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]

Inflation is money becoming less valuable, not products becoming more expensive. Double the number of dollars in circulation and whatever you have now starts to approach half of its previous value relative to existing goods and services as the market adjusts to the sudden influx.

Inflation is making the value of any money that companies have lose value in just the same way that your money is losing value. Any amount of inflation will lead to "record profits" because a company has to take in more dollars to have the same value as previously. Unless inflation is adjusted for you can't just use the raw number. It's no different than the record breaking box office figures for new films. Of course they want to make it seem impressive so they tend not to adjust for inflation or changes in ticket prices.

Inflation is not always due to more dollars in circulation. The Fed has been shrinking its balance sheet quite hard of late. If you have less of what people want to buy, people are willing to spend more money to get it. That's not dollar devaluation, that's pure price inflation of those goods - classic supply and demand. When you have multiple waves of supply chain issues, production issues, and shipping/transport issues all at the same time, like we did due to covid, you're going to get shortages of a lot of stuff.

Purely increasing the money supply generally does not have much of an effect on inflation. That's why previous attempts to increase the money supply (i.e. the multiple rounds of QE) didn't create inflation, or increase economic growth, but at least they insured we would not succumb to deflation like Japan's lost decades. They can put money into circulation but they can't force people to spend it. The velocity of money kept decreasing as the money supply increased, the futility of that was compared by some to pushing on a string. What that does do is cause asset price inflation, because the people who can access that extra money supply are mostly those who already have a lot of money and are able to borrow this new money. They aren't going to buy more burgers or more Toyotas, instead they will buy houses, land, art, stocks, bonds, bitcoin, NFTs. Basically anything they think they will go up in price. Even if they know it is a bubble if they feel they can ride it up and get out before it bursts they'll be in it. Blaming the increase in the money supply on cars or eggs costing more is misplaced. Blaming it on houses and rent costing more is absolutely on the mark.

BTW the record corporate profits were after adjustment for inflation.

 

[Mopetar]

Inflation is not always due to more dollars in circulation. The Fed has been shrinking its balance sheet quite hard of late. If you have less of what people want to buy, people are willing to spend more money to get it.

It's also possible to devalue money if an economy shrinks in size while the supply of money remains the same. To some degree this occurred during the last few years when many businesses were shuttered or saw their revenues decrease due to lockdowns. Normally this isn't an issue though as economies tend to grow and have an overall upward trend.

That's not dollar devaluation, that's pure price inflation of those goods - classic supply and demand. When you have multiple waves of supply chain issues, production issues, and shipping/transport issues all at the same time, like we did due to covid, you're going to get shortages of a lot of stuff.

If it's just one industry or area, it's not really inflation, just businesses passing on their own increased costs to their customers. But that's normal and how markets adjust so that production isn't being funneled into things that aren't valued.

Purely increasing the money supply generally does not have much of an effect on inflation.

This is completely wrong and ignores countless historical examples of countries that have tried to get out from under their debt by running the printing press. Both Zimbabwe and Venezuela are recent examples of this. Their governments kept creating money at rates vastly outpacing their own economic growth, which caused further damage to their own economies making the problem worse.

Modern banking naturally creates new money due to the nature of how fractional reserve lending works, which is why countries hike interest rates to curb inflation. Reducing the number of people borrowing money slows the rate at which new money enters circulation. This of course has its own set of issues, and it's best not to reach that point in the first place.

GPU prices, for example, are higher for a variety of reasons. Costs are up, particularly for wafers on the newest nodes and there's only one fab that can make them. Both NVidia and AMD have products that give them better margins on those wafers, so supply of consumer GPUs is going to be constrained. Higher than average inflation rates many some countries (Euro-area was at 8% last year when normally it's been under 2% the last decade) account for some of the increase as well.

Personally I don't see Apple releasing a cheap Mac, even if their costs haven't changed much. Really they could accomplish something like that by making a keyboard dock for their iPads and a version of iOS that lets it run like a modern OS (or more like one) when in that mode. I think that a lot of the success of Chromebooks in that market was a lack of anything else for schools to buy during the pandemic. It doesn't look like the people who're using them now are going to buy a their own later on in life.

 

[Eug]

I think that a lot of the success of Chromebooks in that market was a lack of anything else for schools to buy during the pandemic.

