Yearly Upgrades of M-series Chips

[4087258]

I think M3 gets announced, along with its high-end variants. It HAS to be. The name "Scary Fast" would be a strange choice if they don't actually have anything to announce that is scary fast relative to current products.

I suppose there's the possibility of some new, even higher-end variant of the M2 chip for their Mac Pros, particularly with the disappointments around how they compare to the now-discontinued Intel ones. Maybe this chip would further expand the tight RAM limit the new Mac Pro suffers from. Given what a niche product the Mac Pro is though, I don't feel like this would deserve a big announcement all by itself and I can't see a chip like that being available for their other Macs.



I think yearly is the only long-term possibility.

I know people basically settled on the 18-month idea at the beginning, but that always seemed incredibly strange to me, both because there was zero evidence for it and because of how insanely weird and impractical a schedule it would be for them to maintain with how the tech is researched and made. I mean, their A-series chips, which are based on the same tech, are already released annually. How exactly would that even work, to not sync the M-series chip releases to that in any way? I think this only sounded so natural to people because of the haphazard way they've had to release Macs in the past with how they were tied to Intel's fall-on-face unpredictable chip releases. It doesn't make sense anymore in the new world of Apple Silicon. Supply issues still leave them with some unpredictability for now but that will resolve over time as they mature and diversify their supply sources and as the COVID-19 crisis moves further and further away in the rearview mirror.

So that just leaves a 1-year or 2-year release cycle. 2 years would be the most sensible given how mature this tech is and how unnecessary frequent speed increases are to most of the target audience. It would also mean they could support their Macs for twice as long with the same effort, and a long support life is particularly important to Mac users. But it would sit poorly with their shareholders because of the reduced hype and profits. So I think we are going to be seeing 1 year. Certainly that's what the evidence has been pointing to very, very strongly for a while now.



This I think is a really clever point and probably the only thing that gives me any real pause about everything I said above. Maybe though they just quietly update it to the M3. Or they leave it on the M2 because there's an understanding that this is a niche, 1.0 release that they have no expectation of widespread adoption for yet. I would also imagine, given the heat, weight, and sizing nightmare they've had with this thing that it's non-trivial to equip it with the latest and greatest chip generation going forward and maybe they are resigned long-term to that reality.

I'll end this by pointing out that they don't have to release products to announce them. They could announce new MacBook Pros for example but not actually release them until next summer. They could even just use this event to showcase the performance of the M3 chips that future devices will come with, sort of like what they did when they announced Apple Silicon in the first place.

Thanks for all the writing. You clarified a few things and made me wonder even more about the possibilities.

 

[no-reply]

Given how far ahead the M-series is, and some aspects that nobody else (even the higher performers) can touch like all-day battery and running cool, without needing to be plugged in to feel it... Apple can update it a lot less than it has been. No harm in producing hits often though. I'm sure someday it will slow down but for now, we can use all the power we can get for efficiency and things like Vision Pro

 

[MBAir2010]

is there a real difference form the M! to M2?
especially with 16 cores?
and
how many years will our friends at  Cupertino support their own M1 chip?

thanks!

 

[leman]

is there a real difference form the M! to M2?

There are plenty of differences. Depends on what you are interested in?

 

[MBAir2010]

There are plenty of differences. Depends on what you are interested in?

I have a MacBook Air m1 and love that, especially with Monterey.
Sonoma is nice fast but no personality like Monterey.
I was on the fence over purchasing either a MacMini M1 ($429) or the refurb M2, ($509)
so I just (13 minutes ago) purchased the M1 with 16 cores since.... well Monterey.

doctor
doctor
what is wrong with me?

/...with me with me...

 

[Chuckeee]

If the M3 will be announced, it will make the Vision Pro look odd with the promised M2 chip… you’d expect the latest and greatest for their new flagship product next year. Or maybe it will get the M3 after all?

I don’t really understand the consternation about if the Vision Pro has an M2 verses an M3 processor. While the vision pro might represent the future, for now it’s a high cost boutique item for a small niche market.

Is anyone going to say, “I was going to get a Vision Pro, but now that it’s stuck with an M2 processor, and not an M3, forget about it”

Or the converse. “I wasn’t interested in a Vision Pro but now that it has an M3 I can’t wait to buy one.”

