More tiering on Apple Silicion chips going forward?

[neinjohn]

With the addition of a "Pro" slogan to a A-series SOC I wonder what A-SoC the vanilla iPhone 16 will receive next year.

Understanding that the Pro slogan will bring a separation of features, Apple may develop two tiers of A-SoCs that will expand to M-SoCs. A tier expanding from the Pro A-Soc and another from a more basic, non-Pro A-Soc. This tiering would follow to two different tiers of products on Apple's catalogue.

Products I assume would receive the Pro variation:

• iPhone Pro
• both iPad Pro
• actual design MacBook Air
• both MacBook Pro
• Mac Mini with the M-Pro SoC
• two-fan iMac 24''
• Mac Studio
• Mac Pro

Products I assume would receive the non-Pro variation:

• vanilla iPhones
• iPhone SE
• iPad Air, Mini and basic iPad (A-Soc)
• older design MacBook Air (if they keep it at a lower price option)
• Mac Mini with non-Pro Soc
• lower variant/basic iMac 24''
• Apple TV

I would not expect different E-cores and P-cores, instead I expect the 1-year cycle delay. Other adjustments could be on additional hardware as raytracing, memory (both type and GB), GPU core count and maybe the NPU.

What are your thoughts?

 

[Chuckeee]

Until this day, Apple has not used the sort of tiering system that Intel and AMD use with version 3, 5, 7 and 9 of their CPU.

Really,
What about A verses M chips used on iPads?
What about plain M, pro M, maxi M and Ultra M chips used in Macs?

 

[sack_peak]

iPhone A chips used to be called A4, A5, A6, etc without Bionic or Pro

They did a rebrand to A12 Bionic, A13 Bionic, A14 Bionic, A15 Bionic, etc

They're doing it again with A17 Pro, A18 Pro, etc.

For the M chips these are the equivalent

- M3 = Core i5
- M3 Pro = Core i7
- M3 Max = Core i9
- M3 Ultra = Xeon

Refresh cycle for iPhone A chips is 12 months.

M chips is 19.5 months apart but rumor is Apple wants it to be shortened to align closer to iPhone A chip tech generations.

M3 is likely to come out Q1 2024 before April 2024.

 

[neinjohn]

@Chuckeee and @sack_peak . Yup, I understood I was presenting my thoughts incorrectly comparing to Intel and AMD tiering and changed a lot of the text. What I mean is Apple start developing the base M-SoC (as non-Pro,Max,Ultra version) from a different base A-SoC.

 

[sack_peak]

@Chuckeee and @sack_peak . Yup, I understood I was presenting my thoughts incorrectly comparing to Intel and AMD tiering and changed a lot of the text. What I mean is Apple start developing the base M-SoC (as non-Pro,Max,Ultra version) from a different base A-SoC.

I think that was based on the A12Z chip.

 

[leman]

Too expensive, I doubt we will see it. We might see more differentiation between the Pro/Max chips, but they are tape-outs from the same design, so no extra expense here.

 

[theluggage]

What about A verses M chips used on iPads?
What about plain M, pro M, maxi M and Ultra M chips used in Macs?

The thing about Intel's i3/i5/i7/i9 labelling was that it was all about "target market segment" - i3 = entry, i5=midrange, i7=pro, i9 (added later) = 'goes up to 11' and had no consistent technical meaning between chip generations or between desktop, mobile and ultra-low-power ranges. I believe that, on a few occasions, this generation's i3 was actually last generations i5. All it really meant was that an i5 ultrabook was probably faster than an i3 ultrabook from the same range. It kept things simple for salespeople - but also dumbed things down (you really had to dig out the model number and look it up to be sure what you were getting).

We only have two data points for Apple's Mx plain/pro/max/ultra naming. Yes, it represents entry/mid/high/workstation class, but so far it also has a clear link to the hardware (regular = dedicated entry-level die, max = larger die with more of everything, Ultra = two Max's fused together, pro = Max die with some GPU cores "sliced off").

Really, Apple's M-series processor naming is a dumpster fire (this is all about marketing, not technical merit) - there's no unambiguous name for the "regular" version (if I say "M2 chips" do I mean the regular M2 or the whole M2 range?), they clash with the names of the actual products (...so the iPad Pro and the 13" MacBook Pro have a regular M2, the Mac Pro has a M2 Ultra chip, the Mac mini with M2 pro isn't a Mac Mini pro... seriously?) plus it puts the generation number up front, despite a M1 Max still running rings around a regular M2. I'm sure Mac users are smart enough to cope, but its still messy for a company with Apple's rep.

