I suspect something similar. I can imagine that they decided to change the schedule so that they can have a better answer to Alder Lake and friends. Like offering a more scalable architecture that can reach higher performance in a desktop environment.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
M2 Platform Wish List and Predictions
- Thread starter playtech1
- Start date
- Sort by reaction score
I suspect something similar. I can imagine that they decided to change the schedule so that they can have a better answer to Alder Lake and friends. Like offering a more scalable architecture that can reach higher performance in a desktop environment.
The M1 is based on the A14 cores. It made sense for the M2 to be based on the A15 cores. There’s probably a development gap between the M2 and its iterations with the A16. Apple would not only do a regular M2, but also M2 Pro, Max, and Ultra. The A16 would probably be too early in the stages for Apple to develop the M2 ultra.
I could be wrong. We’ll find out.
The other interesting factor is Mac Pro at the June WWDC. During the last event they explicitly stated that the M1 family was finished with the M1 Ultra, meaning anything that goes into the Mac Pro will be M2 based (if it is 2 M1 Ultra stitched together it would qualify as one more M1 chip wouldn't it)
So if the M2 architecture is already well into development then perhaps it makes sense for it to lag behind the iPhone chip. We may be looking at 2 iteration cycles in terms of fab advance, i.e. the M3 will go down to 4 or even 3nm, while the M2 stays at 5nm.
There's a recurring pattern in past Apple chips which suggests they might use a cadence of 2 years for refreshing M series chips.
A10 + A10X
A11
A12 + A12X/Z (note: the Developer Transition Kit mac minis used an A12Z)
A13
A14 + M1 (aka A14X/Z)
A15
Even generations get both An and AnX, odd generations get just An. I think it makes a lot of sense for Apple to continue this pattern and base M2 off A16 generation tech.
The reason they can afford to design a new iPhone chip every year without fail is that they sell something like 200 million iPhones per year. Macs sell about a tenth as many units. That's still a lot, but the economics for doing a new chip refresh for every Mac every year probably aren't great.
It just seems odd to me that only the Studio Display gets this feature on the Mac (perhaps it gets it because it's basically an iPad internally), plus the MBP notch is massive relative to the actual camera, so there seems to be space for a more sophisticated camera setup (albeit not much depth).
It made sense before.
But keep in mind that the base M series chip now goes into the Macbook Air, Mac Mini, iMac, iPad Pro 10, iPad Pro 13, iPad Air.
In the past, it just went into the iPad. There are now 5 reasons to refresh at least the base M SoC once a year.
I would like to see Apple refresh the M series each year, and the M Ultra & 2x M Ultra every 2 years.
I doubt that Apple can change its schedule so fast. They plan these 3 or 4 years in advance, right?
If rumors are true that Apple almost announced the M2 MBP 13" in the spring event, then it's got to be the A15.
Though one curious thing is that all signs point to 12 CPU cores for the M2 Pro/Max which could suggest that it's based on an A16.
Also, the A16 will use TSMC 4nm. The A15 still uses TSMC 5nm. It might make sense to design both the A16 and the M2 on TSMC 4nm.
I'm hoping for A16 of course. However, I won't upgrade until a 3nm chip comes out which is the next giant leap in node technology from TSMC.
Yeah, but they could skip a generation. What if an A15 M2 was originally planned for fall 2021, but Apple skips it to release an A16 based M2 in second half of 2022 instead? Pure speculation of course, not that I have strong opinion either way.
Regarding the rest what you say, maybe M2 is based on A15 and M2 Pro/Max is based on A16. That would be a way to include some pro-level features for the prosumer hardware (like RT etc.).
That would mess up the numbering system.
I hope it's A16. I think Apple can reclaim the ST performance crown from Intel and maybe even AMD's upcoming Zen4 if it's A16.
That would mess up the numbering system.
Not at all. Would still be M2. Apples does not make any public statements about how their tech relates to each other internally.
Hopefully with something better. IMG ray tracing hardware only exists on paper. And Apple needs programmability.
CPU speaking, they nailed it. They can continue what they've been accomplishing for the past 10 years or so, and it will be fantastic.
GPU speaking, there's room for improvement in both software and hardware.
Short shot : They need to do what they did with the iPhone 13 Pro, which is to add more cores to boost performance.
Long shot : They need to work some magic with game developers to make native games. Apple's never been serious in gaming, and honestly 1.5 years after the launch of Apple Silicon, nothing has changed in that department. The proof is the M1 Ultra GPU has been tough to benchmark correctly and objectively by the main Apple YouTubers.
GPU speaking, there's room for improvement in both software and hardware.
Short shot : They need to do what they did with the iPhone 13 Pro, which is to add more cores to boost performance.
Long shot : They need to work some magic with game developers to make native games. Apple's never been serious in gaming, and honestly 1.5 years after the launch of Apple Silicon, nothing has changed in that department. The proof is the M1 Ultra GPU has been tough to benchmark correctly and objectively by the main Apple YouTubers.
