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

 

[Doug S]

Exactly this. All these "big" announcements are just made so Trump can boast about it to his voters. It's all BS designed to fool the gullible. Apple will invest the minimum it can to make it seem like they're bringing back some manufacturing. They'll deliberately slow plant build outs so that by the time Trump is no longer POTUS, they'll just cancel the plants.

No one is tracking Apple's $600b "investment".

Apple committed to investing $430 billion in the US over the next five years in 2021 shortly after Biden took office. And committed to investing $360 billion over the next five years in 2017 short after Trump took office the first time. Cook knows how to play the game, on both sides.

These commitments never really had anything to do with investing in US production capacity beyond whatever already made sense. They are counting all the salaries from Cook's down to today's new hire at the local Apple Store, along with all the utilities, taxes and money they pay the company that runs the cafeteria in the spaceship. Any components sourced in the US like Micron DRAM are counted. All the datacenters, including green energy investments to power them. It sounds like the custom "AI servers" might be getting built here - that's much easier to do since they don't have the same types of product cycles that consumer products like iPhones and Macbook Airs do and are likely more amenable to automation since there's no customization and no fancy cases. Basically every dollar paid in the US, either directly or indirectly, is being counted.

 

[Eug]

I didn't use sales prices to prove that I was right. Stop putting words in my mouth and twisting actual things I said.

Hmmm…

Amazon, Best Buy, Walmart have been selling the M1 Air for $650 - $750 for over a year now. You keep saying my predictions are ridiculous but you keep getting proved wrong over and over again.

 

[johnsonwax]

This point about warranties is hard to over-emphasize. For laptops used for college / professionally, many of us have mission critical needs and insurance against downtime and repair/replacement cost is very important.

Anecdotally, before I purchased a fully spec’ed M4 Max, I researched comparable x86 options from the top 3-4 vendors. It seemed that no one offered an AppleCare-like warranty beyond three years. The indefinite warranty protection offered through AppleCare+ seems unique to the industry and I find it incredibly valuable. But maybe I missed a program offered by a major vendor somewhere.

In addition, adding phone support so that you have an AppleCare like experience brings the cost of the laptop to about the same as a highly specced M4 Max!

Yeah, when the cost of a 3 year warranty raises the price by ⅓ (which was about what we found - this was some years ago, so maybe things have improved), the actuaries are trying to tell you something important.

 

[Tigerick]

@Tigerick:

Ironically, back in 2020 before the Apple Silicon launch, I had predicted an A-series MacBook the entry-level Apple Silicon laptop, but that never came to fruition back then. Now in 2025, I personally won’t try to claim I was actually right considering my prediction was for 2020 or maybe a little later, not 5-6 years later.

Anyhow, in 2025-2026 I could see an A18 Pro MacBook, as the SoC is more than capable.

Again, not so fast. I have created a thread in MacRumors dedicated for iPad lineup in 2026. And I suspect the so-called low cost Macbook is 13-inch iPad in disguise. Read and hope you get the ideas....

13-inch iPad for $599? iPad 2026 Full Lineup Discussion Threads

I am wondering whether to post new topic discussing iPad 2026 Lineup or not. The news of upcoming low cost MacBook got me excited and confused. To my understanding, Apple won't use A18 Pro with 64-bit memory bus for MacBook; even M1 MacBook was using 128-bit memory bus. Then I look into next...

forums.macrumors.com

 

[mikegg]

Hmmm…

What? I was referring to the fact that Macbook Air prices indicate that Apple can release a much cheaper laptop. I'm surprised you don't understand that reasoning.

I never said Macbook Air on sale is the Macbook SE device from my original prediction.

I'm not sure what's hard for you to understand.

1. I made prediction 5 years ago for a low cost Mac in the $700 range.

2. You said it'll never happen.

3. I said it'll happen because Apple wants market share, sell services, and Apple Silicon allowed it to have margins good enough to lower prices even further.

4. You said it's "ridiculous".

5. I showed you prices that were already in the $700 range when Macbook Air was on sale.

6. Now you're saying #5 is me telling you I was right? What the heck?

Just admit that you were wrong. It's easy. In 2020, you had an outdated mindset about Apple. You still thought they were a premium-only company. In reality, they were already selling $400 iPhone SE, $279 Apple Watch, ~$250 iPads, $100 Homepods. Clearly Apple was always going to make a cheap Macbook. For whatever reason, you just didn't have the foresight to see it. It was plain and obvious back then and it's plain and obvious now.

Even with 5 years of inflation and tariffs, Apple might still be able to deliver a Macbook in the ~$700 range. How wrong are you?

 

[mikegg]

It sounds like the custom "AI servers" might be getting built here - that's much easier to do since they don't have the same types of product cycles that consumer products like iPhones and Macbook Airs do and are likely more amenable to automation since there's no customization and no fancy cases. Basically every dollar paid in the US, either directly or indirectly, is being counted.

I think they'll make AI servers in TSMC Arizona. AI server chips need more mature nodes because they have big dies.

