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

 

[DrMrLordX]

Are we sure that it can even play that trombone?

It can, but it's laggy.

 

[mvprod123]

https://twitter.com/x/status/1966036127514767820

 

[mvprod123]

https://twitter.com/x/status/1966036130572415300

 

[ashFTW]

Like 10 years ago with iPhone X when people were wondering what comes next I was saying Apple should just name it for the year. At least they abandoned the 's' stuff, and 10 years later did the year based naming but for iOS rather than the iPhone. Pretty sure the codenames and SoC names don't matter to them because the typical customer doesn't see that stuff. It is just annoying for us lol

I am kind of annoyed by the iPhone Air naming too tbh. If they update it for next year will it be like iPhone SE and iPad Pro where people variously identify them as "second generation" or "2026 model" or whatever? That's just unnecessarily confusing. I can't help wondering if that naming means that maybe iPhone Air won't be updated yearly, and maybe won't even be updated on the same schedule as the other iPhones. It seems like a decent option, I don't really make use of the fancy camera on my Pro Max so I could sacrifice that (and one GPU core) in exchange for a lighter phone. Kinda wonder how it does as far as hot spots given the thin case, titanium rather than aluminum, and no vapor chamber.

Yeah the naming of Apple products has gotten very inconsistent. iPhone Air was very very (maybe I should say pro max ultra) tempting, but in the end it came down to the camera for me; Air’s battery was fine for my use case. I’m going with the Pro Max. Don’t think I’ll get the fold next year for the similar camera compromises it will end up making.

 

[Eug]

https://twitter.com/x/status/1966036127514767820

Plus 5% increase in clock speed = 11-12% increase in single-core performance.

 

[mikegg]

And TSMC would be even more profitable if they no longer had to build fabs. But that's now how the economy works.

Great. So AI does have a model to follow - similar to TSMC's industry. So why do you say there's no business model?

 

[Raqia]

https://twitter.com/x/status/1966036127514767820

Interesting given that the A18 Pro's big cores have edit:MORE cache than the A19 Pro's (per GB6... error?). Looking forward to a Geekerwan deep dive...

 

[name99]

A18 Pro (+15% faster CPU and +20% faster GPU than A17 Pro). Only 5% better in CPU? Pretty meh. Probably most of the transistor budget was spent on the GPU. (+30% faster than A18 Pro)

As expected.
It's become clear for a while now that CPU and GPU are on alternating schedules, with big changes in each taking two years.
This is not to say the CPU hasn't changed; there are a bunch of security changes like the MTE stuff. It will be interesting to see whether SSVE remains crippled and useless or has actually been made usable.

But as far as performance changes go, it looks like what's been done is mainly laying the groundwork for next year.
My guess is the new front-end is in place (eg trace cache and various tweaks to exploit "almost certain" branches) but that by itself doesn't much boost performance over what we already have (especially not with GB6 or SPEC, you want to test a LARGE footprint workload; maybe visible in the browser tests?) But next year they can spend transistors within the execution core (maybe 6 NEON pipelines? or another load pipeline?) and start to take real advantage of the new front-end. The other thing I'd love to see them do at some point is take advantage of criticality. Almost every other vendor has started to copy and take seriously Apple innovations, but criticality exploitation is the one big area where no-one has really yet turned academic R&D into a product.

 

[name99]

"We conducted a deep evaluation and research process to determine whether MTE, as designed, would meet our goals for hardware-assisted memory safety. Our analysis found that, when employed as a real-time defensive measure, the original Arm MTE release exhibited weaknesses that were unacceptable to us, and we worked with Arm to address these shortcomings in the new Enhanced Memory Tagging Extension (EMTE) specification, released in 2022. More importantly, our analysis showed that while EMTE had great potential as specified, a rigorous implementation with deep hardware and operating system support could be a breakthrough that produces an extraordinary new security mechanism."

Apple is once again making it clear to everyone that they are perhaps the biggest contributor to changes in ARM specifications.

And yet
"Still no SVE2?"

