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

 

[FlameTail]

I hope so. Otherwise, Apple will miss the AI generation.

Lunar Lake, Zen 5 will come with 16GB as base.

I have a grim feeling that Apple will do the bare minimum 8 GB -> 12 GB, and not 8 GB -> 16 GB.

Unless, the extra 4 GB of 16 GB is so crucial to AI...

 

[Eug]

I have a grim feeling that Apple will do the bare minimum 8 GB -> 12 GB, and not 8 GB -> 16 GB.

Unless, the extra 4 GB of 16 GB is so crucial to AI...

I’m not convinced Apple will upgrade from 8 GB base this year (since 8 GB still actually works fine for light usage), so if they do increase the base memory to 12 GB, that would be a very pleasant and welcome surprise. That would satisfy the needs of entry level users.

But like I said, I’m not convinced they will do even that, and will just stick with 8 GB for this generation. However, at entry level, few care about AI anyway.

 

[Doug S]

I’m not convinced Apple will upgrade from 8 GB base this year (since 8 GB still actually works fine for light usage), so if they do increase the base memory to 12 GB, that would be a very pleasant and welcome surprise. That would satisfy the needs of entry level users.

But like I said, I’m not convinced they will do even that, and will just stick with 8 GB for this generation. However, at entry level, few care about AI anyway.

Yeah the people who are buying a new PC for the AI hype aren't going to buying a base model, whether they are buying a Mac or a PC.

 

[dr1337]

Yeah the people who are buying a new PC for the AI hype aren't going to buying a base model, whether they are buying a Mac or a PC.

But they spend so much time advertising their NPU and all of their AI support, it is weird to any borderline informed customer that apple holds back, so hard, on RAM. My entry level 4gb 3050 lenovo I sought out to try AI on modern Nvidia GPUs is faster in stable diffusion than any M1 benchmark I've seen. And ya know, it just so happens that laptop came with a total of 12gb of RAM between the GPU and CPU... heh

Some portion of the market definitely buys new personal computers without considering specifications or price. But I think a pretty strong majority of computer buyers aren't interested in purchasing a new machine unless it has new features. If your computer already works and is fast enough why would you spend more money unless you have A LOT of cash to burn?

Apple is interested in keeping their stake of high end markets. They purposely pair low vram and storage as an 'entry model' in order to upsell SKUs.

Considering 8gb was standard for a macbook pro 10 years ago and its still standard now shows a regression on apple towards consumer friendliness. There is absolutely no logical reason whatsoever to sell a laptop marketed as "pro", that only comes with 8gb of total RAM in 2024, unless you're specifically abusing a captive market.

They are literally the 2nd most valuable tech company in the world right now, if anyone has enough money to make 16gb standard for a PC, its Apple. They don't need those margins to retain 2nd place, and yet in this same world companies 10x smaller than them don't have as extreme of a margin on RAM. There is no logical argument for their insane price gouging and heinously anti-consumer behavior. Rather, the current board/CEO are clawing at everything they can in order to maintain stock prices and pass on that detritus to the average consumer who (they hope) is too ignorant.

And then they can come around and announce the new M4 or M5 that comes with 12/16gb base model and get tons of praise and headlines. Its really not that hard to see Apple's strategy moving into the 20s.

 

[Doug S]

But they spend so much time advertising their NPU and all of their AI support, it is weird to any borderline informed customer that apple holds back, so hard, on RAM. My entry level 4gb 3050 lenovo I sought out to try AI on modern Nvidia GPUs is faster in stable diffusion than any M1 benchmark I've seen. And ya know, it just so happens that laptop came with a total of 12gb of RAM between the GPU and CPU... heh

Some portion of the market definitely buys new personal computers without considering specifications or price. But I think a pretty strong majority of computer buyers aren't interested in purchasing a new machine unless it has new features. If your computer already works and is fast enough why would you spend more money unless you have A LOT of cash to burn?

Apple is interested in keeping their stake of high end markets. They purposely pair low vram and storage as an 'entry model' in order to upsell SKUs.

