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

 

[johnsonwax]

Well, the semiconductor space is getting all of the geopolitical drama. Intels future uncertain, TSMC at risk from China/Taiwan relations, and now martial law in South Korea.

I've always argued that Apple liked not owning their own fabs so they would be able to shop around, but none of the major players are reliable at the moment. I wonder if they're having a rethink on that one.

 

[name99]

TSMC is, apparently. With the transistors sandwiched between multiple oxide layers on both sides apparently getting the heat out of the chip is more of an issue than when the transistors were on the bottom (top when socketed) with only silicon in the way.

The people here who know the technical details of how the chips are made support that assessment and I don't have the knowledge base to argue with them, even though I don't understand the physics of that given how thin the oxide layers are.

Reference?

 

[digitaldreamer]

Well, the semiconductor space is getting all of the geopolitical drama. Intels future uncertain, TSMC at risk from China/Taiwan relations, and now martial law in South Korea.

I've always argued that Apple liked not owning their own fabs so they would be able to shop around, but none of the major players are reliable at the moment. I wonder if they're having a rethink on that one.

I doubt Apple would ever seriously consider building their own Fab(s). Could they afford it? Sure. But, not everything can be solved by throwing money at it. It's about getting the know how, experience, IP, and skirting around the hundreds of patents. Remember TSMC complaining about lack of educated workers for the US plant? And, witness how long it's taking to build a proper in-house modem. Six years on, it's only rumored to be installed in a small phone next year for limited release. And, that's after purchasing the IP and bringing Intel's engineers on board.
Just my 2¢

 

[gdansk]

Apple doesn't bother owning the (comparatively) cheap factories in which their products are assembled.

They're not moving into manufacturing, be it chips or anything else. They might pressure suppliers as best they can to diversify manufacturing locations to avoid risk but that's most likely the limit.

 

[The Hardcard]

I doubt Apple would ever seriously consider building their own Fab(s). Could they afford it? Sure. But, not everything can be solved by throwing money at it. It's about getting the know how, experience, IP, and skirting around the hundreds of patents. Remember TSMC complaining about lack of educated workers for the US plant? And, witness how long it's taking to build a proper in-house modem. Six years on, it's only rumored to be installed in a small phone next year for limited release. And, that's after purchasing the IP and bringing Intel's engineers on board.
Just my 2¢

IBM maintains semiconductor manufacturing research and in fact developed 2nm technology in 2021. They have their own patents and IP. IBM 2 nm development is the foundation for the attempt by Rapidus in Japan to launch as a fab within the next 3 years.

It is very unlikely that Rapidus has exclusive access to. There’s a chance that they have the goose that can lay golden eggs.

Apple doesn't bother owning the (comparatively) cheap factories in which their products are assembled.

They're not moving into manufacturing, be it chips or anything else. They might pressure suppliers as best they can to diversify manufacturing locations to avoid risk but that's most likely the limit.

They may not have a choice if certain things happen. If Samsung, TSMC, and Intel all go down for various reasons, there are very few companies who can step in and revive semiconductor manufacturing.

 

[Doug S]

Reference?

Read the last few pages of the Leading Edge Foundry Node thread.

 

[Doug S]

I've always argued that Apple liked not owning their own fabs so they would be able to shop around, but none of the major players are reliable at the moment. I wonder if they're having a rethink on that one.

No Apple doesn't own their own fabs because it would cost them more and incur greater risk. If TSMC's future roadmap goes off the rails they can talk to Intel and Samsung. If their own roadmap goes off the rails, they have to use it even when it sucks - just like Intel did for several years of 10nm+++++++++++.

Not sure why you think Apple would want to own the most expensive part of production, when they outsource the least expensive part to Foxconn?

 

[jdubs03]

IBM maintains semiconductor manufacturing research and in fact developed 2nm technology in 2021. They have their own patents and IP. IBM 2 nm development is the foundation for the attempt by Rapidus in Japan to launch as a fab within the next 3 years.

It is very unlikely that Rapidus has exclusive access to. There’s a chance that they have the goose that can lay golden eggs.



They may not have a choice if certain things happen. If Samsung, TSMC, and Intel all go down for various reasons, there are very few companies who can step in and revive semiconductor manufacturing.

