2 Pro and Max SoCs will be 3nm.


    designrprogrammer
  • Reply 22 of 29
    techconctechconc Posts: 244member
    tht said:

    The rumor is that the M2 Pro/Max will be TSMC 3nm products. That will buy about +15% clock speed increase or about a -15% decrease in power consumption. So, net-net, it's going to be pretty darn good. We will know in about 2 months when the iPhones come out. If the A16 is fabbed on 3nm, odds are pretty good that the M2 Pro and Max SoCs will be 3nm.
    I'd be surprised if the M2 Pro/Max are tabbed on the 3nm process, even if the A16 is.  This is a new node and the yields would likely be low initially.  In fact, the rumor suggests only the Pro models will get the A16 chip.  I suspect part of that decision (if true) is due to yield issues based on when the 3nm process goes live.  I'd also expect all M2 related products to be on the same node.  We'll see how this plays out.
    designr
  • Reply 23 of 29
    designrdesignr Posts: 754member
    techconc said:
    tht said:

    The rumor is that the M2 Pro/Max will be TSMC 3nm products. That will buy about +15% clock speed increase or about a -15% decrease in power consumption. So, net-net, it's going to be pretty darn good. We will know in about 2 months when the iPhones come out. If the A16 is fabbed on 3nm, odds are pretty good that the M2 Pro and Max SoCs will be 3nm.
    I'd be surprised if the M2 Pro/Max are tabbed on the 3nm process, even if the A16 is.  This is a new node and the yields would likely be low initially.  In fact, the rumor suggests only the Pro models will get the A16 chip.  I suspect part of that decision (if true) is due to yield issues based on when the 3nm process goes live.  I'd also expect all M2 related products to be on the same node.  We'll see how this plays out.
    I agree. It seems most likely for the M3 to be the first (Mac Silicon) generation to use 3nm.
    techconc
  • Reply 24 of 29
    Reading this thread and the others like it, as we try to imagine what Apple has planned, I think it’s probable Apple will eventually find that it makes sense to make an effort to educate its customers on the development cycles for Apple Silicon. Something like what Intel used to do with the old (now positively ancient) “tick tock” model: “tick” was the manufacturing process node, “tock” was the microarchitecture. It’s more complicated than that now, but there’s a lot to be said for cultivating something like that sense of development, moving forward like clockwork. It’s not a coincidence that Intel was most dominant during the “tick tock” period.

    Just don’t repeat Intel’s mistakes when the model inevitably breaks down, making promises they couldn’t keep. 
    edited July 22
  • Reply 25 of 29
    thttht Posts: 4,618member
    Reading this thread and the others like it, as we try to imagine what Apple has planned, I think it’s probable Apple will eventually find that it makes sense to make an effort to educate its customers on the development cycles for Apple Silicon. Something like what Intel used to do with the old (now positively ancient) “tick tock” model: “tick” was the manufacturing process node, “tock” was the microarchitecture. It’s more complicated than that now, but there’s a lot to be said for cultivating something like that sense of development, moving forward like clockwork. It’s not a coincidence that Intel was most dominant during the “tick tock” period.

    Just don’t repeat Intel’s mistakes when the model inevitably breaks down, making promises they couldn’t keep. 
    Fab process development is much harder and resource intensive now than it was in the past. It's a fool's game to de facto promise when the next node is going to come. TSMC 3 nm is about a year late, and there aren't any warm fuzzies that TSMC will have shipped 3 nm silicon this Fall. Getting 2 nm into production will be even trickier. So, say nothing I'd say.