Chromebooks are not only dirt cheap for the hardware, they are also supposedly much easier and cheaper to manage for education institutions, and with free Google software to manage the devices. Furthermore, the education and productivity software suites that both the teachers and students use is also free and straight from Google itself, and that client software is also cross platform and free on those other platforms. That said, Google's applications work best on Chromebooks for obvious reasons.

To summarize, It's a fully integrated and cross-platform education ecosystem with most of the software side completely free, paired with cheap hardware that's easy to manage.

I believe part of the reason everything is so cheap is that the data is harvested - Google's core business - at least for the teens and older. Google claims it does not harvest information for those under 13 without parental consent, but they are currently being sued in the US for doing just that.

It doesn't look like the people who're using them now are going to buy a their own later on in life.

Yes it's true that later in life or even at the same time for home use, consumers will buy other machines - specifically Windows or macOS machines, but that's a completely different discussion.

FWIW, at home we are an all Apple setup, but with Google's edu applications running for the kids on our Macs (via the Chrome browser) and iPads (via dedicated apps). At school they run Chromebooks with that same software, and also iPads with other software.

BTW, AFAIK, almost no institutions buy Macs in large volumes anymore for primary school students. If Apple, they're almost all iPads. That's how Apple competes with the Chromebook price point in the edu space - not with cheap Macs, but with entry level iPads.

 

[Doug S]

This is completely wrong and ignores countless historical examples of countries that have tried to get out from under their debt by running the printing press. Both Zimbabwe and Venezuela are recent examples of this. Their governments kept creating money at rates vastly outpacing their own economic growth, which caused further damage to their own economies making the problem worse.

If you don't understand the massive difference between increasing the money supply and actually printing physical cash in larger and larger denominations, there's no point in continuing this discussion.

 

[Mopetar]

If you don't understand the massive difference between increasing the money supply and actually printing physical cash in larger and larger denominations, there's no point in continuing this discussion.

If you don't understand that running the printing press is a euphemism for increasing the supply of money then I can't help you. Not that it matters whether it's physical or digital, the effect is the same.

This is basic supply and demand, yet you've convinced yourself that money is somehow magical in ways I can't even begin to imagine. Go back to prior history when countries used gold or other precious metals for currency. There are still examples where sudden influxes of gold or discoveries of new sources lead to inflation. More money without a corresponding increase in goods and services makes an existing amount of money lose value relative to what it had before.

Most countries even want a small amount of inflation to exist. It means that there's a cost for just sitting on your money as it slowly loses value over time. People are incentivized to invest or spend money which creates economic activity. As long as the inflation is really small (say 1-2%) then no one really notices or is bothered. Increase it much beyond that and you need constant raises or cost of living adjustments or even a good paying job won't keep you housed or fed after a moderate (or even small) number of years.

If rapidly expanding the supply of money doesn't cause inflation, please provide an explanation for why Zimbabwe was printing one hundred trillion dollar bank notes in the mid-2000s.

$100 trillion. They most have been the richest country on earth with that kind of wealth. Surely creating money at such a rate couldn't have devalued the currency to the point where it wasn't even worth the materials used in its creation.

Maybe we should do the same. Why worry about the national debt if we can just make enough new money to pay it off in one fell swoop. There's no possible negative consequences to doing this after all.

 

[Doug S]

Not that it matters whether it's physical or digital, the effect is the same.

That's is 100% completely and totally incorrect.

 

[A///]

I'm half in the bag. Can someone summarise the last two pages of bickering?

 

[igor_kavinski]

Can someone summarise the last two pages of bickering?

Don't print money!

 

[Mopetar]

That's is 100% completely and totally incorrect.

If we weren't crapping up a completely unrelated thread I'd ask you to elaborate.

I'm in the mood for a good laugh.

 

[Eug]

We should be expecting M3-based iMac and MBP13 in October based on Mark Gurman. At least the first iteration of M3 will be based on TSMC's N3 process

Ming-Chi Kuo agrees with my prediction (from earlier this summer) that there will be no M3 release this year.

https://twitter.com/x/status/1700193866178343036

So are we still thinking N3B for M3, or N3E?

 

[Saylick]

Ming-Chi Kuo agrees with my prediction (from earlier this summer) that there will be no M3 release this year.

https://twitter.com/x/status/1700193866178343036

So are we still thinking N3B for M3, or N3E?

I'm of the opinion that Apple wants to get off of N3B and onto N3E ASAP. TSMC is likely in the same boat.

