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.
New cores (A15)
- Thread starter cmaier
- Start date
- Sort by reaction score
I think it's possible they skip the M2 Pro/Max as they have skipped a few X chips back in the day, but the rumors have pointed to a redesigned MBA with a 10-core GPU so I still expect there to be a standard M2 at least.
As @reallynotnick says, M2 Airs are rumored/expected next year but overall your question is quite valid: the cadence of Mac updates relative to core updates is unknown. They could skip generations, they may not. Some macs/m-series chips may get yearly updates and stay current with phone cores, some may not. At this point, outside of Apple and probably only a small number even there, who knows?
If I had to guess: the rumored huge desktop chips yet to be unveiled may skip core generations and be updated every other year or 18 months while smaller M1-style chips stay current more or less year-to-year.
I’ve heard skipping generations or 18 month release cycles posited before, but I’m always left wondering what a “generation” is in reference to.
For instance, is it M2 Max to M1 Max, or is it M2 Max to M2 (or M1)? Same goes for the Mac Pro SoCs.
I’d think the cadence from one generation to the next of a single variant (base A-series, Pro, Max, etc.) would be 1 year. But, there would be a time offset between the release of the base cores and release of the other variants using the same cores.
Can you elaborate?
Last edited:
I am curious whether Apple will refresh these on an annual basis aligned to the latest A-series architecture, and whether the fact the cores appear to be Firestorm was an aberration caused by the fact it was a huge new design plus possible supply-chain issues.
Will we see a refreshed M2 Pro/Max in October 2022 with the latest improvements that go into the A16.
I tried to make it clear but obviously failed
. I’ll try to post examples:The CPU cores have historically been updated every year with a new phone SOC often on a new process node: this year we get avalanche in the A15 on 5nm+, last year firestorm in the A14 on 5nm, next year it’ll be something else in the A16 on probably a 4nm, maybe 3nm process.
The timing of the rest of the lineup is in flux because of the transition and new bodies to go with new chips. But you could imagine that the small M1-chip gets a new SOC every year. But larger SOCs don’t. If they have an 18 month or longer cycle they would fall very far behind Apple’s bleeding edge cores and TSMC fabrication nodes unless they skip core generations. As and example: maybe next year, no M2 max SOC ever gets made, it skips to the M3 max (or even if they call it the M2 Max it’s still with the cores and fabrication process of the A16 not the A15 as is expected with the M2).
Ah, I see! I have been wondering the same thing, so thank you for clarifying.I tried to make it clear but obviously failed
The CPU cores have historically been updated every year with a new phone SOC often on a new process node: this year we get avalanche in the A15 on 5nm+, last year firestorm in the A14 on 5nm, next year it’ll be something else in the A16 on probably a 4nm, maybe 3nm process.
The timing of the rest of the lineup is in flux because of the transition and new bodies to go with new chips. But you could imagine that the small M1-chip gets a new SOC every year. But larger SOCs don’t. If they have an 18 month or longer cycle they would fall very far behind Apple’s bleeding edge cores and TSMC fabrication nodes unless they skip core generations. As and example: maybe next year, no M2 max SOC ever gets made, it skips to the M3 max (or even if they call it the M2 Max it’s still with the cores and fabrication process of the A16 not the A15 as is expected with the M2).
Considering the scalability of the M1 to the M1 Max (and to whatever the “workstation” SoC configuration might be), I’m thinking that there is enough (performance, use case, etc.) distance between the variants that lagging a generation wouldn’t be all that bad.
I shared @leman’s disappointment of lack of A15 cores in the new SoCs, however that allows Apple to fine tune bottlenecks due to scaling out.
Many people voiced concern that memory bandwidth would be problematic on the new SoCs, and Apple completely blew that away. I don’t mean to put forth a false choice, but the alternative might have been A15 cores and LPDDR4X. I’m excited to see how the 400GB/s works out.
Also, my takeaway from the SLC and bandwidth is that Apple is scaling to the GPU. The CPU is not the focus, which is particularly exciting— 1/3 of the die is the GPU! And there is still the mystery of the extra 16 core neural engine on die.
A number of smart people seem pretty convinced that the presented die shots were faked/altered. It could be an extraneous NPU, but it may be something else that’s “off” but has been designed into the SOC for future iterations that will use the same blueprint. For instance: if the desktop SOC is two M1 Maxes glued together, it could be logic that helps with that … or like when AMD announced v-cache, people went back to the Zen3 die shots and realized that they had predesigned the space for the extra 3D cache. It could be something more prosaic but 🤷♂️
Edit: People will get actual die shots during tear downs so we’ll know a lot more then.