 

[DZero]

Again, not so fast. I have created a thread in MacRumors dedicated for iPad lineup in 2026. And I suspect the so-called low cost Macbook is 13-inch iPad in disguise. Read and hope you get the ideas....

13-inch iPad for $599? iPad 2026 Full Lineup Discussion Threads

I am wondering whether to post new topic discussing iPad 2026 Lineup or not. The news of upcoming low cost MacBook got me excited and confused. To my understanding, Apple won't use A18 Pro with 64-bit memory bus for MacBook; even M1 MacBook was using 128-bit memory bus. Then I look into next...

forums.macrumors.com

Imagine if there are 2 devices instead of one... the iPad with that price and the iMac with the USD 799 tag.

 

[johnsonwax]

I think they'll make AI servers in TSMC Arizona. AI server chips need more mature nodes because they have big dies.

Can they even get capacity in AZ given that AMD/Nvidia have to make their China AI/GPU products there?

 

[Doug S]

I think they'll make AI servers in TSMC Arizona. AI server chips need more mature nodes because they have big dies.

I was talking about assembling the "servers" in the US, not referring to fabbing the chips.

Pretty sure I saw some rumors that Apple was going to be using N2 for their server chips. Just because Nvidia uses megadies doesn't mean Apple would. They could use smaller chips and tie them together Ultra style. If Apple has prepayed for first dibs on N2 they almost have to be using it for AI servers at first. All signs point to N2 mass production starting the last month or two of this year. If so the first finished chips will be going out the door in early Q2 - much too early for iPhone 18. They could do an M6 iPad Pro like last year's M4 surprise, but that doesn't have the volume to use up very many wafers. Making server chips will keep TSMC busy until Apple is ready to start taking deliveries of iPhone chips a few months later.

Heck even a reticle sized die isn't that big of an issue since 'AI server chips' are a highly regular design where you can easily test/disable bad cores. The yield for 100% perfect reticle sized chips in a brand new process might be tiny, but you don't need perfect chips, you only need whatever doesn't have any redundancy designed in to be perfect.

 

[johnsonwax]

5. I showed you prices that were already in the $700 range when Macbook Air was on sale.

That is not a $700 product any more than the promotional shirt I'm currently wearing means that clothing is now free. You will always find products below their indicated price in a market. I'm sure you can find some nice stolen MBAs for less than $700 even.

The question is will Apple release a laptop with a $700-ish list price, and do not see a scenario where that happens. In an inflationary market, holding your base price constant is price-cutting. The $1299 base iMac today is half the price in adjusted dollars as the $1299 iMac when it was first launched in 1998. Apple has made the product drastically more affordable. They don't need to cut prices here in order juice demand - that happens continuously due to inflation. Apple has been doing this for a quarter century so what you are proposing is quite a radical change. What's more, you're proposing that they'll list it for $699-ish post-tariff, which would mean a $550 pre-tariff product. Do you really think Apple is looking to jump from a consumer facing $999 price point to a build cost that would produce a $550 product in one go? Because from Apple's perspective, that's what you're arguing.

But in a tariff environment as high as 25%, they do need to radically cut costs to maintain that $999 or $1299 base price point. A $999 MBA with a 25% tariff is a $799 product pre-tariff which sounds about right for the rumored product. Apple bumps the M5 MBA to $1199, etc. Apple's goal here is to keep the price at pre-tariff levels to the best of their ability (because those are the numbers Apple has judged hit the affordability categories they are chasing). That's been their approach for decades through tariffs, bouts of inflation, downturns in the economy, etc. The price stayed the same the whole time. Apple picked $999. It's not a byproduct of what they made, they chose $999 and built around that price. They will continue to do that. If they want to juice sales, that's what AppleCare and Beats are there for - to value add to that price point. AppleCare has extraordinarily high margins so Apple can use it to value-add. They do the same with Beats. These are their two main promotional items that they entice retailers with at the start of a sales cycle.

At the start of a sales cycle you want to resell a $999 MBA, you are buying it from Apple for $949, and they will discount that further in 1 point margin increments ($10 in this case) if you do other things - onsite repair, selling Beats, other accessories, having a dedicated 'Apple' area that meets their requirements, etc. They will also let you buy AppleCare for basically nothing, and 1:1 Beats (one unit of headphones for every unit of Mac, etc.) for very discounted prices - half off or more. This is how Apple protects their price points, and if you want your retail margins, you gotta flip that AppleCare sale because that's really where your margins are, or flip that Beats sale. It's later in the sales cycle that Apple starts offering discounts to that $949, but usually with a no-return policy. You're stuck with it, and this is why Apple's release schedule being predictable becomes important, because retailers don't want to buy a bunch of M4 MBAs even at a discount a week before the M5s come out.

 

[johnsonwax]

I was talking about assembling the "servers" in the US, not referring to fabbing the chips.

I agree that if possible Apple will seek to assemble the servers in the US, mainly because they don't need the benefits of Chinas logistics to nearly as great of a degree for an internal product, and it's an easy way to get the Trump admin off their back.