 

[Doug S]

I'm of the opinion that Apple wants to get off of N3B and onto N3E ASAP. TSMC is likely in the same boat.

So long as Apple is getting the KGD pricing I don't think they care N3B vs N3E. TSMC obviously wants to forget about N3B like a bad dream ASAP, they would be the ones pushing this with both carrots and sticks.

So if that KGD deal has an expiration date it may not be worth it to Apple to make M3s for a few months on N3B then switch to N3E. Just wait and make them all with N3E.

 

[Eug]

I'm of the opinion that Apple wants to get off of N3B and onto N3E ASAP. TSMC is likely in the same boat.

So long as Apple is getting the KGD pricing I don't think they care N3B vs N3E. TSMC obviously wants to forget about N3B like a bad dream ASAP, they would be the ones pushing this with both carrots and sticks.

So if that KGD deal has an expiration date it may not be worth it to Apple to make M3s for a few months on N3B then switch to N3E. Just wait and make them all with N3E.

So, now people have warmed to the idea of M3 on N3E. A17-based too then?

 

[Doug S]

So, now people have warmed to the idea of M3 on N3E. A17-based too then?

I have said for a couple years that M3 would use A17 cores. If M3 waits a bit longer than I guessed (I was always assuming fall 2023) for N3E which seems more and more likely then IMHO there's a good possibility that even the A17s made on N3B are actually hybrid (pessimistic) N3B/N3E designs - i.e. they won't take advantage of the (slightly) denser cache possible with N3B, nor will they take advantage of the features that N3E offers over N3B. We'll know if this is true if the cache does not shrink at all from A16's when Techinsights does one of their die evaluations.

Then they wouldn't need new mask sets or changes to the design, TSMC would keep using the same masks on N3E once that's ready and the only way you might be able to tell is if you run some sort of CPU burn test designed to drain the battery which would last a bit longer on iPhone 15s with N3E A17s. The cost of that assuming the cores are basically ported over would be M3s would underperform a bit versus what they are capable of, but they should be enough of a jump over M2 it doesn't matter (and that makes it easier for M4 to exceed as well, despite only getting a bump to N3P)

 

[Eug]

It looks like Apple's foray into modem chip design is not going that great.

Qualcomm says it will supply Apple with 5G modems for iPhones through 2026

Wall Street analysts and Qualcomm officials had previously said they expected Apple to use an internally developed 5G modem starting in 2024.

www.cnbc.com

Qualcomm said Monday that it will supply Apple with 5G modems for smartphones through 2026.

Wall Street analysts and Qualcomm officials had previously said they expected Apple to use an internally developed 5G modem starting in 2024.

I'm pleased that when I upgrade my iPhone 12 Pro Max to the iPhone 16 series in 2024, I won't be a guinea pig for a first gen Apple modem. And when my wife upgrades her iPhone 14 Pro Max to the iPhone 18 series in 2026, she may get a Qualcomm modem too.

Will Apple's modems be discrete chips, not integrated into the A series?

 

[Doug S]

Will Apple's modems be discrete chips, not integrated into the A series?

At first they will I'm sure. They have to make them discrete at first for flexibility until the transition from Qualcomm is complete. Even then maybe there are advantages to having them separate, since there are products (lower end iPads, Apple TV, HomePod) that use those SoCs but don't need cellular - and depending on how the patent licensing works they don't want to pay patent license fees for LTE/5G patents for every Apple TV sold.

When Apple & Qualcomm made the original deal it was for four years but could be extended to six. Looks like Apple is taking the extra two years.

Based on what Qualcomm said about 20% of iPhones using Qualcomm modems after the 2026 iPhones launch, the 2025 iPhones will be using Qualcomm and 2026 will be using Apple's. I'm sure they would put it in the SE first to give them time to address software issues (given that baseband software is the hard part about making modems, not the hardware) Since the current SE was 2022, the 4th gen is expected next spring, it makes sense to do it with the 5th gen coming in spring 2026. We probably also see Apple modems in iPads purchased with the cellular option before the main iPhone launch as well.

At least by taking so long there is no conceivable argument against Apple doing what I've always said they would, and only implementing LTE and 5G in their modem. They can probably add the eSIMs to it as well as they can probably drop support for physical SIMs worldwide for products released in 2026. So at least they buy some simplification for taking so long...

