Discussion Apple Silicon SoC thread

[Eug]

M1
5 nm
Unified memory architecture - LP-DDR4
16 billion transistors

8-core CPU

4 high-performance cores
192 KB instruction cache
128 KB data cache
Shared 12 MB L2 cache

4 high-efficiency cores
128 KB instruction cache
64 KB data cache
Shared 4 MB L2 cache
(Apple claims the 4 high-effiency cores alone perform like a dual-core Intel MacBook Air)

8-core iGPU (but there is a 7-core variant, likely with one inactive core)
128 execution units
Up to 24576 concurrent threads
2.6 Teraflops
82 Gigatexels/s
41 gigapixels/s

16-core neural engine
Secure Enclave
USB 4

Products:
$999 ($899 edu) 13" MacBook Air (fanless) - 18 hour video playback battery life
$699 Mac mini (with fan)
$1299 ($1199 edu) 13" MacBook Pro (with fan) - 20 hour video playback battery life

Memory options 8 GB and 16 GB. No 32 GB option (unless you go Intel).

It should be noted that the M1 chip in these three Macs is the same (aside from GPU core number). Basically, Apple is taking the same approach which these chips as they do the iPhones and iPads. Just one SKU (excluding the X variants), which is the same across all iDevices (aside from maybe slight clock speed differences occasionally).

EDIT:



M1 Pro 8-core CPU (6+2), 14-core GPU
M1 Pro 10-core CPU (8+2), 14-core GPU
M1 Pro 10-core CPU (8+2), 16-core GPU
M1 Max 10-core CPU (8+2), 24-core GPU
M1 Max 10-core CPU (8+2), 32-core GPU

M1 Pro and M1 Max discussion here:

Page 78 - Discussion - Apple Silicon SoC thread

Page 78 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

M1 Ultra discussion here:

Page 109 - Discussion - Apple Silicon SoC thread

Page 109 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

M2 discussion here:

Page 127 - Discussion - Apple Silicon SoC thread

Page 127 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

Second Generation 5 nm
Unified memory architecture - LPDDR5, up to 24 GB and 100 GB/s
20 billion transistors

8-core CPU

4 high-performance cores
192 KB instruction cache
128 KB data cache
Shared 16 MB L2 cache

4 high-efficiency cores
128 KB instruction cache
64 KB data cache
Shared 4 MB L2 cache

10-core iGPU (but there is an 8-core variant)
3.6 Teraflops

16-core neural engine
Secure Enclave
USB 4

Hardware acceleration for 8K h.264, h.264, ProRes

M3 Family discussion here:

Page 215 - Discussion - Apple Silicon SoC thread

Page 215 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

M4 Family discussion here:

Page 263 - Discussion - Apple Silicon SoC thread

Page 263 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[Eug]

What I can see...
- 17 is doing great, so the life of this device is secured.
- 17 Air is an interesting surprise but is being niche
- 17 Pro is nice, still, seems that the existance of this one is less justified if we consider a new model coming next year.
- 17 Pro Max is by far the most sold and by far the best of the bunch.

Don't misunderstand me, but despite I like the Pro series, I feel that Apple will pull the plug of the vanilla pro and put the Pro Max as the new Pro or just calling Max as the flagship of Apple.
Why?
- 17 is the new base iPhone and the improvements were so massive that people like it.
- 17 Air is a debut and expecting to be 2 years at least.
- 17 Pro Max is the flagship, so no.
- And there is the rumor of the iPhone Fold / Flip on the works.

And seeing that Apple maintain the 4 iPhones lineup, the vanilla Pro might be sacrified to put the new foldable phone.
Also, this implies that each model will be different each other and not be a clone. Even the e series has a reason to be:
- e series to be the low cost phone
- Vanilla to be the base one
- Air is the lightweight one
- Pro Max is the flagship
- Foldable to be the luxury one.

Let's see how this ends, because the vanilla Pro is selling really well, but has some issues like overheat despite the chamber, which the Pro Max is not noticeable due sizes. And again.. Apple might want a 4 phone lineup for next year, not counting the "e" tier

I disagree. The Pro and Pro Max sizing are fundamentally different. It's not really about the weight, but more about the dimensions. The 17 has comparable dimensions as the 17 Pro and is a wonderful device, but it is not a replacement for the Pro. From what I've observed online, people who found the 17 Pro Max too big didn't go for the 17, but went to the 17 Pro. Indeed, I myself returned a 17 Pro Max for a 17 Pro. And then there is the much, much smaller group who chose the Air. While the 17 may make sense for some people coming from something like a 12 Pro, it doesn't make much sense for people who want an upgrade from the 14 Pro.