I just don’t see how the buying decision for a $3500 item that has a few hours of life, per battery charge, is really going to be effected whether it has an M2 or an M3 chip.

 

[Chuckeee]

Annual updates don’t always have to always be new generation of Mx processors. Updates to displays, batteries, interfaces could be the gimmick for a new generation. Kinda of like auto sales, you want an annual update but every update doesn’t need a new powertrain

 

[picpicmac]

They could announce new MacBook Pros for example but not actually release them until next summer.

Not going to happen. Rumor monger sez 8 Nov 2023 for availability.

We must remember that computer systems are more than just CPUs. And here I am using "CPU" for the entire Mx package because it's silly to keep using "GPU" and "Neural Engine" and claim this package is still in the spirit of the original ARM. The whole thing is just one big fat Central Processing Unit.

Anyway, for Macbooks one can argue that the display and battery life are far more important than the CPU.

And for a product like the iMac, the industrial design, ease of use, and price are far more important than the CPU.

Apple can do yearly updates of its products regardless of TSMC's own roadmap.

And I suspect that is why MBPs are on the menu for Monday night. New displays, faster refresh displays, longer battery life (perhaps enabled due to more efficient CPUs).

If the 5PM Pacific time is indeed intended to allow a companion company in Japan (or elsewhere in that region) to announce gaming support, then I think everyone who self-declares as a "gamer" will look with disappointment if the new iMac or MBP displays can't do significantly better than 60Hz.

"Scary Fast" could apply to 120Hz or even 144Hz screens.

 

[Analog Kid]

I really hope they don't move to an annual cadence for Macs. It is a lot of work to release a new product, and I'd rather them spend that energy making progress rather than just to release a minor update.

It's almost to the point I think the phones should go to a slower cadence, but at least they have the volumes to justify the effort. Macs don't. If there's something really worth releasing, fine. But no point in more frequent releases just to keep abreast of incremental benchmark advancements.

 

[bousozoku]

Thank you, sir! You know, I don’t want to get people upset or aggressive over such a hypothetical subject.

I don’t want to argue with anyone, that’s not my intention… So I just apologize 😄

At any moment in time, people will argue about anything. Some just like to argue or to be seen as correct.

There used to be the Apple vs. Commodore arguments. Then, the Atari vs. Apple vs. Commodore arguments, and so on.

It's better to just wait to see what will happen. When the PowerPC processors arrived, it was amazing and quickly, the excitement faded until there was nothing but disappointment. Hopefully, where Apple is going will be much better than that, but I'm not holding my breath while waiting.

 

[picpicmac]

it was amazing and quickly, the excitement faded until there was nothing but disappointment.

And that sort of applies to life in general, too.

 

[MBAir2010]

Given the recent rumors on Apple releasing the M3 chip at the same time as M3 Pro and M3 Max, I believe this indicates that Apple is going to upgrade the M-series chips on a yearly basis from now on. Yearly upgrades may give Apple a continuum leading on benchmarks on any given quarter, as Intel may not be able to have the lead because of it’s current yearly generations.

Let’s talk about that.

EDIT: I’m sorry if this topic wasn’t a good topic for you; it’s just a share of opinions and excitement. Be kind!

great post, this topic was on my mind lately.
my concern is what OS we like as well as what chip can run that.

I just hope  develops for every chip until the solar flare, judgement day and Yosemite burping.

 

[reklex]

I think AI could change that scenario. I Agree with the rest.

while AI is the future it won’t instantly solve all the problems modern computers face.

If you told me the new laptop could last 4 more hours than my current laptop that would be a more compelling reason to upgrade than stable diffusion being a few seconds faster.

 

[vigilant]

I think AI could change that scenario. I Agree with the rest.

AI could change it for developers.

But rationally speaking, to the average user anything Apple does for the next two to three years.

Most of the AI people use is running off of models that were trained.

AI is going to change sooooo many things including Siri.

The hard part of AI will always come from building the models. We’ll see things accelerate on Apple faster than the Intelosers. But that isn’t where the consumer will see the impact.

 

[theorist9]

I'd strongly prefer to see yearly upgrades for the M-series chips.