Thing is, Apple doesn't really need snappy processor names - unlike Intel, they're not making processors for other people's products so if they offer a "MacBook Pro" with a choice of "M-12/19-2, M2-12/30-2 or M2-12/38-2" those numbers aren't going to offend punters eyes until they have to decide how many GPUs etc. they want.

 

[dmccloud]

I think that was based on the A12Z chip.

A12Z was what Apple used in the Developer Transition Kits when Apple Silicon was first announced rather than what M1 was based on. M1 was based on the A14 cores, and M2 from A15. If M3 is based on 3nm, it will mark the first time the M series is based off the latest A-series core architecture instead of one generation behind.

 

[Basic75]

The thing about Intel's i3/i5/i7/i9 labelling was that it was all about "target market segment"

The other thing is that it's completely redundant. If you have an Intel 12700K you can look up that it's "classified" as an i7, but you can just as well omit that part of the name, it's not like there's an i5 or i9 chip with the 12700K model name/number, too.

 

[sack_peak]

A12Z was what Apple used in the Developer Transition Kits when Apple Silicon was first announced rather than what M1 was based on. M1 was based on the A14 cores, and M2 from A15. If M3 is based on 3nm, it will mark the first time the M series is based off the latest A-series core architecture instead of one generation behind.

Many here claim the A12Z of the iPad is different from the DTR.

I do not see any citation to support this so I wish not to be disagreeable.

What you pointed out helps my point of view that the Vision Pro will use the 3nm M3 for its better performance per watt to lengthen the ~2hr battery life further.

 

[dmccloud]

Many here claim the A12Z of the iPad is different from the DTR.

I do not see any citation to support this so I wish not to be disagreeable.

What you pointed out helps my point of view that the Vision Pro will use the 3nm M3 for its better performance per watt to lengthen the ~2hr battery life further.

Apple's $500 Developer Program Includes Tools and Resources for Transitioning to Apple Silicon, Plus a Loaner A12Z-Based Mac Mini

To help developers prepare for the Mac transition from Intel processors to Apple Silicon, Apple has launched a Universal App Quick Start Program,...

www.macrumors.com

On the hardware side, participants will receive exclusive access to a Developer Transition Kit (DTK), which resembles a Mac mini but uses Apple's A12Z Bionic chip from the latest iPad Pro as its brains. In addition to the A12Z Bionic, the DTK includes 16GB of RAM, a 512GB SSD, a pair of 10 Gbps USB-C ports, a pair of 5 Gbps USB-A ports, and an HDMI 2.0 port. Thunderbolt 3 support is not included.

If you want to see this for yourself, check out Luke Miani's teardown of a DTK.

Apple stated at WWDC that Vision Pro will be using M2. There will not be enough 3nm production to supply iPhones, Macs and Vision Pro for a while, given that TSMCs yields are still in the 50-60% range based on recent reports.

 

[thenewperson]

A12Z was what Apple used in the Developer Transition Kits when Apple Silicon was first announced rather than what M1 was based on. M1 was based on the A14 cores, and M2 from A15. If M3 is based on 3nm, it will mark the first time the M series is based off the latest A-series core architecture instead of one generation behind.

?

M1 and M2 were also based off the latest A-series architecture.

 

[sack_peak]

Apple's $500 Developer Program Includes Tools and Resources for Transitioning to Apple Silicon, Plus a Loaner A12Z-Based Mac Mini

To help developers prepare for the Mac transition from Intel processors to Apple Silicon, Apple has launched a Universal App Quick Start Program,...

www.macrumors.com

If you want to see this for yourself, check out Luke Miani's teardown of a DTK.

Apple stated at WWDC that Vision Pro will be using M2. There will not be enough 3nm production to supply iPhones, Macs and Vision Pro for a while, given that TSMCs yields are still in the 50-60% range based on recent reports.

Apple has historically bumped up to better CPUs without changing MSRP after the officially announcement was made 2x within the Intel era.

Within this week there was rumors of new Mac laptops and the M3 making its release before 2024.

More than a year ago there was news of 3nm M2 chips.

I am talking about the iPad Pro chip itself. Whatever improvements made to make it a DTK is outside of the chip and on the logic board.

Same with A14 or A15-based iPads with USB-C ports having a throughput exceeding 480Mbps. Chips outside of the SoC likely allows for that faster than 480Mbps speed vs same SoC iPhones with Lightning ports.

Again, I am talking about SoC being the same while improvements made on the logic board makes them different.

Pls do not confuse the iPad chip with iPad chip + logic board.

 

[dmccloud]

Many here claim the A12Z of the iPad is different from the DTR.

I do not see any citation to support this so I wish not to be disagreeable.