What I suspect may happen is that somewhere down the line, the Pro and Pro Max will adopt some of the Air design language, but will keep the professional features.

 

[name99]

If this is true it was done for Apple's servers. They beefed up the A19/M5 generation GPU core to be much better at AI. So put as many of those GPU chiplets as can fit on a board with a minimal number of CPU chiplets, cover the backside of the board with DRAM and a few NAND stacks, then replicate as many are needed to fill a datacenter or two.

Any additional options for their consumer product offerings is a side effect.

The prophecy says otherwise
We shall see...

https://patents.google.com/patent/US20250094093A1

If you take it seriously, Apple seems to consider THE problem of datacenter-scale AI to be one of providing uniform memory (a single address space?) to a much larger unit of compute than a single Blackwell.
I've no idea how serious or even aspirational it is. It seems like to make it work would require providing more physical address bits than Apple currently uses AND being able to use high end virtual address bits that are currently used for tagging and security as genuine address bits. Maybe the overhead is small enough that they're willing to make that sort of change in every core down to the lowliest watch E-core?

If they do get it to work, the payoff is pretty nice! Presumably an easier programming model than something like a DGX, and probably also at substantially lower power.
Maybe even cheaper since all the expensive HBM gets replaced with lower-priced LPDDR5?

 

[mvprod123]

M5 Max?

 

[Doug S]

What I can see...
- 17 is doing great, so the life of this device is secured.
- 17 Air is an interesting surprise but is being niche
- 17 Pro is nice, still, seems that the existance of this one is less justified if we consider a new model coming next year.
- 17 Pro Max is by far the most sold and by far the best of the bunch.

Don't misunderstand me, but despite I like the Pro series, I feel that Apple will pull the plug of the vanilla pro and put the Pro Max as the new Pro or just calling Max as the flagship of Apple.
Why?
- 17 is the new base iPhone and the improvements were so massive that people like it.
- 17 Air is a debut and expecting to be 2 years at least.
- 17 Pro Max is the flagship, so no.
- And there is the rumor of the iPhone Fold / Flip on the works.

And seeing that Apple maintain the 4 iPhones lineup, the vanilla Pro might be sacrified to put the new foldable phone.
Also, this implies that each model will be different each other and not be a clone. Even the e series has a reason to be:
- e series to be the low cost phone
- Vanilla to be the base one
- Air is the lightweight one
- Pro Max is the flagship
- Foldable to be the luxury one.

Let's see how this ends, because the vanilla Pro is selling really well, but has some issues like overheat despite the chamber, which the Pro Max is not noticeable due sizes. And again.. Apple might want a 4 phone lineup for next year, not counting the "e" tier

I don't think we should assume there will be another Air. Why did they call it "iPhone Air" rather than "iPhone 17 Air", if it is going to be a part of the yearly lineup going forward? The Mini and Plus didn't get that treatment, because when introduced they were intended to continue on - and would have, if either had sold well enough.

At some point Apple is going to start using silicon-carbon batteries, and I think we all know Apple is NOT going to keep the same battery volume and deliver huge increases in the already long (except for Air) battery life. They'll use it to deliver the same battery life in a thinner/lighter form factor.

So the "Air" might go away simply because next year the other models (at least Pro/Max) become thinner. Not as thin as Air, but thin enough that having a phone whose entire existence is about "thin" is rendered redundant.

 

[LightningZ71]

I kind of like the notion that the Air is a sort of pipe cleaner for the Fold.

 

[DZero]

I don't think we should assume there will be another Air. Why did they call it "iPhone Air" rather than "iPhone 17 Air", if it is going to be a part of the yearly lineup going forward? The Mini and Plus didn't get that treatment, because when introduced they were intended to continue on - and would have, if either had sold well enough.

At some point Apple is going to start using silicon-carbon batteries, and I think we all know Apple is NOT going to keep the same battery volume and deliver huge increases in the already long (except for Air) battery life. They'll use it to deliver the same battery life in a thinner/lighter form factor.

So the "Air" might go away simply because next year the other models (at least Pro/Max) become thinner. Not as thin as Air, but thin enough that having a phone whose entire existence is about "thin" is rendered redundant.

The issue is the vapor chamber. Even with Si-Ca batteries, it takes space, and Apple is using the Air term as the thin phone.
Also vanilla iPhone won't get changes that much.

I kind of like the notion that the Air is a sort of pipe cleaner for the Fold.