First, that's certainly achievable, in a meaningful way. TSMC likely won't be able to offer a new process (e.g., going from 5 nm to 3 nm) annually. But it's certainly able to offer at least annual process refinements (e.g., N5-> N4P, both of which are 5 nm). Plus Apple has been able to update its A-series microarchitecture annually for many years, and the M-series and A-series chips are based on the same microarchitecture. Thus, in summary, we should expect to have both a new microarchitecture, and a new process or process refinement, every year. That's certainly enough to constitute an upgrade.

Second, regular annual upgrades benefit the consumer. People like having the simple predictabilty of the iPhone's annual upgrades. That way, regardless of when you are considering a new iPhone, you know exactly when the new version will come out, and thus can decide to buy or wait on that basis. If you had this with the Mac, there would be no more reading tea leaves to guess whether, if you wait X months, a new Mac will be available. Instead, you'll know, and can make the decision with that information at hand.

Plus, on average, you'll be getting a more recent Mac. With upgrades every two years, the time-averaged purchase will be a machine released one year ago. With annual upgrades, it will be a machine released 6 months ago.

Further, while the Mac divison is smaller than the iPhone division, it's only small by comparsion. The Mac division by itself would be a Fortune 100 company. And I know of no other Fortune 100 company that mainly sells products, yet somehow lacks the capability to update <10 products annually (13" Air, 15" Air, Mini, Studio, 13"/14"/16" MBP's, 24" iMac, Mac Pro), even if each of those is available with multiple configurations. In fact, I know of no Fortune 100 company that mainly sells products and offers as few different products as Apple's Mac division (adding the two displays gives 11, ignoring accessories).

 

[falainber]

And about Intel, I think that it’s interesting how the 14th-generation had such a relatively small gain on performance compared to the 13th-generation.

Apple‘s M3 family may show much better results.

Intel's 14th generation is a new generation in name only. It is using the same process and architecture as the previous two generations. Besides, they released these three "generations" in about 18 months. Their true new generation (Meteor Lake) is coming in December. It does not make sense to compare release cadence of Intel and Apple. For Apple, it makes little sense to release a marginally better model. Nobody is going to buy it. It's different for Intel.

 

[MBAir2010]

. Their true new generation (Meteor Lake) is coming in December.

wonder how long the intelidots decided on this particular name?
is there even a meteor lake in existance?

(no ridicule towards you)

 

[Abazigal]

I always assumed that we would be seeing a 2-year upgrade cycle for the M-chip because Apple doesn't sell enough Macs to make an annual refresh worth the R&D (unlike their iPhones). There's also the question of whether we will see enough of an improvement year on year for it to even make sense.

They don't sell their chips to anyone else either, and people are holding on to their devices longer also. We see it in the iPad lineup, where the Ax series chips typically get a 1.5 year refresh cycle.

Moving forward, I won't be surprised to see their processors get refreshed every 3 years. As long as Apple continues to lead in terms of power efficiency, I don't think they are too worried about intel / AMD chips being slightly faster in the high end, especially when the only way they achieve this is by consuming massive amounts of power, which means that they will never work in the form factors that Apple traditionally favours and currently holds sway in.

 

[falainber]

wonder how long the intelidots decided on this particular name?
is there even a meteor lake in existance?

(no ridicule towards you)

Yes, there is such lake: Meteor Lake, Siskiyou County, California, United States.
Intel uses geographical names for their projects to avoid trademark issues. These are internal names anyways. These are not better or worse than, say, internal Apple names for their projects (examples being: Purple, Safari Pad, Titan, Marzipan).

 

[mectojic]

Like all Apple announcements, they might, or they might not.
The logic of their decisions can fall apart due to need for profits in a certain quarter, or supply chain delays.

 

[Allen_Wentz]

Given the recent rumors on Apple releasing the M3 chip at the same time as M3 Pro and M3 Max, I believe this indicates that Apple is going to upgrade the M-series chips on a yearly basis from now on. Yearly upgrades may give Apple a continuum leading on benchmarks on any given quarter, as Intel may not be able to have the lead because of it’s current yearly generations.

Let’s talk about that.

EDIT: I’m sorry if this topic wasn’t a good topic for you; it’s just a share of opinions and excitement. Be kind!