Last edited:
Concurred on all points. We’ll just have to wait for delivery… and teardown. I imagine TechInsights will be all over it. They were pretty quick with the A15 die analysis.
I think Apple's original plan was:
New A SoC in October
New Mx SoC in November
New Mx Pro/Max SoC in June the following year
But this year, they delayed the MBP to fall.
In June 2022, we might see the laptops get refreshed with the M2 Pro/Max. Then in November, we might see the M3 in a redesigned MBA.
This is why I'm only getting a base 16" right now in case they refresh again in June 2022.
Not sure what you expected. The M1 has a smaller memory interface with less bandwidth and far more port and display support. All this stuff take a bit more power. A bigger chip was never going to be as efficient as the M1 with its trade offs and limited I/O.
The M1 was already a monster and beat the 16” i9. Any more cores is just bonus at this point. The new chip is twice as fast as the i9.
The only thing I wish it has was 2 more efficiency cores. It just kind of annoys me the M1 has more. Those cores can still come in handy on any computer especially a notebook.
Not shocking. If he had legit info then he’d be making money reporting it or someone like Gurman would have already had the scoop before him.
I’ve been warning people here for months to pump the breaks and use common sense and only listen to legitimate reporting. People just want to believe whatever they want to believe. Looks like I was correct when I threw a wet blanket on the “30-32 iMac Pro” nonsense speculation. Ross Young just came out with a new scoop of the next iMac being 27”. Personally I guessed it would either be 5.5k at ~28.6” or they would just keep 5k at 27”. They were never going to make a 32” 6k iMac for a third the price of the XDR. Common sense.
The only surprise to me is that it’s ProMotion MiniLED. Looks like a price increase is coming.
Do you have a link to a discussion or article? I'd be interested in checking it out-- my searches have only turned up a single reddit post.
Edit: I just found AnandTech’s article about the processor and there were a few comments somewhere around pages 25-29, but no in-depth discussion.
Last edited:
Andrei wrote about it in Twitter:
I saw one or two others say the same and he references others that he’s read.
That strikes me as pretty odd. The die shots aren’t high resolution enough to actually reverse engineer the logic. It’s really only useful to identify known structures and IP.
So, if they’re hiding it, maybe they’re licensing IP… from AMD? Infinity Fabric? That’d surprise me, but OTOH they presumably still have a good relationship with AMD… I like it. I’ll stick it on my prediction list.
No idea personally why 🤷♂️ … especially since teardowns and die shots will happen and quickly so this seems like it buys them little to nothing if it’s true.
Well, it looks like the M1 Pro/Max are of the same generation as the A14, rather than the A15 used in the latest iPhone. If Apple releases the next-gen MBP's in a year, it would be nice if they could employ the same generation of chips used in next year's iPhone (A16=>M3) rather than those in this year's iPhone (A15=>M2).
Who knows, maybe (just maybe) Apple will actually start giving us, GASP...!!!, short-term roadmaps towards future products...?!? ;^p
TechInsights in particular usually publishes die shots of flashy new Apple chips on their blog for free rather than requiring you to pay as you usually would; it's great PR for their paid services.
I can easily spitball a few ideas which seem more plausible. Perhaps Apple planned on doubling NPU core count for M1 Max, but chose not to, so the feature ended up being dark silicon. Or perhaps someone in the art department just unintentionally botched up the image while manipulating it to make it look good for presentations, without intending to hide anything.
Yeah, right, like that's gonna happen ?. The only exception I can recall was when they let pro customers know they were working on a trashcan replacement.
Are you guys saying that Jade 2C and 4C will be 2 or 4 discrete SoC with an interconnect?
Sorry not an expert, but how would you handle memory coherency between SoC if the memory is distributed on four SoC.
What would happen with the 3 nm process, would Apple use the same structure and use additional transistors to improve existing blocks rather than adding more cores?
Sorry not an expert, but how would you handle memory coherency between SoC if the memory is distributed on four SoC.
What would happen with the 3 nm process, would Apple use the same structure and use additional transistors to improve existing blocks rather than adding more cores?
That's the most likely current rumor.
The same way the memory coherency is handled within the SoC: the chips are interconnected to create one unified topology. From that perspective, the interconnected SoCs will simply work as one much larger SoC. Apple already kind of demonstrated this technology in M1 Pro and M1 Max: these are more or less scaled up m1 chips, with multiple blocks replicated.