Considering 8gb was standard for a macbook pro 10 years ago and its still standard now shows a regression on apple towards consumer friendliness. There is absolutely no logical reason whatsoever to sell a laptop marketed as "pro", that only comes with 8gb of total RAM in 2024, unless you're specifically abusing a captive market.

They are literally the 2nd most valuable tech company in the world right now, if anyone has enough money to make 16gb standard for a PC, its Apple. They don't need those margins to retain 2nd place, and yet in this same world companies 10x smaller than them don't have as extreme of a margin on RAM. There is no logical argument for their insane price gouging and heinously anti-consumer behavior. Rather, the current board/CEO are clawing at everything they can in order to maintain stock prices and pass on that detritus to the average consumer who (they hope) is too ignorant.

And then they can come around and announce the new M4 or M5 that comes with 12/16gb base model and get tons of praise and headlines. Its really not that hard to see Apple's strategy moving into the 20s.

They will provide enough RAM to make their AI run well. Apple gets a lot of out of their RAM due to aggressive use of page compression and Safari isn't as much of a memory hog as Chrome.

Don't discount the impact of being 64 bit only either. Windows systems have to keep both 32 and 64 libraries resident in memory for most people because they run a mix of 32 and 64 bits. (off topic, but are all parts of Windows 11 64 bit now? I remember for years stuff some of the Administrative tools still had the old NT 4.0 look years after they've moved on to XP and even 7, so I wouldn't be surprised if they're still 32 bit executables)

That said I agree Apple needs to move on from 8 GB. They may run a lot better with 8 GB than a PC does, but that's no excuse given that even their low end Macs are sold as premium products.

 

[okoroezenwa]

Yeah, nothing $1k and up should be hobbled with low RAM/SSD amounts.

 

[poke01]

But they spend so much time advertising their NPU and all of their AI support, it is weird to any borderline informed customer that apple holds back, so hard, on RAM. My entry level 4gb 3050 lenovo I sought out to try AI on modern Nvidia GPUs is faster in stable diffusion than any M1 benchmark I've seen. And ya know, it just so happens that laptop came with a total of 12gb of RAM between the GPU and CPU... heh

Some portion of the market definitely buys new personal computers without considering specifications or price. But I think a pretty strong majority of computer buyers aren't interested in purchasing a new machine unless it has new features. If your computer already works and is fast enough why would you spend more money unless you have A LOT of cash to burn?

Apple is interested in keeping their stake of high end markets. They purposely pair low vram and storage as an 'entry model' in order to upsell SKUs.

Considering 8gb was standard for a macbook pro 10 years ago and its still standard now shows a regression on apple towards consumer friendliness. There is absolutely no logical reason whatsoever to sell a laptop marketed as "pro", that only comes with 8gb of total RAM in 2024, unless you're specifically abusing a captive market.

They are literally the 2nd most valuable tech company in the world right now, if anyone has enough money to make 16gb standard for a PC, its Apple. They don't need those margins to retain 2nd place, and yet in this same world companies 10x smaller than them don't have as extreme of a margin on RAM. There is no logical argument for their insane price gouging and heinously anti-consumer behavior. Rather, the current board/CEO are clawing at everything they can in order to maintain stock prices and pass on that detritus to the average consumer who (they hope) is too ignorant.

And then they can come around and announce the new M4 or M5 that comes with 12/16gb base model and get tons of praise and headlines. Its really not that hard to see Apple's strategy moving into the 20s.

Notice how they included the 16GB as a standard config instead of built to order for both the M3 Air and M3 Pros in March. 100% sure that was because of the bad press they got last year due to 8GB drama. Now third party stores will have discounts on the 16GB models well.

Thats why reviewers have to keep on pressuring Apple about this. Confront them.

 

[FlameTail]

Notice how they included the 16GB as a standard config instead of built to order for both the M3 Air and M3 Pros in March. 100% sure that was because of the bad press they got last year due to 8GB drama. Now third party stores will have discounts on the 16GB models well.

Very good. Very good.
Hopefully, the next step is the complete elimination of the 8 GB model.

