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

 

[FlameTail]

the iPad is a tablet, running a table os,

Ah, so the iPhone must be running a chair os.

 

[poke01]

Ah, so the iPhone must be running a chair os.

I just realised what I wrote. table os

 

[poke01]

Anyway…

I’m expecting big increases in battery life for these iPhones 16 pros. They finally mentioned better battery life in the keynote and the last time they said was for the 13 pros.

The iPhone 13 Pro Max had better battery life than the iPhone 14PM and even 15PM was mediocre.

 

[okoroezenwa]

The iPhone 13 Pro Max had better battery life than the iPhone 14PM and even 15PM was mediocre.

I’m aware of the 14PM but how was the 15PM mediocre? IIRC it had better battery than the 13PM, though not by much.

 

[DZero]

Anyway…

I’m expecting big increases in battery life for these iPhones 16 pros. They finally mentioned better battery life in the keynote and the last time they said was for the 13 pros.

The iPhone 13 Pro Max had better battery life than the iPhone 14PM and even 15PM was mediocre.

How about the 12 Pro Max?

 

[The Hardcard]

CDMA 3G is dead, no reason to be limited to Qualcomm modems in the US anymore or using them only for AT&T/T-mobile like Apple did with their Intel modems.

I got blurry with the situation. All 5G uses Qualcomm tech. Samsung and Mediatek have to pay Qualcomm and do so.

Apple was trying to get around at least some of Qualcomm patents. There are a couple of Qualcomm patents that they went to court to have them removed, but they lost. Some articles claim they would have still had to pay Qualcomm no matter what to have a working modem. Maybe they are trying to minimize the amount.

 

[The Hardcard]

This goes to show that TOPS is a meaningless metric. Benchmarks confirm this as well you need good APIs to advantage.

TOPS Is a meaningful metric, but as Jensen says, in his presentations doing the software work is the much larger task, Hoover building the hardware. Nvidia, which is so much further along than everyone else continues to release software frameworks that massively boost the capabilities of their chips. just last month, they released new software that boosted the Llama 3.1 throughput of the H100 by 30 percent.

If Nvidia can make those huge boost from software optimizations In late 2024, that probably means that none of these chips are anywhere near fully optimized yet. just like Apple getting a 25% boost with the upcoming iOS/macOS releases. there are probably several more big boosts from optimizations to come for every player.

But you must have the TOPS to optimize the TOPS. It does mean that it is impossible to do straight hardware comparisons. AI benchmarks are always going to be hardware and software framework mixtures.

 

[poke01]

How about the 12 Pro Max?

The 11 Pro Max was better at battery life and Apple also mentioned very good battery life for the 11 pro as well.

 

[Doug S]

I got blurry with the situation. All 5G uses Qualcomm tech. Samsung and Mediatek have to pay Qualcomm and do so.

Apple was trying to get around at least some of Qualcomm patents. There are a couple of Qualcomm patents that they went to court to have them removed, but they lost. Some articles claim they would have still had to pay Qualcomm no matter what to have a working modem. Maybe they are trying to minimize the amount.

There are a ton of LTE and 5G patents owned by many companies - including some owned by Apple. They're all covered by FRAND so they are generally not that expensive (though Qualcomm controversially charges based on the value of the device the chip goes into, which many believe is against the spirit of FRAND) Qualcomm has more cellular patents than about anyone (I think Huawei may hold more 5G patents) so they are paid in a lot more than they have to pay out - but they do have to pay Apple for the LTE/5G patents Apple owns - its just that Qualcomm owns a lot more so Qualcomm is obviously the winner in that.

There's no point in trying to work around patents that are part of the 5G standard - they are FRAND because they are part of the standard that must be implemented. The patents Apple tries to work around / invalidate are ones that Qualcomm claims on smartphones themselves or otherwise incidental to cellular standards. That's where Qualcomm really gets you when you buy chips from them, because they force you to license those patents in order to get them to sell you their modems. When Apple is able to get their own modems going they'll free themselves from that. Because if Qualcomm tried to enforce their smartphone related patents against Apple absent the chip supply contract, Apple has a crapton of CPU related patents they would be able to enforce on Qualcomm. So basically that's not a battle Qualcomm would be willing to fight.

 

[Mopetar]

They can only do want they can. The issue is patents. When Qualcomm started working on CDMA, most of the rest of the industry thought it was physically impossible, or at least not worth the effort. Qualcomm scored big time on that. it is almost impossible to have a modem targeting the US market that is both able to avoid Qualcomm patents and not be inferior in capability and/or power consumption.

It doesn't matter since Qualcomm (or anyone with patents that are part of the standard used) has to license them under FRAND (fair reasonable and non-discriminatory) terms. Apple also bought a bunch a patents years ago so every company is paying them for the use of any that are part of any standard where they're necessary and Apple has to license those patents.