    Usually, we would get a rumor that TSMC has started mass production of 3 nm wafers by now if Apple is shipping a 3 nm A16 in September. That's hard to hide. We should have gotten an A16 Geekbench leak by now, but this is easier to keep secret. Apple should already have an allotment of 3 nm A16 silicon right now. It's just about 8 weeks to end of September now. Apple will need about 20m chips, if not 40m, by the end of year. All the parts have to be hitting the mass manufacturing right now.
    tenthousandthings
  • Reply 26 of 29
    MarvinMarvin Posts: 14,831moderator
    tht said:
    Reading this thread and the others like it, as we try to imagine what Apple has planned, I think it’s probable Apple will eventually find that it makes sense to make an effort to educate its customers on the development cycles for Apple Silicon. Something like what Intel used to do with the old (now positively ancient) “tick tock” model: “tick” was the manufacturing process node, “tock” was the microarchitecture. It’s more complicated than that now, but there’s a lot to be said for cultivating something like that sense of development, moving forward like clockwork. It’s not a coincidence that Intel was most dominant during the “tick tock” period.

    Just don’t repeat Intel’s mistakes when the model inevitably breaks down, making promises they couldn’t keep. 
    Fab process development is much harder and resource intensive now than it was in the past. It's a fool's game to de facto promise when the next node is going to come. TSMC 3 nm is about a year late, and there aren't any warm fuzzies that TSMC will have shipped 3 nm silicon this Fall. Getting 2 nm into production will be even trickier. So, say nothing I'd say.

    Usually, we would get a rumor that TSMC has started mass production of 3 nm wafers by now if Apple is shipping a 3 nm A16 in September. That's hard to hide. We should have gotten an A16 Geekbench leak by now, but this is easier to keep secret. Apple should already have an allotment of 3 nm A16 silicon right now. It's just about 8 weeks to end of September now. Apple will need about 20m chips, if not 40m, by the end of year. All the parts have to be hitting the mass manufacturing right now.
    TSMC said 3nm volume manufacturing would start H2 2022, it's H2 just now so it's possible they've started or will be starting soon but it would be too tight a schedule for iPhones as they have to assemble the product:

    https://www.anandtech.com/show/17452/tsmc-readies-five-3nm-process-technologies-with-finflex
    https://www.tsmc.com/english/dedicatedFoundry/technology/logic/l_5nm

    For 5nm, they started volume manufacturing Q2 2020 and Apple had it in the iPhone/iPad in Q3 in September/October.

    The following says A16 will use N5P and that N4 won't be much of an improvement over N5P:

    https://www.tweaktown.com/news/86529/apples-next-gen-a16-soc-in-iphone-14-baked-on-tsmc-n5p-process-node/index.html

    Given the low shipment volume for Pro Macs, likely under 8 million units per year (1-2 million for launch), it's still possible they can manufacture M2 Pro/Max/Ultra on 3nm because these will be announced in October Q4 and can ship November/December. For iPhone/iPad, the shipment volume is high at 40 million for launch on 3nm in just 8 weeks.

    Apple managed to get a 40% performance boost from N5P (M2) over N5 (M1). That would still be a worthwhile improvement for M2 Pro/Max/Ultra but hopefully they will use 3nm as this will give the best result for a Mac Pro launch. I don't think they will use a dual UltraFusion for M2. I reckon they'd just connect multiple Ultras similar to how AMD/Intel connect multiple tiles:

    https://www.techpowerup.com/292204/intel-sapphire-rapids-xeon-4-tile-mcm-annotated

    September: A16 (N5P)
    October: M2 Pro/Max/Ultra (N3 or N5P)
    H1 2023: Apple AR and M3 (N3)
    H2 2023: A17 (N3), Macs either still N3 or some improved variant
    tenthousandthings
  • Reply 27 of 29
    tht said:
    Reading this thread and the others like it, as we try to imagine what Apple has planned, I think it’s probable Apple will eventually find that it makes sense to make an effort to educate its customers on the development cycles for Apple Silicon. Something like what Intel used to do with the old (now positively ancient) “tick tock” model: “tick” was the manufacturing process node, “tock” was the microarchitecture. It’s more complicated than that now, but there’s a lot to be said for cultivating something like that sense of development, moving forward like clockwork. It’s not a coincidence that Intel was most dominant during the “tick tock” period.