The Fold might deliver something unique that can become an option

 

[The Hardcard]

There is a ceiling, it doesn't continue forever. Double performance could mean from 0.02s to 0.01s, marginal

Other real examples and quantify the percentage of users? Until now we have:

- horribly written ultra niche app for Truckers
- lidar
- "data science"

at best that's 4% of total users

There is no foreseeable ceiling since the demands on compute run directly alongside the increase in compute power. Your new percentage is still way too low.

Computational photography taxes the CPU, GPU, and NPU each year. While it’s a small percentage who need it regularly, a much higher percentage have those certain occasions or random occurrences where they want lots of pictures. And 2025 cameras fully tax 2025 SOCs. I’ve seen lots of complaints about camera lag during bursts, especially the Samsung S25 Ultra.

The apps become complex. The games become more demanding. The world continues to welcome more powerful SOCs and will for years.

 

[The Hardcard]

I don't think we should assume there will be another Air. Why did they call it "iPhone Air" rather than "iPhone 17 Air", if it is going to be a part of the yearly lineup going forward? The Mini and Plus didn't get that treatment, because when introduced they were intended to continue on - and would have, if either had sold well enough.

At some point Apple is going to start using silicon-carbon batteries, and I think we all know Apple is NOT going to keep the same battery volume and deliver huge increases in the already long (except for Air) battery life. They'll use it to deliver the same battery life in a thinner/lighter form factor.

So the "Air" might go away simply because next year the other models (at least Pro/Max) become thinner. Not as thin as Air, but thin enough that having a phone whose entire existence is about "thin" is rendered redundant.

I think they are planning to keep or even increase battery volume. As they try to come up with on-device AI killer apps, they have no doubt noticed that running generative AI drains the battery at a similar rate to AAA gaming.

Any victory in finding use cases that make people want to run AI models regularly enough for them to buy new phones, will also turn all day with 8 hours of screen time phones into 4 hours with 2 1/2 hours of screen time phones.

 

[johnsonwax]

The Fold might deliver something unique that can become an option

I can't imagine Apple would expand the line with both an Air and a Fold. Either the Air stays and the Fold replaces the Max, or the Fold replaces the Air. Apple probably already knows what that decision will be after the first few weeks of sales. My sense is that if they do a book Fold style, that'll replace the Max, and if they do a flip style that'll replace the Air.

 

[johnsonwax]

Computational photography taxes the CPU, GPU, and NPU each year. While it’s a small percentage who need it regularly, a much higher percentage have those certain occasions or random occurrences where they want lots of pictures. And 2025 cameras fully tax 2025 SOCs. I’ve seen lots of complaints about camera lag during bursts, especially the Samsung S25 Ultra.

iPhone users took a trillion photos last year. Almost everyone is using that computational photography compute. iPhones got NVMe before most PCs did because of that. They got NPUs years before any PC did because of that.

People think of compute as 'how big of a spreadsheet would you run on your phone' and not 'doing motion stabilization in realtime on 4K 60FPS video' which is WAY the f harder than any spreadsheet will push your computer. This is why A series had higher TOPS than most M series until recently.

 

[Eug]

I can't imagine Apple would expand the line with both an Air and a Fold. Either the Air stays and the Fold replaces the Max, or the Fold replaces the Air. Apple probably already knows what that decision will be after the first few weeks of sales. My sense is that if they do a book Fold style, that'll replace the Max, and if they do a flip style that'll replace the Air.

It wouldn't be a good idea for the Fold to replace the Max. The Max is one of Apple's bestsellers. The Air is projected to be Apple's worst seller this year, and the Fold is expected to be a very low seller too. The Fold replacing the Air doesn't make sense either, since it totally defeats the purpose of the Air. The Fold would weigh as much and be as thick as the Pro Max, which is very un-Air-like.

As mentioned previously, I would expect the Pro and the Max to adopt some of the Air's design language, while maintaining the professional features. Remember, the Pro and Max can actually be used as true professional video cameras (especially since the release of the 17 series). The Air is severely lacking in the camera department, and even if another Air is released with an upgraded camera setup, it will still be inferior to the Pros for several reasons. The 17 Pros finally addressed several major drawbacks of prior generations for videography, including proper cooling, open gate ProRes RAW support, Genlock, and battery life. The Pro Max is the pro cameraphone of choice, partially because of its larger screen, and Apple leans into this by offering the highest storage in that model only now.