Put those two together...
They seem to think SVE is a bad idea? We can at least see how they treat SSVE in this new design to get some hint of their direction.

 

[name99]

It's a little out of character for Apple to do this unless they want to do more in the radio space. They've sort of had that ambition with U1, but it's not like Broadcom isn't perfectly competent to provide these kinds of chips. Why make effectively a commodity component unless you intend to push it out of the commodity space? Cellular was a little different as they were trying to escape a monopoly situation and there was similar opportunity for power savings as they found with AS. I can't imagine N1 power draw is high enough to justify it.

If you control the radios you can do unusual things that your vendor may not be willing to do.
For example I suspect it's not too hard, technically, to hotspot WiFi the same way you hotspot cellular, but it likely requires small tweaks to the chips that Broadcom does not provide.

Why might this be useful? Any situation where you have to pay for each WiFi connection, maybe, perhaps on a plane? (Dad gets one connection and the rest of the family hotspots to his iPhone or iPad...)
Or perhaps some sort of future "shareable" CarPlay where my iPhone is connected to the car (wirelessly) but other people can AirPlay their choice of music to my phone and thus to the car audio system?

 

[jpiniero]

For example I suspect it's not too hard, technically, to hotspot WiFi the same way you hotspot cellular, but it likely requires small tweaks to the chips that Broadcom does not provide.

I am not sure if iOS (or Android) would let you do it... but you should be able to do it on Windows or Linux provided the WiFi has at least 2 antennas.

 

[name99]

8% better than A18 Pro. I wonder how much performance got nerfed because of MIE implementation.

It's honestly not bad considering how it stayed on N3 family, got MIE, and added Neural Accelerators to the GPU which probably took up most of the transistor budget. Matmul acceleration in GPUs usually take up an additional 10% die space.

I think M6 generation is going to be the mother of all upgrades. I've been holding out from upgrading. My M1 Pro 16" MBP is still going strong along with my M4 Mini. My next laptop is definitely going to be M6 Max w/ maximum memory (192/256GB?) for local LLMs. Hopefully top-end LPDDR6 for 917 GB/S bandwidth. N2 node. Tandem OLED. Supposedly thinner and lighter than the current generation of MBPs. Now that'd be an excellent upgrade for M1 holdouts like me.

Annual reminder that these first leaks of GB6 results tend to be 1 to 2% slower than the results in a few months. Generally, I assume, because the compiler hasn't yet been fully tweaked to best match the CPU.

Eg the equivalent of these next day GB6 results for the A18 Pro were 3409, whereas the GB6 average today (across many many submissions) is 3446, so ~1% higher.

 

[name99]

https://twitter.com/x/status/1965885239315882480

The other thing to look for is the MOPS instructions. Conceptually these are somewhat like Intel's REP instructions in that they speed up copying data or setting data (memcpy and memset) but done in a more RISC-like fashion. These could add a % or 2 to a lot of code given how frequent these operations are BUT we won't see their effect in benchmarks until both the compilers and then the benchmarks are updated.

Reason to believe they might be present is that they are present in the new ARM Lumex cores (but ARM appears to have fscked something up in the implementation, so they're not actually that performant this generation *sad trombone*) Hopefully Apple did a better job...

 

[name99]

I am kind of annoyed by the iPhone Air naming too tbh. If they update it for next year will it be like iPhone SE and iPad Pro where people variously identify them as "second generation" or "2026 model" or whatever? That's just unnecessarily confusing. I can't help wondering if that naming means that maybe iPhone Air won't be updated yearly, and maybe won't even be updated on the same schedule as the other iPhones. It seems like a decent option, I don't really make use of the fancy camera on my Pro Max so I could sacrifice that (and one GPU core) in exchange for a lighter phone. Kinda wonder how it does as far as hot spots given the thin case, titanium rather than aluminum, and no vapor chamber.

I expect the Air naming will be resolved based on how well it sells.

Will it be another Mini, matching a demand that only exists in the mouths of internet whiners?
Does Apple in fact already believe that, and the PRIMARY reason for the existence of the Air is actually to prototype one half of the 20th Anniversary iPhone, the Foldable XX?