But then, this news came out recently;

Apple continues to insist that 8GB of RAM in Macs is enough for most tasks

In a new interview, Apple's executives have continued to emphasize that 8GB of RAM is enough in a Mac when it comes to handling most daily tasks.

www.xda-developers.com

So I am not sure...

 

[FlameTail]

People are missing the most interesting detail here, the top memory config of 512 GB.

3) The step up from 192 GB to 512 GB means either they are using denser packages (so the max memory possible for all models goes up)

That 512 GB max memory is going to be a hit. It will be amazing to run AI models.

Right now if you want to run AI models, you can buy a consumer GPU like the RTX 4090 - but you are limited by the 24 GB vRAM. If you want a card with more vRAM, you'll have to shell thousands or tens-of-thousands of dollars to get a card like the A100 or H100.

Because of this, some people are running AI models using their Macs (M Max/M Ultra chip configured with high capacity of RAM). Because of the Unified Memory architecture, the NPU and GPU can also make use of all that RAM.

These people will gladly welcome the beefier Neural Engine and increased RAM capacity of the M4.

 

[FlameTail]

Is this graph for ST or MT performance?

Looking at that 15W power consumption for M3...

 

[Eug]

Gurman's M4 release date predictions already seem to be a bit fluid.

Now he is saying that M4 will come around the end of 2024, but only in the 14" MacBook Pro and 24" iMac, and NOT the MacBook Airs.

For M4 Pro/Max MacBook Pros, he states they will come end of 2024/early 2025.

He also suggests M3 Ultra has been shelved. Next Mac Studio could be M4 based in mid 2025, and next Mac Pro could be M4 Ultra in the second half of 2025.

 

[Mopetar]

Clearly the M4 will come out exactly 15 months after the M3, allowing both those claiming a 12 month or an 18 month cadence to claim that they were right and that Apple was slightly late/early for reasons and ensure the pointless bickering will continue for at least one more product cycle.

 

[SpudLobby]

View attachment 96964
Is this graph for ST or MT performance?

Looking at that 15W power consumption for M3...

Obviously MT, they are just measuring the CPU/SOC power draw alone, not including the DRAM and stuff. And in that sense it’s accurate.

 

[FlameTail]

Obviously MT, they are just measuring the CPU/SOC power draw alone, not including the DRAM and stuff. And in that sense it’s accurate.

I am not sure about that. Notebookcheck has measured that the M3 CPU consumes a max of 22W in all core workloads. I believe this was measured using PowerMetrics, which directly gives CPU power (albeit at less accuracy).

 

[Curious_Inquirer]

There’s a pretty interesting post on SemiWiki on how Apple might use Advanced Packaging on the M4. https://semiwiki.com/forum/index.ph...which-might-be-used-on-apple’s-m4-chip.20034/

“
TSMC has reportedly secured four major clients for its latest SoIC packaging – AMD, Nvidia, Broadcom, and Apple. The chip manufacturer is actively working on increasing its next-generation chip packaging (CoWoS) production capacity.

Apple is testing the next-generation packaging solution​

Apple will reportedly use SoIC with Hybrid molding (thermoplastic carbon fiber board composite molding technology). The company is currently conducting a small-scale trial production and will reportedly start mass production in 2025. The company is expected to use this next-generation packaging method in its next generation of AI-powered Apple Silicon or the M4 chip.

TSMC’s SoIC is the industry’s first high-density 3D chip stacking technology which enables heterogenous integration of chips with different sizes using Chip-on-Wafer packaging. This innovative packaging method was first introduced in 2018. Reportedly, the planned advanced packaging facility in Chiayi, Taiwan will not only include two CoWoS (Chip on Wafer on Substrate) plants but also an SoIC facility.

Notably, AMD is TSMC’s first customer to adopt SoIC technology. The company used this SoIC tech along with CoWoS on the Instinct MI300 AI accelerator chips, designed for data centers.