What you pointed out helps my point of view that the Vision Pro will use the 3nm M3 for its better performance per watt to lengthen the ~2hr battery life further.

1. The A12Z in the DTK was identical to the A12Z used in the iPad, just on a logic board that was a strange hybrid of an Intel Mac Mini form factor and something you would see in an ARM-based system. This was something that was clarified in the developer documentation for the DTK.

2. Apple announced Vision Pro would use M2 back at WWDC:

These groundbreaking innovations are powered by Apple silicon in a unique dual-chip design. M2 delivers unparalleled standalone performance, while the brand-new R1 chip processes input from 12 cameras, five sensors, and six microphones to ensure that content feels like it is appearing right in front of the user’s eyes, in real time. R1 streams new images to the displays within 12 milliseconds — 8x faster than the blink of an eye. Apple Vision Pro is designed for all-day use when plugged in, and up to two hours of use with its external, high-performance battery.

Given that 3nm yields are still below 70% (some estimates put it closer to 50%, lower than even Samsung's yields with their own 3nm process), there would be supply issues at play with the available silicon. If Apple was to put an M3 into Vision Pro, they would have to cannibalize the available supply of 3nm silicon from either the iPhone or Mac lineups.

3. The quoted section above also illustrates why switching to M3 would likely have no significant impact on battery life, and honestly battery life is one of those "nice to have" things unless you insist on using the device without any cords connecting it. The original Oculus required a hardline connection to a PC in order to operate as well, so there is an established precedent for tethered VR setups.

 

[dmccloud]

Apple has historically bumped up to better CPUs without changing MSRP after the officially announcement was made 2x within the Intel era.

Within this week there was rumors of new Mac laptops and the M3 making its release before 2024.

More than a year ago there was news of 3nm M2 chips.

I am talking about the iPad Pro chip itself. Whatever improvements made to make it a DTK is outside of the chip and on the logic board.

Same with A14 or A15-based iPads with USB-C ports having a throughput exceeding 480Mbps. Chips outside of the SoC likely allows for that faster than 480Mbps speed vs same SoC iPhones with Lightning ports.

Again, I am talking about SoC being the same while improvements made on the logic board makes them different.

Pls do not confuse the iPad chip with iPad chip + logic board.

You contradict yourself multiple times here. In this thread, you have previously claimed that the A12Z used in the DTK was different from the A12Z used in iPads, but you admit you don't have a source for that. Yet here you claim that the A12Z wasn't changed but the logic board was, which completely contradicts the first claim that the A12Z itself was different. As far as the Intel era goes, two examples of a CPU swap in 14+ years does not make a trend, and was most likely affected by what Intel was shipping to Apple. If anyone is confusing "the iPad chip with the iPad chip + logic board", it's you, not me.

 

[sack_peak]

You contradict yourself multiple times here. In this thread, you have previously claimed that the A12Z used in the DTK was different from the A12Z used in iPads, but you admit you don't have a source for that. Yet here you claim that the A12Z wasn't changed but the logic board was, which completely contradicts the first claim that the A12Z itself was different. As far as the Intel era goes, two examples of a CPU swap in 14+ years does not make a trend, and was most likely affected by what Intel was shipping to Apple. If anyone is confusing "the iPad chip with the iPad chip + logic board", it's you, not me.

I think you misunderstood what I read.

Could you bullet point what you are trying to say in all your replies to me within the last 24 hours?

 

[mr_roboto]

Given that 3nm yields are still below 70% (some estimates put it closer to 50%, lower than even Samsung's yields with their own 3nm process), there would be supply issues at play with the available silicon. If Apple was to put an M3 into Vision Pro, they would have to cannibalize the available supply of 3nm silicon from either the iPhone or Mac lineups.

I disagree. The first generation Vision Pro isn't going to be a volume product, and Apple knows it - that's why they set the entry level price at $3500, which is going to price out everyone but rich "influencers" and developers who think it has potential. If they wanted to upgrade to M3, they could do so without much fear of causing supply chain issues.

But right now, they need the Vision Pro team to be 100% focused on shipping what they've already designed. Apple needs units in the hands of developers as soon as possible, because those are the people whom Apple hopes will design killer apps to drive consumer demand for the eventual second generation Vision Pro. The only scenario where they might want to reconsider is if any other problems happen push the release date out enough that upgrading to M3 won't cause additional delays.

 

[tenthousandthings]

iPhone A chips used to be called A4, A5, A6, etc without Bionic or Pro

They did a rebrand to A12 Bionic, A13 Bionic, A14 Bionic, A15 Bionic, etc

They're doing it again with A17 Pro, A18 Pro, etc.