But mikegg specifically referenced TSMC AZ, which 1) I don't think Apple will be able to pry away from AMD/Nvidia given it's criticality to those China sales, and 2) TSMC AZ won't remotely have 2nm any time soon so if Apple did make something there it's going to be radios or whatever they're using older nodes for. They could make AI chips on older nodes - Nvidia clearly is - but I don't think Apple can outspend Nvidia for capacity there.

Pretty sure I saw some rumors that Apple was going to be using N2 for their server chips. Just because Nvidia uses megadies doesn't mean Apple would. They could use smaller chips and tie them together Ultra style. If Apple has prepayed for first dibs on N2 they almost have to be using it for AI servers at first. All signs point to N2 mass production starting the last month or two of this year. If so the first finished chips will be going out the door in early Q2 - much too early for iPhone 18. They could do an M6 iPad Pro like last year's M4 surprise, but that doesn't have the volume to use up very many wafers. Making server chips will keep TSMC busy until Apple is ready to start taking deliveries of iPhone chips a few months later.

Heck even a reticle sized die isn't that big of an issue since 'AI server chips' are a highly regular design where you can easily test/disable bad cores. The yield for 100% perfect reticle sized chips in a brand new process might be tiny, but you don't need perfect chips, you only need whatever doesn't have any redundancy designed in to be perfect.

I would also note the profile of what Apple is using their AI servers for is fairly different from what Nvidia's chips are designed to do. Apple isn't doing cloud based LLMs. They're doing smaller, more focused models but distributed across more users. There's also no revenue being thrown off this service, so cost is a different consideration.

 

[oak8292]

But mikegg specifically referenced TSMC AZ, which 1) I don't think Apple will be able to pry away from AMD/Nvidia given it's criticality to those China sales, and 2) TSMC AZ won't remotely have 2nm any time soon so if Apple did make something there it's going to be radios or whatever they're using older nodes for. They could make AI chips on older nodes - Nvidia clearly is - but I don't think Apple can outspend Nvidia for capacity there.

Apple is saying they are using TSMC in Arizona for chips.

“That includes TSMC in Arizona, which is producing tens of millions of chips for Apple using one of the most advanced process technologies in America. Apple is this factory’s first and largest customer.”

Apple increases U.S. commitment to $600 billion, announces ambitious program

Apple announced a new $100 billion commitment to America, a significant acceleration of its U.S. commitment that now totals $600 billion over four years.

www.apple.com

 

[Doug S]

Apple is saying they are using TSMC in Arizona for chips.

“That includes TSMC in Arizona, which is producing tens of millions of chips for Apple using one of the most advanced process technologies in America. Apple is this factory’s first and largest customer.”

Everything that is on an N4/N5 node can be made there, but a lot of Apple's products on that node are probably leftover stock at this point. The only thing I think they're producing in such quantities today on that node is the C1 modem (the 40nm baseband chip, at least the radio chip is on N6RF and I'm not sure if they have that process in AZ)

Anyone know when TSMC starts N3 production in Arizona? With their "only older nodes outside of Taiwan" policy they could start production as early as the end of the year when N2 production begins in Taiwan.

 

[johnsonwax]

Not sure where TSMC AZ is now, but S10 is N4P. H2 can't be any newer than that and is almost certainly N5 or older. These are 10s of millions of units per quarter, but pretty small die sizes. Apple successor to the W series will probably not be on a more advanced node than C1, and we're expecting a MUCH higher volume C2 this year, so already in production. Apple really does have a lot of volume apart from their flagship chips.

 

[oak8292]

There are some other interesting comments about ‘Apple Silicon’ from that Apple published document. Apple is working with both GF and Samsung. Here are some excerpts;

• Apple is also working with Samsung at its fab in Austin, Texas, to launch an innovative new technology for making chips, which has never been used before anywhere in the world. By bringing this technology to the U.S. first, this facility will supply chips that optimize power and performance of Apple products, including iPhone devices shipped all over the world….

• GlobalFoundries and Apple have also entered an agreement to bring more semiconductor manufacturing to the United States, focused on manufacturing cutting-edge wireless technologies and advanced power management — critical technologies that enable longer battery life and enhanced connectivity in Apple devices. The partnership will bring new capabilities, jobs, and technology to the GlobalFoundries semiconductor facility in Malta, New York….

Apple is also working with Broadcom and GlobalFoundries to develop and manufacture additional cellular semiconductor components in the U.S. These components are crucial for 5G communications in Apple products.

Earlier this year, construction began in Houston on the new factory supporting production of advanced Apple servers, and in July, the facility produced its first test unit. The 250,000-square-foot server manufacturing facility is slated to begin mass production in 2026.

Apple increases U.S. commitment to $600 billion, announces ambitious program

Apple announced a new $100 billion commitment to America, a significant acceleration of its U.S. commitment that now totals $600 billion over four years.

www.apple.com