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

 

[poke01]

N3P vs N3E tho.

It brings greater efficiency and performance than the node jump provides. These are mainly architectural improvements.

 

[Doug S]

If they wanted to launch a new A-series-based inexpensive Mac I could see Apple holding off on the M-series products to give that new product some time to breathe.

Minis don't have to wait for an OLED or anything else really. There's always a possibility they could launch an A-series laptop this fall alongside an M-series Mini while holding off the Pro chips and products until Spring.

I don't know how much I like the idea of them doing that, but it's certainly something they could do or is at least conceivable for them to do. Apple is always fairly predictable until they aren't.

But why hold off the Pro chips? N3P is a mature node from the day it was born. They might need a delay for larger dies (esp. Max) with N2 right out of the gate, but not N3P.

It wouldn't make sense for Apple to introduce a cheaper Mac and hold off the more profitable higher end Macs in its favor. Look at how they do it with iPhone, the SE and now 'e' versions follow the real stuff six months later to give the more profitable ones time on the market before throwing out the lower profit little brother.

Not saying they won't release an A series Macbook this year and M5 Macs next, but if they do it won't be because they didn't want to give the cheap one time alone on the market to "breathe". And if they do I'd guess we'll see it use A18P not A19P.

 

[smalM]

ChipWise A19 die shot


A18 GPU core vs. A19 GPU core

 

[Eug]

ChipWise A19 die shot

View attachment 130860View attachment 130861
A18 GPU core vs. A19 GPU core

 

[poke01]

ChipWise A19 die shot

View attachment 130860View attachment 130861
A18 GPU core vs. A19 GPU core

Pretty impressive. It’s basically the same size considering they added the Neutral cores. but with 30-40% faster rasterisation.

 

[Covfefe]

Pretty impressive. It’s basically the same size considering they added the Neutral cores. but with 30-40% faster rasterisation.

I'm constantly amazed by the seemingly endless advancements for GPUs (by all companies). It makes you think, if Apple had today's GPU arch 5 years ago, how much faster could the M1 GPU have been?

 

[mvprod123]

Lol

https://twitter.com/x/status/1970841735443210501

 

[DZero]

Lol

https://twitter.com/x/status/1970841735443210501

XDDDD
And how much goes with the G5 performance?

 

[name99]

I think this is simply because AI was not 'the thing' when M3 and A17 were being designed. ChatGPT first launched 2 years 10 months ago (yes, the AI revolution is less than 3 years old). M3 was already designed then. So Apples AI ambition in the Mac was extremely narrow while the M3 was being designed, while the NPU in the A17 was there for doing pretty serious computational photography/FaceID, in realtime or near-realtime, so it couldn't really be modest and still work. Again, Apple shipped NPUs in A series starting in 2017 - 8 years ago - back in the A11. It's doing an ENTIRELY different thing. The biggest use case for ANE was Face ID. Still is. And I wouldn't be surprised if more TOPS-hours were spent doing FaceID than all chatbots combined.

By launch, the M3 sort of mismatched to where the market had quickly gotten to, but it's not like anyone else in the PC space was shipping an NPU in their CPU - Apple was still the only one even offering an NPU. Everyone was doing their AI on GPU. x86 didn't get there until the M4 got there. They had the same timing problem.

That's one example but there are other cases of this sort of mismatch.
For example the E core on the M3 Max is more sophisticated than the E core on the M3 and M3 Pro.

That's my larger point, that Apple's HW design seems "componentized" enough that it's practical for IP blocks to be moved forward from a later project, or, alternatively, have the older/pre-existing IP block reused if the new block looks like it's being delayed by a month or three.

 

[name99]

When I heard the rumors about M5 Macs not launching until next year I immediately posted wondering if that was about the timing of N2. I can't really see any other reason why Apple would wait, since an N3P based M5 cannot be the source of the delay. At least there's nothing chip related I can see, it would have to something else like switching to OLED displays that affected the launch timing - but if they were waiting until next year for that they might as well do N2 as well.

LPDDR6 is not going to be the reason for delay, nor is it going to deliver more bandwidth out of the gate. The initial JEDEC speed is -10666, and you can already get LPDDR5X at that speed. I don't know about vendor plans, maybe they will skip forward to -12800 at launch but regardless it isn't even in mass production yet. Do you really believe Apple would introduce that level of third party schedule risk just to get a (maybe) small bump in bandwidth?

You won't necessarily get a bandwidth bump but you should get a power reduction which is always nice.

Of course whether there's a schedule risk is precisely the point -- we don't know the timing but you could imagine Apple thinking something like "Let's wait till February; if the numbers look OK we'll go ahead with LPDDR6 otherwise we fall back to LPDDR5X"

 

[Doug S]

You won't necessarily get a bandwidth bump but you should get a power reduction which is always nice.