IMO it is clearly wrong-headed thinking to suggest "that Apple is going to upgrade the M-series chips on a yearly basis."
A) Chips are not Apple's products, they are simply one component of any product. Various products may indeed be on annual cycles, but a new chip does not always define each product iteration.
B) Each chip has multiple usages that Apple may vary over time. E.g. M3 may drive Studio while an M2 still serves MBA just fine. It is all about (hugely complex) yields. Buyers need best value, not some chip number.
C) Chip production, new nodes, etc. are not annual and will not be annual. Especially now that TMSC chip process is allegedly down to the (really small!) ~3 nm size range, we should expect process evolution to be different than it has been: less rapid obvious physical changes.
D) Software changes, i/o changes, comms, video, etc. may all evolve independently concurrent with an existing chip.

 

[TracerAnalog]

I don’t really understand the consternation about if the Vision Pro has an M2 verses an M3 processor. While the vision pro might represent the future, for now it’s a high cost boutique item for a small niche market.

Is anyone going to say, “I was going to get a Vision Pro, but now that it’s stuck with an M2 processor, and not an M3, forget about it”

Or the converse. “I wasn’t interested in a Vision Pro but now that it has an M3 I can’t wait to buy one.”

I just don’t see how the buying decision for a $3500 item that has a few hours of life, per battery charge, is really going to be effected whether it has an M2 or an M3 chip.

Consternation is a big word. I don’t understand why Apple would commit to a M2 chip for a product that is destined to be the future of spatial computing. The M3 will have raytracing support (amongst other improvements)! Why would Apple ‘freeze’ the Vision Pro with an M2 chip knowing that the M3 is just around the corner and available a half year before they intend to start selling it (allegedly in small volumes)? It just looks odd to me.

Whether people will buy it or not depends on many factors, but the older M2 chip would be another item on my ‘doubt list’. I think I will probably wait for a chip upgrade, precisely because it is so expensive.

 

[IG88]

I'd strongly prefer to see yearly upgrades for the M-series chips.

First, that's certainly achievable, in a meaningful way.

That's how you end up with 14nm++++++++++++

 

[bousozoku]

And that sort of applies to life in general, too.

You make progress with what you control, and the rest, you can accept for whatever is available.

Once the PowerPC 604e went to 275 MHz, the Intel crowd was angry. Then, suddenly, we were looking at the less-than-powerful G3. Too bad there wasn't any Mac OS X at that point.

 

[leman]

I'd strongly prefer to see yearly upgrades for the M-series chips.

First, that's certainly achievable, in a meaningful way. TSMC likely won't be able to offer a new process (e.g., going from 5 nm to 3 nm) annually. But it's certainly able to offer at least annual process refinements (e.g., N5-> N4P, both of which are 5 nm). Plus Apple has been able to update its A-series microarchitecture annually for many years, and the M-series and A-series chips are based on the same microarchitecture. Thus, in summary, we should expect to have both a new microarchitecture, and a new process or process refinement, every year. That's certainly enough to constitute an upgrade.

Second, regular annual upgrades benefit the consumer. People like having the simple predictabilty of the iPhone's annual upgrades. That way, regardless of when you are considering a new iPhone, you know exactly when the new version will come out, and thus can decide to buy or wait on that basis. If you had this with the Mac, there would be no more reading tea leaves to guess whether, if you wait X months, a new Mac will be available. Instead, you'll know, and can make the decision with that information at hand.

Plus, on average, you'll be getting a more recent Mac. With upgrades every two years, the time-averaged purchase will be a machine released one year go. With annual upgrades, it will be a machine released 6 months ago.

Further, while the Mac divison is smaller than the iPhone division, it's only small by comparsion. The Mac division by itself would be a Fortune 100 company. And I know of no other Fortune 100 company that mainly sells products, yet somehow lacks the capability to update <10 products annually (13" Air, 15" Air, Mini, Studio, 13"/14"/16" MBP's, 24" iMac, Mac Pro), even if each of those is available with multiple configurations. In fact, I know of no Fortune 100 company that mainly sells products and offers as few different products as Apple's Mac division (adding the two displays gives 11, ignoring accessories).

Excellent post!

I agree that regular yearly updates are great both for the customer and for Apple. It helps keep buyer enthusiasm up, reduce prices, make sustained use of manufacturing capabilities, as well as deliver consistent improvements. The upgrades don't need to be large to make sense. In fact, I think that smaller regular upgrades are a better strategy. To keep the business going it's enough if a significant enough group of users has reasons to upgrade. But you don't actually want everyone to upgrade at the same time as it leads to shortages and customer frustration.