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

M5 Family discussion here:

Page 431 - Discussion - Apple Silicon SoC thread

Page 431 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[name99]

If true, that is a massive disappointment. Five months later than the M5.

I don’t see how that doesn’t push back the M6-series redesign until 2027. Unless they want to piss everyone off after just an eight month cycle for the high-end versions.

People mostly still think that the product cycle is that each IP block for say the A19 is designed, then more of those blocks are glued together for the M5, then more still for the M5 Pro and M5 Max. Meaning that when each of M5/Pro/Max are released is basically Apple's choice based on consumer sentiment, fab availablity, and so on.

I think this is no longer as true as it once was. It's more like there are teams working on each IP block, and teams working on A19, M5, Pro, etc, each with a target date (OR feature set) in mind, and each of the chip teams uses whatever the state of an IP block is at the time the design has to be frozen.
So, for example, we have already seen a situation where one model (is the M3?) the Max has an updated version of the E-core relative to the base M and A chip.

Usually I expect the difference is more like the Pro/Max get the 1.04 version of the P-core, but, especially in turbulent times, you could imagine something like the Pro/Max getting the 1.1 version of the M5 ANE, and the 1.2 version of the M5 GPU, in both cases significant improvements that could be rushed into the design having done so?

If you look behind the scenes at the patents, some of the most interesting but least sexy patents have to do with trying to implement uniform standards across all IP blocks - eg a standardized set of control registers of for boot/sleep, or for power control. Obviously the more this can be done, the more it's feasible and low risk to implement the sort of strategy I've suggested here, where IP blocks can be picked up as late in the design process as possible.

 

[fastandfurious6]

unpopular opinion but using latest node / TSMC N3 for phone chips is a total waste

- less than 1% of iphone users use their CPUs more than 50%

- older iphones very popular right now for one main reason: similar to latest but way cheaper. why get latest CPU? just to make UI 4% snappier?

most phones should use 5nm/7nm and leave the best nodes for real machines


I guess it brings free money to Apple & TSMC with no real benefit to consumer

 

[jpiniero]

most phones should use 5nm/7nm and leave the best nodes for real machines

lol

 

[fastandfurious6]

It's a thing: latest nodes only for flagships and their derivatives (cut cores / yields)

rtx6090 - rtx6070: N2
rtx6060: N3/whatever

zen6: N2
zen6 mid-range: N3/whatever

M6: N2
iphone cpu: 5nm/whatever

Why waste cutting edge capacity to build phone chips? What is more demanding on a phone than the game Genshin Impact which runs on potato phones too?

 

[poke01]

It's a thing: latest nodes only for flagships and their derivatives (cut cores / yields)

rtx6090 - rtx6070: N2
rtx6060: N3/whatever

zen6: N2
zen6 mid-range: N3/whatever

M6: N2
iphone cpu: 5nm/whatever

Why waste cutting edge capacity to build phone chips? What is more demanding on a phone than the game Genshin Impact which runs on potato phones too?

Nodes aren’t just for performance but for efficiency as well.

AMD is an irrelevant company compared to Apple. So go back to the Zen6 thread and be happy with a 30% improvement every two years and RDNA3.5.

We can all survive without AMD and Apple. These are just by products of late stage capitalism.

 

[poke01]

@fastandfurious6
Actually let’s talk about how irrelevant AMD is.

Apple makes more profit in a single quarter than AMD makes for the whole 12 months. AMD is peanuts, you should really get out your bubble.
I’m saying all this cause Ryzen and gaming GPUs they are not a big market at all when compared to iPhones.

You have no idea of the scale that Apple operate in terms of iPhone. If anything it’s the Mac chips that should be on N-1, not iPhone.

 

[branch_suggestion]

Why waste cutting edge capacity to build phone chips? What is more demanding on a phone than the game Genshin Impact which runs on potato phones too?

Because they are small dies with very high, consistent volumes that serve as pipecleaners for new nodes?

Nodes aren’t just for performance but for efficiency as well.

AMD is an irrelevant company compared to Apple. So go back to the Zen6 thread and be happy with a 30% improvement every two years and RDNA3.5.

We can all survive without AMD and Apple. These are just by products of late stage capitalism.

Nice doomerism but the trend at TSMC is very clear, the smartphone market is mature with very slow growth whilst HPC is now #1 and growing rapidly.
NV is now the biggest revenue driver at TSMC, AMD is 3rd and catching Apple faster than you'd think, MI400/500, Venice, next gen consoles etc is a lot to digest.
Also you know Mx Pro/Max comp next gen is RDNA5, don't play dumb.

@fastandfurious6
Actually let’s talk about how irrelevant AMD is.

Apple makes more profit in a single quarter than AMD makes for the whole 12 months. AMD is peanuts, you should really get out your bubble.
I’m saying all this cause Ryzen and gaming GPUs they are not a big market at all when compared to iPhones.

You have no idea of the scale that Apple operate in terms of iPhone. If anything it’s the Mac chips that should be on N-1, not iPhone.