At that point, a multinational joint venture would probably be the most appropriate entity to ensure consistent state of the art chip volume.

 

[johnsonwax]

They may not have a choice if certain things happen. If Samsung, TSMC, and Intel all go down for various reasons, there are very few companies who can step in and revive semiconductor manufacturing.

That was my point. China may decide the fate of TSMC. Nobody knows what's going on in South Korea right now, and events may at some point overtake Samsung. We have an unpredictable leader entering the US. Intel couldn't possibly meet Apple's demand right now. At some point you decide to take your fate in your own hands. Apple didn't want to do maps - Google forced their hand. They didn't want to do browsers, Microsoft forced that. I don't think Apple wanted to make radios, but Qualcomm made that unavoidable.

 

[johnsonwax]

At that point, a multinational joint venture would probably be the most appropriate entity to ensure consistent state of the art chip volume.

I don't think multinational. I think a domestic joint venture is more likely because that's what the federal government would put their support behind.

 

[jdubs03]

I don't think multinational. I think a domestic joint venture is more likely because that's what the federal government would put their support behind.

But they’d (US government) likely want TSMCs IP to ensure that it doesn’t go to China. So it could be a TSMC-Intel JV. Though I’m sure it’d come to the US regardless if China were to invade.

I’m still of the opinion that China won’t invade because of the silicon shield.

 

[poke01]

The iPad Pro equipped with the M5 processor is expected to enter mass production in 2H25. The company’s business momentum in 2H25 is anticipated to benefit significantly from Apple’s new product launches.
-kuo

Looks M5 iPad Pro is a 2026 product. M5 looks like its coming to Mac first this time.

BYD Electronic and Tianma Microelectronics as Key Suppliers for Appleâs Smart Home Products

For an analysis of Appleâs smart home strategy, refer here.

medium.com

 

[jdubs03]

The iPad Pro equipped with the M5 processor is expected to enter mass production in 2H25. The company’s business momentum in 2H25 is anticipated to benefit significantly from Apple’s new product launches.
-kuo

Looks M5 iPad Pro is a 2026 product. M5 looks like its coming to Mac first this time.

BYD Electronic and Tianma Microelectronics as Key Suppliers for Appleâs Smart Home Products

For an analysis of Appleâs smart home strategy, refer here.

medium.com

That’s how it should be tbh. It’s kind of wild to think that the iPad Pro released almost 7 months ago. Twas a good cycle to buy that early. Still is up until midway next year.

 

[igor_kavinski]

I think one thing that doesn't get mentioned is the advanced silicon design and simulation tools Apple engineers must be using to not require unnecessary silicon tapeouts. I'm pretty sure that they have designed those tools in-house and they are proprietary. And another important factor in their rapid progress is that they must be working on two generations at once. So right now, they must be testing or running simulations for M5 while a subset of their team must be working on completing the M6 design.

 

[Doug S]

I think one thing that doesn't get mentioned is the advanced silicon design and simulation tools Apple engineers must be using to not require unnecessary silicon tapeouts. I'm pretty sure that they have designed those tools in-house and they are proprietary. And another important factor in their rapid progress is that they must be working on two generations at once. So right now, they must be testing or running simulations for M5 while a subset of their team must be working on completing the M6 design.

They acquired a lot of that with their purchase of Intrinsity in 2010.

And they are working on more than two generations at once. Everyone is. They have likely started work on A22/M8 already.

 

[mvprod123]

Тrue. Jim Keller mentioned it in his interview. He told us some very interesting things about his work at Apple. Starting from 55:25.

 

[poke01]

Тrue. Jim Keller mentioned it in his interview. He told us some very interesting things about his work at Apple. Starting from 55:25.

Watching this from Jim Kepler interview and knowing he doesn’t sugar coat things really puts into perspective how Apple’s silicon team worked/works.

Thanks for this.