As for the Fold, it's also may get inferior cameras to the Pros, and the extra features of the Fold are completely superfluous for professional videography, yet it will be much higher cost.

iPhone users took a trillion photos last year. Almost everyone is using that computational photography compute. iPhones got NVMe before most PCs did because of that. They got NPUs years before any PC did because of that.

People think of compute as 'how big of a spreadsheet would you run on your phone' and not 'doing motion stabilization in realtime on 4K 60FPS video' which is WAY the f harder than any spreadsheet will push your computer. This is why A series had higher TOPS than most M series until recently.

Yes, computational photography IMO is equally important if not more important than general compute capabilities these days in a phone. Unfortunately people on these forums always often this aspect of the SoCs because we can't benchmark it or play games on it or whatever.

 

[Doug S]

I think they are planning to keep or even increase battery volume. As they try to come up with on-device AI killer apps, they have no doubt noticed that running generative AI drains the battery at a similar rate to AAA gaming.

Having AI suck down the same level of sustained power as AAA games is the ultimate example of a solution looking for a problem. People aren't gonna be giving their phones problems that the AI will grind away on for an hour before coming up with an answer. Anyone who has anything more complex than interactive level work is going to the cloud for it. You aren't gonna need a bigger battery for AI. Apple certainly isn't going to design their phones that way unless the need for it is proven - it isn't a "build it and they will come" situation.

 

[The Hardcard]

Having AI suck down the same level of sustained power as AAA games is the ultimate example of a solution looking for a problem. People aren't gonna be giving their phones problems that the AI will grind away on for an hour before coming up with an answer. Anyone who has anything more complex than interactive level work is going to the cloud for it. You aren't gonna need a bigger battery for AI. Apple certainly isn't going to design their phones that way unless the need for it is proven - it isn't a "build it and they will come" situation.

On device AI will be interactive. It won’t be an hour long problem that drains the battery, it will be the repeated interactive engagement that drains the battery. Apple will have no choice but to design phones that way, the power to run AI agents and models is as important as the models themselves and we are 3 to 5 years away from it being a primary factor in device design.

 

[name99]

It wouldn't be a good idea for the Fold to replace the Max. The Max is one of Apple's bestsellers. The Air is projected to be Apple's worst seller this year, and the Fold is expected to be a very low seller too. The Fold replacing the Air doesn't make sense either, since it totally defeats the purpose of the Air. The Fold would weigh as much and be as thick as the Pro Max, which is very un-Air-like.

The Air serves TWO purposes.

One is to be very thin and thus have significant "hand feel" and "pocket feel". Like the Mini before it, this is something very important to a small number of vocal people, and apparently *unimportant* to most. Air sales may start to pick up as more people encounter it in the wild, actually try it rather than just see pictures. But, honestly, I doubt it,

Two is to signal "I have a different phone from the rest of you commoners" (which may be meant to mean "I am rich" or "I am unique" or whatever, that's a level of detail Apple no longer cares about).
THIS is, IMHO, the primary role the Air is serving this year. And it's the role that will be taken up by the Fold. So, Highlander Rules...

The remaining question is timing. People seem to be talking about the Fold next year. I don't know. This requires the iPhone XX to be a third unique and striking phone in 2027, and that seems a bit much. I could believe that the Air gets two years (maybe 2026 model gets A20, C2, etc; maybe it stays the same but gets a price drop?) and the Fold IS the XX?

A final question is whether we have any reason to believe (beyond just that this is what everyone else has done) that the Fold is in fact a clamshell...
An alternative is to have a thin flexible screen that can be rolled out (like that $85,000 LG OLED R TV). This looks different from every other phone and doesn't suffer from crease damage.
It does require carefully thinking how the roller mechanism works to both look nice and not be a hassle.
Do you want a longer phone (rollout from the bottom?) or a wider phone (rollout from the side)?
The "neat package" solution rolls out from the top, and keeps the camera bump aesthetics of the Air BUT it seems hard to get that to work with Facetime HW up there and in the way.

 

[Doug S]

On device AI will be interactive. It won’t be an hour long problem that drains the battery, it will be the repeated interactive engagement that drains the battery. Apple will have no choice but to design phones that way, the power to run AI agents and models is as important as the models themselves and we are 3 to 5 years away from it being a primary factor in device design.

Interactive engagement isn't going to be generating all that much load, because human response time is glacial compared to CPUs. The LLM isn't engaged while you're taking in what the AI says, or while you are typing/speaking your response. The average person won't spend hours in a "conversation" with their phone's AI. I just think you're massively overestimating the amount of energy use from a local LLM in real world scenarios.