I mean Apple is gonna have a nice peak in this memory crunch as they have a massive prepay and iPhone is a premium product which will take share from cheap Android which cannot sell at inflated prices.
But when memory is sufficient again there will be whiplash, of course Apple will remain ahead of the likes of AMD for roughly the next decade but if they cannot establish a DC business...
Apple is a supply chain company, as is NV at this point.

 

[jdubs03]

Nice doomerism but the trend at TSMC is very clear, the smartphone market is mature with very slow growth whilst HPC is now #1 and growing rapidly.
NV is now the biggest revenue driver at TSMC, AMD is 3rd and catching Apple faster than you'd think, MI400/500, Venice, next gen consoles etc is a lot to digest.

I take it you think AI is not a bubble..

 

[fastandfurious6]

We can all survive without AMD and Apple. These are just by products of late stage capitalism.

The spirit of Karl Marx has awaken!

 

[branch_suggestion]

I take it you think AI is not a bubble..

Oh it is, but the trend was established well before the bubble.

 

[mikegg]

I take it you think AI is not a bubble..

Do you think it is a bubble? If so, how inflated is the bubble?

 

[fastandfurious6]

It's not a bubble — it's a paradigm shift...

 

[poke01]

Nice doomerism but the trend at TSMC is very clear, the smartphone market is mature with very slow growth whilst HPC is now #1 and growing rapidly.
NV is now the biggest revenue driver at TSMC, AMD is 3rd and catching Apple faster than you'd think, MI400/500, Venice, next gen consoles etc is a lot to digest.
Also you know Mx Pro/Max comp next gen is RDNA5, don't play dumb.

NV is propped up by AI and has Open AI and MS/Meta made any profit with Blackwell products? No

Very slow growth for phones?
Umm have you seen iPhone 17 sales? Sold 17 million in 4 months just in China alone.


Consoles sell like 100 million every 5 years. Medusa Halo doesn’t even come out of till 2028. Client especially mobile is last of AMDs priority list.

The only where AMD is making actual money is Zen6 server and instinct GPUs that sell to China and other American companies.

But when memory is sufficient again there will be whiplash, of course Apple will remain ahead of the likes of AMD for roughly the next decade but if they cannot establish a DC business...
Apple is a supply chain company, as is NV at this point

DC isn’t the only way to get profit. Apple is proof of that.

Nvidia is not a supply chain company, they can’t even maintain their inventory as good as Apple.

 

[poke01]

Do you think it is a bubble?

AI is a bubble. Apple doesn’t even sell phones with AI in China and people buy them. Even in the US, Apple Intelligence was a failure so far and the 17 series sold millions.

Users do not care about AI, they just want good battery life and cameras and new designs.

 

[fastandfurious6]

View attachment 137761

Can somebody explain this? What is the real reason Chinese people prefer iOS?

Hardware-wise, android/other phones are usually equal or better and always cheaper.

Does the same discrepancy apply in other markets too i.e. iphone selling 6 times more than second in list?

 

[mikegg]

NV is propped up by AI and has Open AI and MS/Meta made any profit with Blackwell products? No

Why do you think they haven't?

 

[poke01]

Why do you think they haven't?

well they buy very expensive hardware and then it’s very expensive to maintain. They sell the AI subscription for peanuts and many users don’t even pay for AI.

 

[mikegg]

well they buy very expensive hardware and then it’s very expensive to maintain. They sell the AI subscription for peanuts and many users don’t even pay for AI.

Meta doesn't sell AI subscriptions.

https://about.fb.com/news/2026/01/2026-ai-drives-performance/

Note: This is without their many billions of investment into their leading edge LLMs called Meta Superintelligence Labs.

 

[mikegg]

well they buy very expensive hardware and then it’s very expensive to maintain. They sell the AI subscription for peanuts and many users don’t even pay for AI.

Are you talking about OpenAI?

 

[poke01]

Are you talking about OpenAI?

Yes. People are falling a scam it’s laughable

Note: This is without their many billions of investment into their leading edge LLMs called Meta Superintelligence Labs

meta is a useless company. Have you seen anyone talk about their Rayban glasses?

 

[mikegg]

Yes. People are falling a scam it’s laughable

At first, you mentioned MS and Meta. I was confused.

Why do you think OpenAI is a scam?

meta is a useless company. Have you seen anyone talk about their Rayban glasses?

Why? They generated $200 billion in revenue in 2025.

I haven't seen their Rayban glasses. They're dialing back on the Metaverse investment. I think they're admitting it was a mistake to spend on VR.

 

[poke01]

They generated $200 billion in revenue in 2025.

advertising is big money.

Why do you think OpenAI is a scam

I’ll ask you a question, what do you think of AI and the space in general?

 

[branch_suggestion]

Very slow growth for phones?
Umm have you seen iPhone 17 sales? Sold 17 million in 4 months just in China alone.

It is a short term burst that isn't sustainable.
Neither is DC spend but for very other reasons.