Qualcomm may have other patents which aren't part of any standard, but Apple wouldn't need those in order to build a functional cellular modem. If it seems like it's hard for other companies to match what Qualcomm can do it's because that's Qualcomm's core business that they've been involved in for nearly four decades.

 

[Doug S]

Qualcomm may have other patents which aren't part of any standard, but Apple wouldn't need those in order to build a functional cellular modem. If it seems like it's hard for other companies to match what Qualcomm can do it's because that's Qualcomm's core business that they've been involved in for nearly four decades.

Qualcomm being in phones all over the world since the beginning is their biggest advantage. The RF frontend, ADCs, demodulators etc. are all fairly standard stuff (OK RF is a bit of a black art, but there are many capable RF engineers out there) That is to say, the hardware part of the modem is comparatively easy.

The hard (really hard) part is the baseband software. That's where Qualcomm being in phones all over the world is a massive advantage. Like most standards, the LTE & 5G specs that detail interfacing client devices like smartphones with towers doesn't specify every little detail. A lot is "implementation defined", and in many cases that is defined as "how Qualcomm does it" because they've been there since the beginning. Then there is all the stuff related to handoffs between towers, roaming, and so forth where there are even fewer defined standards.

If all Apple wanted to do was make a modem to handle the US they could have been there years ago. There are few true carriers, most of the multiplicity you see like Mint are MVNOs of true carriers like AT&T or Verizon. But if you want a modem that handles the entire world, which is what they need to avoid having different SKUs like Samsung has for ages despite having their own modem, you need to handle all those carriers all over the world. With all their idiosyncracies for tower handoff, for roaming, etc.

The good news is that when you screw up that latter stuff usually the worst you get is a dropped call. That's less of a problem now that everything is data so you aren't circuit switching - Apple can do a modem that does VoLTE/Vo5G only for calls, and handles LTE/5G only. A lot of the ugliest stuff is found in 2G/3G standards but Apple can simply ignore all that and not implement it. But it is still a really really big effort that's going to require testing all over the world. They can't do this stuff in a lab in Cupertino.

 

[SpudLobby]

And there it is. Larger CPU caches than the A18. Could just be SLC though, but it might be L2. It is interesting they didn’t mention these power figures for the regular A18 vs the A17 Pro.

 

[Doug S]

And there it is. Larger CPU caches than the A18. Could just be SLC though, but it might be L2. It is interesting they didn’t mention these power figures for the regular A18 vs the A17 Pro.

They compared the regular A18 with the A16, because that's what the non-Pro iPhones used last year.

 

[The Hardcard]

It doesn't matter since Qualcomm (or anyone with patents that are part of the standard used) has to license them under FRAND (fair reasonable and non-discriminatory) terms. Apple also bought a bunch a patents years ago so every company is paying them for the use of any that are part of any standard where they're necessary and Apple has to license those patents.

Qualcomm may have other patents which aren't part of any standard, but Apple wouldn't need those in order to build a functional cellular modem. If it seems like it's hard for other companies to match what Qualcomm can do it's because that's Qualcomm's core business that they've been involved in for nearly four decades.

I don’t know and can’t speak factually, but I got the impression that the issues wasn’t based on meeting the standards, but how to do operate a modem efficiently. There were claims the at one point Apple had a oversized modem that was half the size of the iPhone logic board and required far more than practical power to run. Nothing about not handling the standards.

There is also the issue of Apple trying to legally force relief from two Qualcomm patents that are in effect until 2029. It would without knowing seem that these are outside of FRAND yet still affecting their modem aspirations. So probably not standards but performance and efficiency.

Though according to some industry analysts the “delay” is due mostly to the fact that it is taking the natural amount of time to produce the modem, but Apple’s early targets were unrealistic.

 

[SpudLobby]

They compared the regular A18 with the A16, because that's what the non-Pro iPhones used last year.

yeah, I am referencing exactly that. They did compare performance, but not power, whereas they did mention power for the Pro. A18 Pro might be more efficient it seems.

That or they left it off and it’s fairly similar and this is just an SLC cache difference.

Edit: yes they did. Nevermind

 

[smalM]

It doesn't matter since Qualcomm (or anyone with patents that are part of the standard used) has to license them under FRAND (fair reasonable and non-discriminatory) terms. Apple also bought a bunch a patents years ago so every company is paying them for the use of any that are part of any standard where they're necessary and Apple has to license those patents.

There are thousands of 5G patents and most of them are non essential.

Share of standard essential patents in 2019

I don't know why so many think Qualcomm is particulary important in 5G.
They owned CDMA but in 5G they are just one of many.

Qualcomm may have other patents which aren't part of any standard, but Apple wouldn't need those in order to build a functional cellular modem. If it seems like it's hard for other companies to match what Qualcomm can do it's because that's Qualcomm's core business that they've been involved in for nearly four decades.