Of course whether there's a schedule risk is precisely the point -- we don't know the timing but you could imagine Apple thinking something like "Let's wait till February; if the numbers look OK we'll go ahead with LPDDR6 otherwise we fall back to LPDDR5X"

You can't make controllers that do both 5X and 6 without wasting a LOT of area, because 5X controllers are 16 bit and 6 controllers are 24 bit.

I just don't buy that Apple is suddenly to be all aggressive in pursuing new memory types immediately after they are available when they have never done that in the past. A similar power benefit has always been there with every iteration from 4 -> 4X -> 5 -> 5X, but they've decided the cost penalty was not worth it.

 

[Mopetar]

But why hold off the Pro chips? N3P is a mature node from the day it was born. They might need a delay for larger dies (esp. Max) with N2 right out of the gate, but not N3P.

It wouldn't make sense for Apple to introduce a cheaper Mac and hold off the more profitable higher end Macs in its favor. Look at how they do it with iPhone, the SE and now 'e' versions follow the real stuff six months later to give the more profitable ones time on the market before throwing out the lower profit little brother.

Not saying they won't release an A series Macbook this year and M5 Macs next, but if they do it won't be because they didn't want to give the cheap one time alone on the market to "breathe". And if they do I'd guess we'll see it use A18P not A19P.

Like I said, I don't think they would do something like that, just that it was a hypothetical possibility based on possible release dates. To speculate further if they were wafer limited on N3P for whatever reason and needed to spread the manufacturing out, but I doubt that's the case.

Apple just does whatever they want at the end of the day for reasons that sometimes don't make the most sense from an outsider perspective, but no one will complain if M5 Macs don't come until early next year. Or maybe more specifically no one will care about anyone who does complain.

I'm still more than fine with M1 MBP so I don't care when M5 ships. I just wish they'd make a good iMac again so I can upgrade that. Unfortunately that seems less likely than the other hypothetical scenarios I had proposed.

 

[Eug]

I'm still more than fine with M1 MBP so I don't care when M5 ships.

I'm amazed that my new iPhone is faster than M1, and even M2 in some measures.

 

[mvprod123]

It is truly impressive how well Apple's new E-cores perform.

 

[DZero]

The e cores are like Pentium M of this generation, for servers could be extremely interesting for example.

 

[name99]

You can't make controllers that do both 5X and 6 without wasting a LOT of area, because 5X controllers are 16 bit and 6 controllers are 24 bit.

I just don't buy that Apple is suddenly to be all aggressive in pursuing new memory types immediately after they are available when they have never done that in the past. A similar power benefit has always been there with every iteration from 4 -> 4X -> 5 -> 5X, but they've decided the cost penalty was not worth it.

That's a reasonable point.
I'm not wedded to the memory issue, my point is more that schedule can be driven by any number of things.
Sometimes it's not the most obvious element (ie the SoC) that's determining the schedule.
And sometimes even when it is the SoC, the element that's changed the schedule is non-obvious (that was my point that M3 Max has a different E core from M3 earlier). It's unlikely, but hardly impossible IMHO, for example that Apple had a new ANE planned, it couldn't quite make the A19 generation so they used the previous model, but they will delay the M5's to pick up this new ANE.

 

[Doug S]

That's a reasonable point.
I'm not wedded to the memory issue, my point is more that schedule can be driven by any number of things.
Sometimes it's not the most obvious element (ie the SoC) that's determining the schedule.
And sometimes even when it is the SoC, the element that's changed the schedule is non-obvious (that was my point that M3 Max has a different E core from M3 earlier). It's unlikely, but hardly impossible IMHO, for example that Apple had a new ANE planned, it couldn't quite make the A19 generation so they used the previous model, but they will delay the M5's to pick up this new ANE.

Oh I fully agree with you that Apple has clearly shown they have split the development of the various SoC blocks out from the release cycles of the ever growing number of SoCs that use them. So if for example there was a massively improved P core that didn't make the cut for A19, that could a reason for the M5 delay, because they thought it was worth the wait to make it part of M5. And yeah given the AI hype I suppose I could see them delaying M5 for an updated ANE, though it would have to be a pretty major improvement, enough that just taking the current one from 16 to 24 or 32 cores wasn't "good enough" if they felt it needed a boost. But I admit I don't know what sort of AI features they might be planning for M5 Macs that could potentially depend on features in a new ANE that more cores couldn't address.

That sort of thing would be an example of schedule risk Apple is willing to take on, because it is fully under their control. They have a fixed release timing for iPhone SoCs, so it gets what it gets depending on when things are ready. The release cycle for the Mac is more flexible than iPhone's, and they would know exactly how "close" the new P core or new ANE was to being ready at the time they have to make the final decision - which I suppose is when they're doing floorplanning prior to tapeout.

So if they knew "this means we can't release on October but instead have to release in January" they would have the information they needed to make the call. I don't think they'd do that with risk they don't control, because instead of being a new P core that is a few months out from being ready if it was LPDDR6 the decision whether to include those controllers or LPDDR5 controllers would have to be more OVER A YEAR prior to the initial mass production. This would probably also assume Apple is designing its own LPDDR6 controller. I don't know if they design their own or license the IP. If its the latter, I have to wonder if there would have been LPDDR6 controller IP available at that time.