 

[LightningZ71]

One of the things I noted from Jim's interview with respect to Apple is something that we bring up a lot on here. Apple has a full grasp of and full control of the entire stack from the display screen all the way down to the individual gate arrangements on the CPU silicon. He referenced "hardware widgets" as being a key component to how a lot of things that Apple does on a high level gets handled optimally at a low level. They have a notable margin advantage over the likes of AMD and Intel under Windows/Linux as implemented on, say, Dell computers to be able to afford to throw "twice" the transistors at a processor while still making a profit which enables them to have custom accelerators and other related hardware solutions to make otherwise difficult or slow things work in an optimal way. They also have control over the software stack, enabling them to transparently flow those tasks to those accelerators in a way that the programmers don't have to typically concern themselves with directly.

This is a huge advantage for Apple. In the Wintel and Lintel world, you have no idea what your program AND OS will be executed on. It could be an over a decade old Sandy Bridge CPU, or a 2 core, 2 thread atom based celeron all the way up for a 24 core Arrow Lake processor and beyond to a 192 core Epyc processor. There are dozens of different combinations of ISA extensions possible in that array, with not always the best OS support for all the processor features present. Then, with Linux, you have many projects that are part of Kernel space that are often maintained by less than half a dozen people with spotty documentation and known bugs that have been passed on for a decade or more. MS has as much in the way of politics and profit motive tied into Windows as they also have their own mish mash of bugs and poorly implemented features. There is little incentive for programmers to target those ISA extensions and advanced features if only a small portion of the market actually has them, and the OS does a poor job of handling their other effects along the way.

As we see from the Phoronix benchmarks, x86 on leading edge processors can sometimes be extremely performant in highly optimized programming situations. But, from a top down perspective, Apple does about as good of a job as can be imagined these days on both building a top of the line processor and also wringing every last drop of performance from it.

 

[name99]

One of the things I noted from Jim's interview with respect to Apple is something that we bring up a lot on here. Apple has a full grasp of and full control of the entire stack from the display screen all the way down to the individual gate arrangements on the CPU silicon. He referenced "hardware widgets" as being a key component to how a lot of things that Apple does on a high level gets handled optimally at a low level. They have a notable margin advantage over the likes of AMD and Intel under Windows/Linux as implemented on, say, Dell computers to be able to afford to throw "twice" the transistors at a processor while still making a profit which enables them to have custom accelerators and other related hardware solutions to make otherwise difficult or slow things work in an optimal way. They also have control over the software stack, enabling them to transparently flow those tasks to those accelerators in a way that the programmers don't have to typically concern themselves with directly.

This is a huge advantage for Apple. In the Wintel and Lintel world, you have no idea what your program AND OS will be executed on. It could be an over a decade old Sandy Bridge CPU, or a 2 core, 2 thread atom based celeron all the way up for a 24 core Arrow Lake processor and beyond to a 192 core Epyc processor. There are dozens of different combinations of ISA extensions possible in that array, with not always the best OS support for all the processor features present. Then, with Linux, you have many projects that are part of Kernel space that are often maintained by less than half a dozen people with spotty documentation and known bugs that have been passed on for a decade or more. MS has as much in the way of politics and profit motive tied into Windows as they also have their own mish mash of bugs and poorly implemented features. There is little incentive for programmers to target those ISA extensions and advanced features if only a small portion of the market actually has them, and the OS does a poor job of handling their other effects along the way.

As we see from the Phoronix benchmarks, x86 on leading edge processors can sometimes be extremely performant in highly optimized programming situations. But, from a top down perspective, Apple does about as good of a job as can be imagined these days on both building a top of the line processor and also wringing every last drop of performance from it.

As I have pointed out before, Apple area's are not notably different from the competition. If they get twice the transistors in the same area, it's because they have made the deliberate decision to push for smarts (IPC) rather than GHz...

But even that is a dumb analysis. The real analysis is that they are not "monists" - they don't think there's a single solution to every problem. technology changes, and they adapt their designs to be an optimal match to technology. There was a window when transistors (and wires, just as important) were "free" and they designed for that going very wide and very smart. When that dried up around N5, they switched to pushing frequency (the transistors can go faster, it's just density has not climbed much) BUT designed to burn as little energy as possible. (Again throw smarts at the problem...)
When we get BSPDN, the designs will again evolve.