I think this is the main point.

 

[jdubs03]

Searched on the GB scores for the various A18/Pro results. They seem to be quite all over the place which I guess isn’t too unexpected so early on.
Did find this gem though, a 16 Pro:

iPhone17,1 - Geekbench

Benchmark results for an iPhone17,1 with an ARM processor.

browser.geekbench.com

3429 ST, 8790 MT. My guess is this will be near as high as it can go.

 

[Eug]

Enjoying a bit of drama unfolding in a Teams meeting with participants (including directors and CEOs) from all over the country and this one dude who was supposed to do a presentation about his properties company is using a frickin' iPad and he doesn't know how to use it or its limitations while using Teams. So after wasting about 10 minutes for him to get his act together, he is now going to share the presentation to someone with an actual computing device so he can voiceover. Such amazing and "free" publicity for the iPad.

I bet the presentation would've gone better if he had let his 5 year old kid help

I had a similar experience trying to use PowerPoint through Zoom for a presentation on the M4 iPad Pro. While that application otherwise works well on the single screen, once you throw presentation mode with an external screen and Zoom into the mix, things don’t work right.

However, I tested this in advance so I didn’t have to test live on the audience. For the actual presentation, I just pulled out my 2017 m3 MacBook and used that. It presented no problems other than the fact the machine is sluggish even for PowerPoint.

Even in 2024, despite being the fastest computer in my house, my M4 iPad Pro is still limited to being a single user machine with limited multitasking support, and is best limited to a single screen.

Searched on the GB scores for the various A18/Pro results. They seem to be quite all over the place which I guess isn’t too unexpected so early on.
Did find this gem though, a 16 Pro:

iPhone17,1 - Geekbench

Benchmark results for an iPhone17,1 with an ARM processor.

browser.geekbench.com

3429 ST, 8790 MT. My guess is this will be near as high as it can go.

Geez. Freezer test?

That means when I get my iPhone 17 Pro Max next year, it will be faster in multi-core than my work desktop, an M1 Mac mini.

And people wonder why I’m not complaining about my 12 Pro Max’s SoC performance and upgrading now.

 

[jdubs03]

Geez. Freezer test?

That means when I get my iPhone 17 Pro Max next year, it will be faster in multi-core than my work desktop, an M1 Mac mini.

And people wonder why I’m not complaining about my 12 Pro Max’s SoC performance and upgrading now.

I’m thinking yeah. That multicore result is humming.

 

[mvprod123]

Geekerwan interviewed Johny Srouji, who is in charge of Apple Silicon !!!

 

[Nothingness]

I had a similar experience trying to use PowerPoint through Zoom for a presentation on the M4 iPad Pro. While that application otherwise works well on the single screen, once you throw presentation mode with an external screen and Zoom into the mix, things don’t work right.

That's not an iOS-only issue. Teams and Zoom also stink on x86 Linux. I used to have to kill them on my Linux machine even when not used.

They perhaps misbehave on macOS too, but my M1 is so powerful that I wouln't notice.

 

[poke01]

Geekerwan interviewed Johny Srouji, who is in charge of Apple Silicon !!!

Wow, this is a great video

 

[FlameTail]

Geekerwan interviewed Johny Srouji, who is in charge of Apple Silicon !!!

Johnny Srouji said something very interesting about the GPU.

(Paraphrasing) "We use the same architecture in iPhone, iPad and Mac but they have different implementations. It's not just core counts and clock speeds, but more low level details".

I wonder what those low-level differences are.

 

[mavere]

Searched on the GB scores for the various A18/Pro results. They seem to be quite all over the place which I guess isn’t too unexpected so early on.
Did find this gem though, a 16 Pro:

iPhone17,1 - Geekbench

Benchmark results for an iPhone17,1 with an ARM processor.

browser.geekbench.com

3429 ST, 8790 MT. My guess is this will be near as high as it can go.

That maybe freezer test vs an almost-certainly freezer iPhone 15 PM test: https://browser.geekbench.com/v6/cpu/compare/7735922?baseline=7747963

Interesting to note the spread of losses and gains amongst in the ST tests. Wonder what they tweaked this time. A few of the tests have straight up regressions in IPC.

The MT results imply that the efficiency cores did not get SME but still continues the trend of getting faster at a faster pace.

Doing some napkin math while focusing on the Clang subtest in MT (clang/llvm subtests' characteristics supposedly better resemble day-to-day user tasks), seems like E cores are about 40% faster.

 

[johnsonwax]

Apple Intelligence is going to be in beta for the time being. Funny how some users here claimed that Apple has been doing AI for years

Windows Recall is going to be in beta for the time being. Funny how some users here claimed that Microsoft has been making operating systems for years.