And frankly, people really tend to over-interpret these simple (often taken out of context) statements.
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.
Could we see M3 before A17?
- Thread starter senttoschool
- Start date
- Sort by reaction score
And frankly, people really tend to over-interpret these simple (often taken out of context) statements.
Hey, when he said that, he looked to the left and his right hand index finger was curled a little bit. When you couple that with the position of his left foot, that must mean that an 8x Mac Pro is going to be released and connected wirelessly to the headset. It’s the only possible explanation!
🤔
Last edited:
After WWDC 2023, I highly doubt that M3 will even come out before A17.
- WWDC 2023 is over and yet no M3 based Mac ever announced. I dont think they wont gonna announce any more Macs in this year.
- Apple Vision Pro will use M2 and yet it wont be available till early 2024. Which means there wont be M3 even later this year.
- iPhone 15 Pro's chip just started to mass manufacture and they dont have time to make M3 chips yet.
- WWDC 2023 is over and yet no M3 based Mac ever announced. I dont think they wont gonna announce any more Macs in this year.
- Apple Vision Pro will use M2 and yet it wont be available till early 2024. Which means there wont be M3 even later this year.
- iPhone 15 Pro's chip just started to mass manufacture and they dont have time to make M3 chips yet.
I’m hoping they are going to give the Vision Pro an M3 and it doesn’t have one yet because it isn’t ready yet … if it has an M2 and it delays the M3 then the Vision Pro will be based on a 2.5 year old architecture by the time it launches…
Why is this a problem if it meets the needs of the device?
Gaming consoles keep stable hardware for years... it's a precedent that helps provide a stable target for software developers, and gives a consistent experience for users.
It's a first-gen device that will be quickly updated. It's about the package, not just processor. Even if you gave the original iPhone, people would still want 3G.
This is both acceptance of mediocrity and hopeful wishful thinking. What possible reason do you have to believe the headset will be updated relatively quickly? If it is updated quickly, why not just ship it with M3 which should be ready by then anyway (one would hope)? If Apple has the engineering talent to update the headset quickly to M3, why do they not also have the resources to keep the M series up to date?
We know that early 3nm yields were around 1/3 of 5nm back in the first quarter of 2023. Depending on which chips are being made on the process, M3 might not even be ready yet. Both the M1 and M2 were based on existing A series architectures rather than an entirely new core design. In the case of the M2, the cores are the same as found in the A15 Bionic. Assuming the M3 will be built on 3nm, it would likely be based off the A17 cores rather than the A16 cores. "Keeping the M series up to date" isn't as simplistic an undertaking as you might think, especially when you add the complexity of moving to a new process on top of everything else that goes into processor/SoC redesigns.
The M1 was based on the A14 which launched 2ish months before the M1. I think we'll see the M3 after A17, which means there should be no problem making sure that it is based on the A17 cores. Why should it launch significantly longer after A17 than M1 did after A14? Keeping it up to date should not be this hard for an organization with the resources of Apple... The A16 seems like a stop gap because N3 wasn't ready, but that was a year ago. They have had a whole year and with N3 and with the process now in better shape (a year of refinement) there is no reason Apple shouldn't be able to launch M3 this fall following A17. Again, excuse making for Apple to justify mediocrity and failure to execute. We didn't put up with this from Intel and we shouldn't from Apple either.
1) There is no indication the processor is the bottleneck.
2) Have you considered that the silicon team was probably focused on R1?
What first gen product wasn't updated quickly? Whether it's iPhone, iPad, or Watch they've all been updated quickly. Given it's a new product category, it's extremely unlikely they'll hit a home run on the first try.
"Mediocrity"? "Failure to execute"? Making the move to 3nm isn't something Apple does entirely on its own, TSMC plays as much (if not more) of a role in that aspect of things. You also completely overlook the multiple reports regarding production levels being lower for 3nm parts. Given how much of a backlog there was for the iPhone 14 Pro and Pro Max last holiday season with full-capacity 5nm production, what could possibly convince you that there are somehow more 3nm parts available in 2023? If 3nm production is not up to "full speed", then Apple has to prioritize which of its SoCs are built on the new process.
2) Then the Silicon team is too small
If by quickly you mean yearly that is true, but yearly is what we should be expecting anyway and why should we expect that they suddenly can keep up with a yearly pace next year if they can't keep up with a yearly schedule now?