I wouldn't count those as proofs. The design of the M1 GPU is vastly different from those of traditional dGPUs. I would say that almost all GPU benchmarking software are not design with the M1 GPU architecture in mind. Using Metal APIs alone probably does not mean maximising the potential of the M1 GPU cores.
Personally, I'm leaning towards a Mac Studio but I'm going to wait a month or two for any bugs to emerge. It's not like the Studio is going to be replaced anytime soon. The early worm gets the bird!
Might hold out until WWDC although there's no guarantee that anything relevant will be launched/announced then.
At this stage, I'd hold off an anything M1 or M1 Pro based until the (base) M2 emerges - since the regular M2 is quite likely to beat the M1 Pro (certainly the lower-end "binned" version) on some workflows. I'm assuming that any M2 Pro and M2 Ultra will lag 6 months or so behind the regular M2 (if they're going to happen - maybe they'll only do max/ultra versions every other generation).
I don't think it'll be much difference, basically the difference between the A14 and A15. (10-15% faster) and that's about it. 2nd gen products don't usually have much changes. There maybe could be a new interconnect so they can go 4 max's. I can see that for a new high end machine.
What would I like to see, nothing, I'm happy with my M1 Studio Max and I wont be upgrading for some time.
What would I like to see, nothing, I'm happy with my M1 Studio Max and I wont be upgrading for some time.
Considering M1 is basically an A14X and the chip previous to that was the A12X (disregarding the A12Z) and previous to that was the A10X, the trend is generally Apple go with every other A series chip to base their X series chipset off of.
So with that logic I expect the M2 will be based off the A16…
Remember the A14 launched in September 2020, whereas the M1 (or A14X) came along in November 2020, so I expect A16 in September, then M2 (or A16X) coming along in October or November… I don’t expect the M2 Pro/Max to come along until late 2023, with the M2 Ultra coming in Early 2024.
Both 3DMark and geekbench should use Metal fairly well. And I sm am not sure that alone the fact of being different is enough to “excuse” a GPU - you would still expect it do well at common things.
The problem with existing benchmarks (and you do mention it, I just wanted to stress it more) is that done of them are very flawed while others simply treat M1 as an afterthought due to it being a fairly novel/rate platform. For example, Geekbench compute tests are too short to be meaningful for any practical purpose, GFXBench is too light, resulting in obscenely high FPS where the differences in benchmark results on different hardware might be measuring irrelevant implementation details etc. I kind of trust 3DMark because it produces the results that make sense to me.
That's only a problem if Apple think they can whack $200 on to the price for the luxury. 16GB of RAM and 512GB SSD is not an unusual base spec for a premium laptop starting at $1000. It shouldn't be an expensive option in 2022. There's nothing magical about the LPDDR and flash chips that Apple are using, and they're charging a massive markup for the BTO options.
It's possible. Thing is, its hard to see what they're going to do with the Mac Pro - given they've declared (accurately or otherwise) that the Studio Ultra already thrashes the 28-core Mac Pro, the distinguishing "Mac Pro" features are now its extreme 1.5TB RAM capacity and ability to support 64 lanes worth of PCIe cards. Not clear whether that is possible at all by gluing a bunch of laptop chips together, even if they're M2s. Also, a lot of Apple Silicon's advantages come from "unified" on-package RAM shared by on-die GPUs and hardware accelerators... it's really not a good fit for anything similar to the 2109 MP so they're gonna have to "think different" - and possibly require workflow changes from Mac Pro customers. Making a whole new Apple Silicon die, with extra PCIe and external RAM support just for the Mac Pro market could be horrendously expensive.
One possibility might be to use multiple M1 Ultra SoCs (until the 2019 Xeon W came along, using multiple 'Scalable Xeon' chips was the only way of getting Mac Pro-level specs) so a possible model for a AS Mac Pro could be a backplane for multiple MPC-like M1 Ultra "compute modules" using something like NUMA to get them working together and sharing RAM.
I think the "minimum viable product" for Mac Pro could even be a 1U rackmount version of the Studio Ultra - that could be racked up with additional processors and PCIe backplanes using Thunderbolt.
The M2 is going to be based on A16, 4nm process. This is already reported.
Apple makes all of its Mac money on the BTO upgrades. Can anyone find another $1000 laptop on the market that comes with only 8 GB of RAM?
It really isn't an argument about whether or not 8 GB is enough for base model users (it is). It is a question of whether the price is appropriate for it (its not). It is all about charging you that extra $200 for 16 GB.
It really isn't an argument about whether or not 8 GB is enough for base model users (it is). It is a question of whether the price is appropriate for it (its not). It is all about charging you that extra $200 for 16 GB.
Good deduction, I think the same tbh. However, that means no M3 until 2024… yeah, makes sense