For the M chips these are the equivalent

- M3 = Core i5
- M3 Pro = Core i7
- M3 Max = Core i9
- M3 Ultra = Xeon

Refresh cycle for iPhone A chips is 12 months.

M chips is 19.5 months apart but rumor is Apple wants it to be shortened to align closer to iPhone A chip tech generations.

M3 is likely to come out Q1 2024 before April 2024.

I'm sure you know this, but the labels do have meaning and purpose. A10 "Fusion" was the first to "fuse" two types of cores (efficiency and performance), then A11-A16 "Bionic" indicates the Neural Engine. Now, A17 "Pro" indicates mesh shading and hardware ray tracing. Software that uses the Neural Engine requires a Bionic SoC, and now software that takes advantage of mesh shading and hardware ray tracing requires a Pro SoC.

That said, the choice of the "Pro" label for A17 raises questions, given the precedent of the M1 Pro/Max and the M2 Pro/Max. You have to ask what they'll do with that. I feel confident in saying that Apple's marketing people could have easily come up with a different name to replace Bionic if the only purpose were to indicate the new graphics capabilities. But "Pro" was chosen. I believe it will make sense once the M3 cycle is complete and the iPhone 16 is released. There is, however, no obvious answer right now that I can see.

I don't think the OP neinjohn's initial guess of some kind of two-tier structure across both mobile devices and Macs is likely. I guess maybe a non-Pro A17 or A18 might be coming for basic iPhone and iPad, but IMHO there's just no way that would translate to Macs. The plain M3 graphics will have mesh shading and hardware ray tracing.

Maybe that's the meaning of the choice. You're getting M-series graphics in your A-series device if you've got a "Pro" SoC.

On the timetable, I'm still betting on a two-year cycle for M silicon, subject to the needs of product releases, especially those that require events. Sometimes it will be less than two years, other times more. The original rumor was an 18-month cadence. We'll see.

 

[sack_peak]

I'm sure you know this, but the labels do have meaning and purpose. A10 "Fusion" was the first to "fuse" two types of cores (efficiency and performance), then A11-A16 "Bionic" indicates the Neural Engine. Now, A17 "Pro" indicates mesh shading and hardware ray tracing. Software that uses the Neural Engine requires a Bionic SoC, and now software that takes advantage of mesh shading and hardware ray tracing requires a Pro SoC.

That said, the choice of the "Pro" label for A17 raises questions, given the precedent of the M1 Pro/Max and the M2 Pro/Max. You have to ask what they'll do with that. I feel confident in saying that Apple's marketing people could have easily come up with a different name to replace Bionic if the only purpose were to indicate the new graphics capabilities. But "Pro" was chosen. I believe it will make sense once the M3 cycle is complete and the iPhone 16 is released. There is, however, no obvious answer right now that I can see.

I don't think the OP neinjohn's initial guess of some kind of two-tier structure across both mobile devices and Macs is likely. I guess maybe a non-Pro A17 or A18 might be coming for basic iPhone and iPad, but IMHO there's just no way that would translate to Macs. The plain M3 graphics will have mesh shading and hardware ray tracing.

Maybe that's the meaning of the choice. You're getting M-series graphics in your A-series device if you've got a "Pro" SoC.

On the timetable, I'm still betting on a two-year cycle for M silicon, subject to the needs of product releases, especially those that require events. Sometimes it will be less than two years, other times more. The original rumor was an 18-month cadence. We'll see.

Of course there is a corresponding meaning of the rebrand.

But essentially the person needs a phone, tablet, laptop, etc and they need the least amount of time to ID what to buy and the branding helps.

 

[Wokis]

It's called Pro because Apple financed TSMCs N3B node and designed a SoC, the A17 Pro, around that.

Now, this node is rumored to be pretty terrible in terms of yield (= expensive chips) and could within a year be phased out entirely for N3E. That node is very different from N3B and will likely require a substantial redesign of the A17 with some different characteristics and performance numbers. Possibly for the worse.

Hence, they can drop "Pro" for the newer A17 that is put into the iPhone 16.

If they will repeat this pattern, I don't know. For sure TSMC wont have an experimental new node aligning perfectly with the iPhone launch every year, though.

 

[Boil]

pro = Max die with some GPU cores "sliced off"

Mn Pro is not a Mn Max with "some GPU cores sliced off"...

Apple has three discrete Mn-series dies; Mn, Mn Pro, & Mn Max...

Two Mn Max dies can be connected on a package (via the UltraFusion connector) to make a Mn Ultra...

There may also be a future product which connects four Mn Max dies to make a Mn "Extreme" package...