As per Mark Gurman, Apple has officially commenced the development of the M4 chip, which will likely debut with the next-generation MacBook Pro. As TrendForce noted, there’s a possibility that Apple will switch to a 2nm process node for the M4 chip.

Regarding the release of the next Apple Silicon, we can see nearly a one and a half year gap between each of the previous iterations. The Apple M1 was released back in November 2020, the M2 was released in June 2022, and the M3 chip was released towards the end of October (a bit earlier) last year. So we can expect Apple to release the M4 chip in the first half of next year.

https://www.gizmochina.com/2024/04/12/apple-testing-tsmc-cowos-chip-packaging-for-m4-chip/“

 

[FlameTail]

As TrendForce noted, there’s a possibility that Apple will switch to a 2nm process node for the M4 chip.

M4 is 3nm. (N3E)
Only with M5, there is a possibility of using 2nm, if not even that will be 3nm. (N3P)

Regarding the release of the next Apple Silicon, we can see nearly a one and a half year gap between each of the previous iterations. The Apple M1 was released back in November 2020, the M2 was released in June 2022, and the M3 chip was released towards the end of October (a bit earlier) last year. So we can expect Apple to release the M4 chip in the first half of next year.

That contradicts Gurman's report which said M4 coming Q4 of this year.

 

[Curious_Inquirer]

M4 is 3nm. (N3E)
Only with M5, there is a possibility of using 2nm, if not even that will be 3nm. (N3P)

That contradicts Gurman's report which said M4 coming Q4 of this year.

Honestly I’m leaning more to a January-February release. They just launched the M3 Air a month ago. They’ll probably want at least 9 months gap to get some sales out of the Airs.

 

[Doug S]

Honestly I’m leaning more to a January-February release. They just launched the M3 Air a month ago. They’ll probably want at least 9 months gap to get some sales out of the Airs.

Yes and they released the first M3 stuff in October. So what stops them from releasing the first M4 stuff this October and waiting until March for the M4 Air?

 

[SpudLobby]

M4 is 3nm. (N3E)
Only with M5, there is a possibility of using 2nm, if not even that will be 3nm. (N3P)

That contradicts Gurman's report which said M4 coming Q4 of this year.

If M5 is 2025 Q3/4 a year after M4 it’ll be N3P at this point I think. If it’s 2026 H2, then N2

 

[Doug S]

If M5 is 2025 Q3/4 a year after M4 it’ll be N3P at this point I think. If it’s 2026 H2, then N2

Yeah I expect the first Apple Silicon made on N2 will be M6, arriving fall 2026 shortly after A20.

 

[SpudLobby]

Yeah I expect the first Apple Silicon made on N2 will be M6, arriving fall 2026 shortly after A20.

Yep if they stick with annual cadences from here on out this is exactly what I’d expect too.

A somewhat interesting question at this point to me is: by the time we get to N2 and the M6, what do the CPU clocks look like on both the E and P cores?

 

[FlameTail]

A somewhat interesting question at this point to me is: by the time we get to N2 and the M6, what do the CPU clocks look like on both the E and P cores?

Unless Apple gets their marbles together and makes a comeback with their CPU team (it will most likely happen in the M4 or M5 generation), we are talking 5+ GHz.

 

[SpudLobby]

Unless Apple gets their marbles together and makes a comeback with their CPU team (it will most likely happen in the M4 or M5 generation), we are talking 5+ GHz.

Lol, so you’ve come around to my opinion on Apple and IPC increases

 

[FlameTail]

Lol, so you’ve come around to my opinion on Apple and IPC increases

I still maintain my stance that Apple could make a comeback in the next two years and bring a major IPC increase to their P cores, whereas you seem to think otherwise.

 

[SpudLobby]

I still maintain my stance that Apple could make a comeback in the next two years and bring a major IPC increase to their P cores, whereas you seem to think otherwise.

No, I didn’t say it was impossible, I think it’s unlikely based on the notable recent trends, talent loss.

But if it does happen — a comeback of sorts to regain some lead by going even bigger etc, then I expect it should come the next few years especially now that they’re losing the same size of perf/GHz lead they had (even iso-node) on others.