 

[Mopetar]

I don't understand the appeal of foldable phones. It might reduce the height of the device, but it makes it thicker and that's less comfortable in a pocket. If it's folded, I have to open it each time before use and then fold it again when done. If I'm not folding it, why did I buy it in the first place? It will definitely wear out faster than a non-foldable phone and I don't want to deal with a device that fails outside of warranty that could have otherwise lasted me another five years.

Maybe it makes sense when phones become so thin that when doubled over they're still thinner than the thinnest devices today, but for now it's a gimmick as far as I'm concerned. I could see the utility for a larger tablet device that could become pocketable, but even that's a bit of a stretch, at least for me personally.

 

[poke01]

Anyone know what node the Broadcom wifi and bt chips were fabricated on in the iPhone 16?

 

[name99]

The other thing to look for is the MOPS instructions. Conceptually these are somewhat like Intel's REP instructions in that they speed up copying data or setting data (memcpy and memset) but done in a more RISC-like fashion. These could add a % or 2 to a lot of code given how frequent these operations are BUT we won't see their effect in benchmarks until both the compilers and then the benchmarks are updated.

Reason to believe they might be present is that they are present in the new ARM Lumex cores (but ARM appears to have fscked something up in the implementation, so they're not actually that performant this generation *sad trombone*) Hopefully Apple did a better job...

Update.
We don't get FEAT_MOPS this generation.
But we DO get FEAT_CSSC which is a collection of instructions to handle common operations that currently take more than one instruction (think eg min, max, abs, and some bit manipulation instructions). Like FEAT_MOPS, these will help a small amount with performance (though not nearly as much as FEAT_MOPS) and will only be visible in benchmarks after compiler updates and then recompiles.

 

[jdubs03]

https://twitter.com/x/status/1966036130572415300

That frequency for the Snapdragon Elite 2 of 4.05 GHz doesn’t comport with the actual geekbench result which states 4.74 GHz. But even that seems off because isn’t the 8E2 supposed to be max out at 4.60 GHz?

Even so 3393 points at 4.60 GHz is a hell a lot different.

Anyone have any ideas on that?

samsung SM-S947U - Geekbench

Benchmark results for a samsung SM-S947U with a Qualcomm ARMv8 processor.

browser.geekbench.com

 

[mvprod123]

That frequency for the Snapdragon Elite 2 of 4.05 GHz doesn’t comport with the actual geekbench result which states 4.74 GHz. But even that seems off because isn’t the 8E2 supposed to be max out at 4.60 GHz?

Even so 3393 points at 4.60 GHz is a hell a lot different.

Anyone have any ideas on that?

View attachment 130070

samsung SM-S947U - Geekbench

Benchmark results for a samsung SM-S947U with a Qualcomm ARMv8 processor.

browser.geekbench.com

The core operated at 4.05 GHz during the test

 

[jdubs03]

The core operated at 4.05 GHz during the test

Ah I kind of figured that. They probably should’ve put in a max frequency value there to clarify there is more headroom.

Well, by just doing a linear approximation to 4.6 GHz yields a 3665 score. And we know there are diminishing marginal returns as you approach the higher power levels.

In terms of IPC that Arm C1-Ultra seems promising. Just a tad lower than the A19s.

 

[Eug]

For example I suspect it's not too hard, technically, to hotspot WiFi the same way you hotspot cellular, but it likely requires small tweaks to the chips that Broadcom does not provide.

Why might this be useful? Any situation where you have to pay for each WiFi connection, maybe, perhaps on a plane? (Dad gets one connection and the rest of the family hotspots to his iPhone or iPad...)

This is niche enough that I'm pretty sure Apple wouldn't bother.

Or perhaps some sort of future "shareable" CarPlay where my iPhone is connected to the car (wirelessly) but other people can AirPlay their choice of music to my phone and thus to the car audio system?

I already do this in CarPlay with Spotify. Spotify on my iOS 26 iPhone 12 Pro Max plays through my car's CarPlay but my daughter's iPhone 16e controls the actual song playback via Spotify Queue and Jam.