    Just don’t repeat Intel’s mistakes when the model inevitably breaks down, making promises they couldn’t keep. 
    Fab process development is much harder and resource intensive now than it was in the past. It's a fool's game to de facto promise when the next node is going to come. TSMC 3 nm is about a year late, and there aren't any warm fuzzies that TSMC will have shipped 3 nm silicon this Fall. Getting 2 nm into production will be even trickier. So, say nothing I'd say.

    Usually, we would get a rumor that TSMC has started mass production of 3 nm wafers by now if Apple is shipping a 3 nm A16 in September. That's hard to hide. We should have gotten an A16 Geekbench leak by now, but this is easier to keep secret. Apple should already have an allotment of 3 nm A16 silicon right now. It's just about 8 weeks to end of September now. Apple will need about 20m chips, if not 40m, by the end of year. All the parts have to be hitting the mass manufacturing right now.
    Thanks, I didn’t mean to imply Apple should be publishing roadmaps or making promises in that regard. I meant that I think maybe there is an opportunity to seize the narrative and educate the public about things that are widely understood inside the industry, but not outside of it. It wouldn’t be a big step beyond the technical information they already provide, just providing a better framework for understanding it. 
  • Reply 28 of 29
    designrdesignr Posts: 754member
    Reading this thread and the others like it, as we try to imagine what Apple has planned, I think it’s probable Apple will eventually find that it makes sense to make an effort to educate its customers on the development cycles for Apple Silicon. Something like what Intel used to do with the old (now positively ancient) “tick tock” model: “tick” was the manufacturing process node, “tock” was the microarchitecture. It’s more complicated than that now, but there’s a lot to be said for cultivating something like that sense of development, moving forward like clockwork. It’s not a coincidence that Intel was most dominant during the “tick tock” period.

    Just don’t repeat Intel’s mistakes when the model inevitably breaks down, making promises they couldn’t keep. 
    The only issue I see here is that Apple is not selling Apple Silicon. It's the "secret sauce" of their real products. This is quite different from Intel. Intel's product is the chips so they had to communicate their process to potential customers (companies like Apple, Dell, et al).

    Apple seems much more inclined to let the devices they sell speak for themselves and not let customers worry too much about the details.

    Yes, they do delve into some details about the chips in their presentations. I see these serving two purposes: 1) targeting developers, 2) showing off their "bonafides" for those who are kinda interested and seeing that Apple is making progress in some more tangible, less hand-wavy way.
  • Reply 29 of 29
    Marvin said:
    September: A16 (N5P)
    October: M2 Pro/Max/Ultra (N3 or N5P)
    H1 2023: Apple AR and M3 (N3)
    H2 2023: A17 (N3), Macs either still N3 or some improved variant
    Hmm. Something will have to give. I think pressure to launch the M2 Pro/Max MacBook Pros will be intense, and that will take priority. 

    [1] Right now, Apple still sells two Intel Macs, the Mac Pro and the Space Gray Mac Mini.
    [2] iMac 24" launched April 2021. (new)
    [3] MacBook Pro launched October 2021. (new)
    [4] Mac Studio launched March 2022. (new)
    [5] MacBook Air launched June 2022. (new)

    These four above are all of the Macs designed (so far) for Apple Silicon from the ground up. The M1/M2 Touch Bar MacBook Pro and the M1 Mini are transitional designs. I think it's significant that the M1 Mini did not get an M2 refresh. That means a redesign is on the horizon, or it is dead. I've left it out here. I'm writing this on a higher-end M1 Mini with a Studio Display, so it affects me, but I don't know what to think. Most likely I will move to a Mac Studio when Apple updates the Pro Display XDR. A Mac Pro is total overkill for me.

    So here's what we get,

    October 2022:
    [6] M2 iMac 24" (refresh)
    [7] M2 Pro/Max MacBook Pro (refresh)

    March 2023:
    [8] M2 Max/Ultra Mac Studio (refresh)
    [9] M2 Ultra/Duo Mac Pro (new)

    Finally, June 2023 the TouchBar MacBook Pro is dropped for a new, next-generation M3 design in the same price position. Then and only then are the Intel Mac Pro and Intel Mini discontinued.
    edited July 24
Sign In or Register to comment.