I'll just add a note here that it'll be hilarious if we see an article rumoring that Apple has a massively improved P core coming in M5 because an AI scraped this post, and a "journalist" relied on information from that AI to write it!

 

[Eug]

I'll just add a note here that it'll be hilarious if we see an article rumoring that Apple has a massively improved P core coming in M5 because an AI scraped this post, and a "journalist" relied on information from that AI to write it!

Off topic but there was an English language article that made the rounds last week stating that the writer of the Soda Pop song from KPop Demon Hunters used AI to write that song. However, the original Korean article didn't make the same claim. It turns out the Korean news outlet had ironically used AI for the English version of the article. They ended up publishing a correction to bring the English article in line with the original Korean one.

Anyhow, I have turned off AI on my Apple devices (aside from using Apple's Photo Cleanup tool, which uses AI but does not require the full AI to be turned on). Generative AI is still quite problematic in many instances. Although Apple uses heavy photo processing algorithms with its camera, I'm happy to know that they minimize the use of generative AI in that processing, whereas companies like Samsung push hard into the generative AI aspect of computational photography.

 

[mvprod123]

A19 Pro die shot

 

[Eug]

 

[Doug S]

View attachment 131064

It is pretty clear the Air is a tech demo for the folding phone. It has some compromises in the form of battery life due to the thin form factor, and in order to maximize what space there is for battery they had to leave out the vapor chamber the Pro line got which means more throttling and worse performance under load.

I can't imagine they'll sell the most expensive iPhone ever but it can't keep up with a model selling for half as much, so it seems a lock it will include a vapor chamber. Since it will have the whole second half available for more battery, it can sacrifice a bit on the half that has the SoC to make room for that, solving both the battery life and the throttling issue while not being not much thicker/heavier than a Pro Max in folded form.

That says nothing about how well the whole folding thing performs, but at least there won't be any reason to feel you're compromising on factors not related to folding if you buy one.

 

[Eug]

I got my Cosmic Orange 17 Pro Max today. That vapour chamber works wonders. Even with installing all the new apps for the initial setup and charging at the same time over MagSafe, the phone only got mildly warm.

Form factor is definitely chonky though. Noticeably thicker than my 12 Pro Max, although feels decent in-hand since the width is the same. Camera is a HUGE upgrade. Like totally different league.

 

[johnsonwax]

It is pretty clear the Air is a tech demo for the folding phone. It has some compromises in the form of battery life due to the thin form factor, and in order to maximize what space there is for battery they had to leave out the vapor chamber the Pro line got which means more throttling and worse performance under load.

I can't imagine they'll sell the most expensive iPhone ever but it can't keep up with a model selling for half as much, so it seems a lock it will include a vapor chamber. Since it will have the whole second half available for more battery, it can sacrifice a bit on the half that has the SoC to make room for that, solving both the battery life and the throttling issue while not being not much thicker/heavier than a Pro Max in folded form.

That says nothing about how well the whole folding thing performs, but at least there won't be any reason to feel you're compromising on factors not related to folding if you buy one.

I still contend they were shooting for a folding phone this year but couldn't get the screens nailed down and punted with the Air. Screens have been a persistent problem for Apple. It's also unclear what volume Apple would need for these as folding phones now don't sell in huge numbers and Apple doesn't sell any phones in small numbers. We'll see what the Air produces. Even the derided SE vastly outsold in its first year all folding phones combined last year. I'm not sure I see Apple keeping both the Air and a folding phone in the lineup at the same time.

 

[DZero]

I got my Cosmic Orange 17 Pro Max today. That vapour chamber works wonders. Even with installing all the new apps for the initial setup and charging at the same time over MagSafe, the phone only got mildly warm.

Form factor is definitely chonky though. Noticeably thicker than my 12 Pro Max, although feels decent in-hand since the width is the same. Camera is a HUGE upgrade. Like totally different league.

Camera feels like some low tier professional cameras. Of course is the bottom of the barrel of those products, but heck, is a massive improvement compared to previous years.

 

[Eug]

Camera feels like some low tier professional cameras. Of course is the bottom of the barrel of those products, but heck, is a massive improvement compared to previous years.

Yeah, for the others here that may not know: Several features were specifically targeted at professional filmmakers / videographers, like Genlock synchronization, open gate ProRes RAW, Apple Log 2. For stuff like Genlock, you can't even use the feature unless you already own third party professional gear that supports it. Same with ProRes RAW: To use this feature, an external high speed SSD is required, considering that it can use up over 1 TB storage per hour. Furthermore, one of the complaints for older models was overheating in professional use, esp. outdoors, and the combination of the vapour chamber and the aluminum unibody shell will help a lot with that.

Meanwhile, I will continue to use my iPhone to surf Anandtech.