Furthermore, my 2025 Toyota Camry supports 2 connected Bluetooth devices so I can use navigation software from my phone while she plays Bluetooth music from hers.

 

[Doug S]

I don't understand the appeal of foldable phones.

I don't either, but not everyone has to want something only enough people. It remains to be seen if that's an "enough people" thing or not.

I was reminded of this by my girlfriend, because I'd seen stuff mentioning the cross body strap over the last couple weeks and was thinking "wtf??" and saw her wearing a little cross body bag yesterday. Mentioned the cross body strap thing to her and she said her pilates class was talking about it and is rather excited. So I'm driving around town on errands today, going to the gym, etc. and I'm kinda looking for this a bit I guess and I'm shocked by how many college girls are wearing bags like that. I honestly never noticed before, that's not something I pay attention to. I guess that's why you need "diversity" in your hiring, a bunch of white men like me would never come up with something like that lol

 

[retnuh]

I don't either, but not everyone has to want something only enough people. It remains to be seen if that's an "enough people" thing or not.

I was reminded of this by my girlfriend, because I'd seen stuff mentioning the cross body strap over the last couple weeks and was thinking "wtf??" and saw her wearing a little cross body bag yesterday. Mentioned the cross body strap thing to her and she said her pilates class was talking about it and is rather excited. So I'm driving around town on errands today, going to the gym, etc. and I'm kinda looking for this a bit I guess and I'm shocked by how many college girls are wearing bags like that. I honestly never noticed before, that's not something I pay attention to. I guess that's why you need "diversity" in your hiring, a bunch of white men like me would never come up with something like that lol

The general lack of pockets or pockets that are tiny compared to men’s clothing is shocking. Think about what you’d need to carry if your hand only fit half way in your front pocket. I used to work in fashion rental and it was the number one thing all women wanted, normal pockets.

Some stats: styles with pockets / total styles
Dresses: 295 / 6018
Everyday: 637 / 3681
Workwear: 550 / 3579

I don’t have any info on whether they’re good pockets but they are considered an embellishment. One of best things of working there was the diversity and gaining insight into things like this, had people from many countries and iirc we hit ~40% female engineers at some point while I was there.

The look of jealous betrayal when I showed a pair of jeans with pockets so deep I could fit half my forearm in, too big honestly, and my jacket pockets that could hold a water bottle vertical no problem I will never forget.

Foldable phone wise if the screen doesn’t have a crease when open I’d consider it.

 

[mikegg]

I don't understand the appeal of foldable phones. It might reduce the height of the device, but it makes it thicker and that's less comfortable in a pocket. If it's folded, I have to open it each time before use and then fold it again when done. If I'm not folding it, why did I buy it in the first place? It will definitely wear out faster than a non-foldable phone and I don't want to deal with a device that fails outside of warranty that could have otherwise lasted me another five years.

Maybe it makes sense when phones become so thin that when doubled over they're still thinner than the thinnest devices today, but for now it's a gimmick as far as I'm concerned. I could see the utility for a larger tablet device that could become pocketable, but even that's a bit of a stretch, at least for me personally.

No one here understands foldable phones because everyone here has at least 2 computers (phone + laptop/desktop).

For most of the world, their phones are literally the only computer they have. It's the source of their entertainment, work, social, study. When it's the only computer you have, you go for the biggest screen possible. Just travel to 3rd/2nd world countries and see how people use their phones. A great majority do not have laptops or desktops or even tablets. Heck, just look at younger Gen Zs in 1st world countries. They don't know how to use a Mac or Windows computer. Everything they do is on iOS or Android.

Everyone here needs to at least try to understand that they are NOT the average consumer. People are old here with a special hobby.

Most normal people will want a foldable if it's cheap and reliable enough.

 

[johnsonwax]

Great. So AI does have a model to follow - similar to TSMC's industry. So why do you say there's no business model?

Except that TSMC does make enough money to pay for their infrastructure. AI doesn't. You can't just say 'if we ignore the infrastructure we're profitable'. Good christ man.