All of this was available to Intel. The micro-architectural ideas that Apple uses are mainly productization of ideas that have been around, sometimes for 20 years. Many of them were funded by Intel dollars! But every time there was a choice, Intel chose raw GHz over IPC. With knock-on effects, all the way down to their fab being uninteresting for most would-be customers because it was optimized for GHz rather than density...
Even the few successes at Intel (like Lunar Lake) feel like they were achieved by endless civil war with the old guard (about hyerthreading, smarts vs GHz, DRAM on package, etc), and feel like the people who achieved them are probably so sick of the fighting, the stupidity, and the criticism from the rest of the company that they will probably move to Apple soon.

"There are dozens of different combinations of ISA extensions possible in that array, with not always the best OS support for all the processor features present. "
All this sort of thing is SELF-INFLICTED. INTEL made the choice to split the ISA 15 ways to Sunday to micro-segment the market. No-one forced to do this. Apple isn't out there restricting AMX/SME to M Pro and M Max, so that it's not available on phones or Macbook Air...
Intel has (or had, until they blew themselves up) massive agency.
They chose to go down this path, against the advice of many many people.

"As we see from the Phoronix benchmarks, x86 on leading edge processors can sometimes be extremely performant"
Phoronix is part of the problem! Phoronix is not an unbiased actor interested in understanding CPUs, they're a propaganda site through and through. They mix up benchmarks that test throughput (ie throw multiple at the problem) vs those that test latency (ie single-threaded performance). They mix up different libraries even when those libraries (eg calculating random numbers) take up a large amount of the time. IO is even worse, when they refuse to understand tradeoffs between IO performance vs "safety" and how different platforms have made different choices.

Phoronix is the sort of enabler that allows addict Intel to keep going, bcs Phoronix-type sites are always there whispering "I got you, baby. Sure you have a problem but you're still my baby, still the fastest CPU out there no matter what the rest of the world says".

I don't want this to be a rah rah Apple post; but I am SO SICK of the Intel excuses. They boarded this train when they refused to listen to Apple's requests for a phone CPU 20 years ago. EVERY YEAR since then they've had an opportunity to look at the market and reconsider. And EVERY YEAR they have decided to make the stupid choice.

 

[Eug]

Luke Miani demonstrates installation of reverse engineered Mac Studio NAND modules

https://www.kickstarter.com/projects/polysoftservices/studio-drive

 

[Doug S]

I'm sorry. I'm sure that means something to you, but it means nothing to me. Thread on what site? Here? On which board?

Its literally another thread in the same location as this one. You will see it as you click down to this thread, unless you have this one bookmarked and never read anything else. Don't ask for references then complain it is too hard to find when it is the same shelf as the book you're currently reading.

 

[name99]

Read the last few pages of the Leading Edge Foundry Node thread.

You ignored the part where I said "especially at mobile power levels?"...
I've no knowledge of or interest in behavior at kW power levels.

I continue to assert that
- this will not be an issue at Apple power levels.
- likewise the claims that this is only beneficial for high power devices. For Apple (again I don't care about kW devices) the BIG win is alleviating wiring congestion, which in turn allows for continued scaling; especially in SRAMs, which have been limited in scaling over the past few generations bcs of wiring issues.

 

[name99]

Its literally another thread in the same location as this one. You will see it as you click down to this thread, unless you have this one bookmarked and never read anything else. Don't ask for references then complain it is too hard to find when it is the same shelf as the book you're currently reading.

Sorry, that shouldn't have been posted. Consequence of this site (like Macrumors) bundling successive not-posted comments together unless you are careful and notice what's at the top of your last comment :-(

I found the thread.

 

[LightningZ71]

As I have pointed out before, Apple area's are not notably different from the competition. If they get twice the transistors in the same area, it's because they have made the deliberate decision to push for smarts (IPC) rather than GHz...

But even that is a dumb analysis. The real analysis is that they are not "monists" - they don't think there's a single solution to every problem. technology changes, and they adapt their designs to be an optimal match to technology. There was a window when transistors (and wires, just as important) were "free" and they designed for that going very wide and very smart. When that dried up around N5, they switched to pushing frequency (the transistors can go faster, it's just density has not climbed much) BUT designed to burn as little energy as possible. (Again throw smarts at the problem...)
When we get BSPDN, the designs will again evolve.