The 14 Pro and Pro Max use the tweaked N4 process not N5.
Everything is somewhat guesswork since we don't know TSMCs exact yields, the only info we have is that they have been claiming volume production since September last year. When N5 launched, again I'll reiterate, the A14 and M1 launched within a small window...
Or maybe the M-series was always on an 18-month schedule for cost and practical reasons.
It's not like the headset project was sprung on Johny Srouji's team at the last minute.
The headset is slated to come out early 2024. They should have M3 processors done at that point. TSMC started 3nm production back in late 2022, so surely having an entire year should be enough time to manufacture the A17 chip and the M3 chip.
How do we know this? We've been told several times by leakers that Apple plans to have M-series on 18 month schedules.
For R1 to show up now, when M3 is expected, is a clear indication the design activities overlapped. So, there would have been resource constraints when R1 was dumped on the team. Now that we see the appearance of R1, it appears R1 had the priority for the resources and so we have a delayed M3. How big of a delay it not known at the moment but I think it could be as much as 12 to 18 months. It would likely take that long to bring a new team up to speed on the M series roadmap and get them productive. In hindsight, the delay likely explains why there have been no M3 leaks.
Only N3B began production in late 2022. N3E, which is what M3 is likely using, doesn't start until H2'23 per TSMC.
N3B is far more expensive and low-yielding than N3E.
Who was expecting M3? It took a year between M1 and M1 Pro. We saw an 18-month cycle between M1 and M2.
Those who were expecting M3 this year didn't have anything to back it up. It was nothing other than "it ought to be this way!" If you look at what actually happened with M1/M2 and the sales numbers for Mac, it doesn't support a 12-month cycle.
We don't even know what is R1. Does it have the complexity of H2 or S8? You started with one expectation and built an entire theory based that single thing.
You dont know this. It is entirely reasonable to think that the iPad + Mac sales can support a yearly cycle. We got A5X, A6X, A8X, A9X, A10X, A12X, when only the iPad Pro was supporting that special chips development. With the mac added in they have basically doubled the number of machines supporting the development of the bigger chip. So unless Apple’s SoC team has gotten dramatically less efficient recently a yearly cycle makes perfect sense.
The A#X chips are the predecessor to the M series And are not more complex than A series in a way that the later A#X series were not also more complex.
You haven't considered the cost of chip development and production. Ever wonder why fewer and fewer companies are using the latest node? Why there is basically only one company in the world producing competitive 3nm? The cost doubles with every generation. You can't look at the heyday of 32nm A6X and apply it to 5nm M1.
If you look at the revenue numbers for Mac and iPad, they are not increasing at the same rate as chip costs. With the M1 boom over, Mac sales are about the same as what Apple saw in 2019. iPad still hasn't returned to the glorious days of 2012-2014 where everyone was buying a new iPad every year.
So no, it's not reasonable to think M-series can support an annual cycle. Not until it's normal consumer behavior to replace iPad and Mac every year. Apple sells 200 million iPhones each year. iPad + Mac don't come even close. Think about that for a second.
Last edited:
You are assuming it costs the same to bring an M series chip to a new node as it does to bring the first A series chip. Is this true? I would assume that bringing an A series chip to a new node would make bringing the matching M series less expensive. Note that Apple should be able to do this, NVIDIA sells about 7 million desktop GPUs per quarter, the majority of which are not their super high end 4090s and that covers 3-4 different GPU silicone designs. Are you seriously trying to tell me that Apple sells, who sells that many Macs per quarter plus about another 10 million iPads (lets say 2 million of which are iPad Pros) cannot maintain just 2.5 pieces of additional silicon (M2,M2 Pro, M2 Max - the pro is a chopped max, hence 0.5)? Apple charges a huge premium for their products and I am not at all convinced that your case is compelling.
The fact that it takes 6-12 months for Apple launch Pro versions of the M1/M2 gives you an idea about the complexity and cost. It's not easy or Apple would be launching all variants on day one.
GPUs are much easier to design and scale, especially when you don't care about power. There's a reason why NVIDIA, Intel, AMD, Moore Threads, etc. all make them.
Just because Apple charges a premium for Macs doesn't mean consumers will upgrade more frequently. How many people want to upgrade from their $999 M1 MacBook Air just because M3 is out? Does M3 do something M1 doesn't? iPhone gets new cameras and AirTags and AirPods. Mac doesn't have anything to push it other than "it's faster this year."