 

[Eug]

At first they will I'm sure. They have to make them discrete at first for flexibility until the transition from Qualcomm is complete. Even then maybe there are advantages to having them separate, since there are products (lower end iPads, Apple TV, HomePod) that use those SoCs but don't need cellular - and depending on how the patent licensing works they don't want to pay patent license fees for LTE/5G patents for every Apple TV sold.

Yes, that makes sense. I was thinking that in the first iteration, whenever that may be, not only the iPhone SE but also possibly the non-Pro iPhones could get the discrete Apple modem, while the Pros would get Qualcomm. Or else different mixes in different markets.

And then there's that whole issue of non-cellular iPads and Apple TV and displays, etc. that use A-series chips. BTW, I wonder how much bigger an A-series chip would have to be to get an integrated modem. Then there's M-series too for the higher end iPads.

 

[Doug S]

possibly the non-Pro iPhones could get the discrete Apple modem, while the Pros would get Qualcomm. Or else different mixes in different markets.

No chance they do it separately for Pro and non Pro iPhones. That would basically be saying "we don't think this is ready for our most important customers". Different mixes in different countries is something I hadn't considered but could be possible as a transition. There is a lot of variation around the world in base station hardware, carrier protcols and so forth. That's a big reason that companies that have their own modems like Samsung and HiSilicon only make them available in limited areas, and use Qualcomm for anything they hope to sell worldwide. They have got those issues licked in China or South Korea or Africa or whatever but not everywhere.

That's honestly probably Apple's biggest issue. I wouldn't be surprised if their modem was ready to go today for use in the US, but wouldn't work well in China, or India or Japan or South America or whatever. Obviously Apple couldn't sell phones with a US only modem today when so many Americans travel and everyone is used to bringing their phone with them and using it around the world when they travel. Qualcomm has had decades to figure all that stuff out - and enough clout that when something goes wrong they can pressure base station vendors to fix their software or carriers to fix their protocols.

Apple WILL have that clout, but only AFTER they release their stuff. Right now they have no customers in the field using their modems, so they can't put much pressure on others to work with them - at least not the degree they will be able to once there are millions of their modems shipping every week.

 

[smalM]

I don't think Apple will waste that precius N2P capacity for an integrated modem and will instead produce a discrete modem on an acient node like N4...

 

[Ajay]

I don't think Apple will waste that precious N2P capacity for an integrated modem and will instead produce a discrete modem on an ancient node like N4...

I agree. I don't see the benefit of an integrated modem. Silly talk to me.

 

[Mopetar]

Integrated modem means simpler packaging and board design overall since it's a single SoC. I don't see Apple adding more and more cores to the iPhone/iPad so they'll have the room for it.

If anything they have no reason to make a separate chip as they don't sell their chips to other companies so they don't save on wafers unless they want to build the baseband chip on an older node. That makes them use more power though so there's trade offs.

Their problem right now is that what they have just isn't competitive with Qualcomm's offering, so it's not worth putting it into production. Whether they ever get to that point or not is another matter.

Maybe it seems like they're burning through a ton a money to get absolutely no where, but they're making hundreds of millions of phones and tablets (and other devices) every year that could use their own baseband. If they do get there, it saves them billions of dollars per year to use their own design. It makes the investment worth it, particularly if it lets them control the ability to add features that would require yet another separate chip.

Regardless of what they ultimately end up doing, they won't go into production for even a stand-alone baseband chip until they have something that's at least as good as alternatives. It's not worth the negative publicity for something that's only 95% as good. Once they get to (or surpass) the competition, I expect them to integrate it and move towards producing a single SoC.

 

[Doug S]

I agree. I don't see the benefit of an integrated modem. Silly talk to me.

Then why does Qualcomm do it, rather than selling OEMs separate Snapdragon and modem dies (which they did for the first gen of 5G then resumed integrating)

 

[eek2121]

Then why does Qualcomm do it, rather than selling OEMs separate Snapdragon and modem dies (which they did for the first gen of 5G then resumed integrating)

To force companies to buy both, rather than just one. Qualcomm’s entire business practice involves around this. Even patent licensing.

 

[Eug]

New Apple Watch S9 SoC:
1.6X transistors of S8 (TSMC N7P), 1.3X speed of S8
4-core neural engine, 2X speed of S8

 

[Nothingness]

"The SoC for the iPhone 15 is the A16. Which was intro'ed last year for the iPhone 14 Pro"

No A17 yet.

EDIT: So A17 is here but only for iPhone Pro. Is that a first? I'm not familiar with Apple phones