All of this was available to Intel. The micro-architectural ideas that Apple uses are mainly productization of ideas that have been around, sometimes for 20 years. Many of them were funded by Intel dollars! But every time there was a choice, Intel chose raw GHz over IPC. With knock-on effects, all the way down to their fab being uninteresting for most would-be customers because it was optimized for GHz rather than density...
Even the few successes at Intel (like Lunar Lake) feel like they were achieved by endless civil war with the old guard (about hyerthreading, smarts vs GHz, DRAM on package, etc), and feel like the people who achieved them are probably so sick of the fighting, the stupidity, and the criticism from the rest of the company that they will probably move to Apple soon.

"There are dozens of different combinations of ISA extensions possible in that array, with not always the best OS support for all the processor features present. "
All this sort of thing is SELF-INFLICTED. INTEL made the choice to split the ISA 15 ways to Sunday to micro-segment the market. No-one forced to do this. Apple isn't out there restricting AMX/SME to M Pro and M Max, so that it's not available on phones or Macbook Air...
Intel has (or had, until they blew themselves up) massive agency.
They chose to go down this path, against the advice of many many people.

"As we see from the Phoronix benchmarks, x86 on leading edge processors can sometimes be extremely performant"
Phoronix is part of the problem! Phoronix is not an unbiased actor interested in understanding CPUs, they're a propaganda site through and through. They mix up benchmarks that test throughput (ie throw multiple at the problem) vs those that test latency (ie single-threaded performance). They mix up different libraries even when those libraries (eg calculating random numbers) take up a large amount of the time. IO is even worse, when they refuse to understand tradeoffs between IO performance vs "safety" and how different platforms have made different choices.

Phoronix is the sort of enabler that allows addict Intel to keep going, bcs Phoronix-type sites are always there whispering "I got you, baby. Sure you have a problem but you're still my baby, still the fastest CPU out there no matter what the rest of the world says".

I don't want this to be a rah rah Apple post; but I am SO SICK of the Intel excuses. They boarded this train when they refused to listen to Apple's requests for a phone CPU 20 years ago. EVERY YEAR since then they've had an opportunity to look at the market and reconsider. And EVERY YEAR they have decided to make the stupid choice.

I won't dispute most of what you said, but I will point out one very significant difference between Apple vs. Intel/AMD: the server market. Apple has intentionally chosen not to play in the data center field (excluding a recent effort at AI clustering their own computers). Intel and AMD, to keep their own costs reigned in, reuse large portions of or entire cores from all the way at the top of their server stack down to the sub $300 laptop market. That's a LOT of ground to cover, a portion of which is wholely unaddressed by Apple. Big x86 made a ton of money on servers while Apple was blowing in the wind from PowerPC to x86 to their own ARM products on the Mac line while not earning a lot of money for much of that time from said products.

Now, the tables have turned, largely due to Intel shooting themselves in the foot repeatedly and AMD not being large enough to dictate the market themselves. Apple is fully vertically integrated and Intel is on the verge of imploding. You can blame Intel for going for Ghz, but that was a symptom of bad management, not the cause of their problems.

 

[Gideon]

Now, the tables have turned, largely due to Intel shooting themselves in the foot repeatedly and AMD not being large enough to dictate the market themselves. Apple is fully vertically integrated and Intel is on the verge of imploding. You can blame Intel for going for Ghz, but that was a symptom of bad management, not the cause of their problems.

The biggest problem with that approach was, that the hyperscalers wanted out of this model. Jon Masters said already a decade ago that they were going vertically integrate.

Now over 50% of Amazon CPUs are Graviton and they look to do the same for AI tasks:

AWS Reaps The Benefits Of The Custom Silicon It Has Sown

When the hyperscalers and cloud builders were smaller and the Arm collective had failed to storm the datacenter and AMD was not yet on its path to

www.nextplatform.com

Same is true for Microsoft, Google ...

x86 is currently on the trajectory to the legacy status RISC and PowerPC, etc processors became in the 00-ies (staying mostly relevant for legacy apps and enterprise)

Intel's issues will only sperf it up