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

 

[name99]

Buddy, I retired early because I made millions understanding that.

But even growth companies have a business model that you can see 'once we climb the adoption curve we'll be profitable'. AI has already claimed the adoption curve, though. One of the notable things about it is that they climbed that curve almost immediately.


They know how to price inference, but they don't know how to also cover training.

This also seems to me a casualty of haste (and perhaps business model).
If you want to use anthropic language, the existing models try too hard to cram everything into "system-1" (ie trained weights) rather than using more "system-2" (ie reasoning and RAG).
This is an economics trade-off - training costs a bajillion dollars, but system-1 based inference is cheaper than system-2 based inference.

Did they get their calculations correct? Maybe they didn't even know the trade-offs (making wild guesses about how each side would scale as they planned for GPT-5)?
I think we're seeing some of the fallout of this now : hastily constructed query routers that try to guess whether or not to route to a reasoning model, but don't yet do a great job; and attempts (maybe, maybe not?) successful to try to bring reasonable pricing to expensive (but high quality, let's be fair) inference.

To add to what I said in my earlier post, I can believe that Apple here are positioned in a very different way from what a lot of people expect:

- could they have substantially better HW (especially with a few tweaks) for LARGE VLM training in house if they can offload all the "dumb mechanical video preprocessing" to ANE?

- could they target models that do a lot more reasoning (and thus need substantially fewer system-1 weights) because they can reliably offload short reasoning to local hardware and long reasoning to their data centers (where they control the billing, based on whatever tier of iCloud+ you're paying for...)

 

[Eug]

Geekbench has enough run to run variation that if enough people run it enough times, eventually they'll post a score that's near the top of its range for that hardware. When the first Geekbench results are produced for a product after release, we always see the "best score" go up a few percent until it pretty much stabilizes. In a week when millions of them are out there instead of thousands, it might go up a few more points yet.

Nobody was getting 4000 single-core and most were well under 3900, and only a minority were hitting just 10000 multi-core, so a much higher than average score would be around 3900/10000.

Then all of a sudden we have a single outlier at over 4000/11000, so I'd say there's a strong chance cooling is involved.

 

[name99]

AFAIK, there are no cars in existence that support AirPlay, and nobody is asking for it either.

There also used to be no cars in existence that supported CarPlay or Wireless CarPlay.
And no-one asking for them.

 

[name99]

Nobody was getting 4000 and only a minority were hitting just 10000, with a much higher than average score being around 3900/10000.

Then all of a sudden we have a single outlier at 4000/11000, so I'd say there's a strong chance cooling is involved.

OR finally a benchmarker who understands that you need to let the phone quiesce with its iOS updates and indexing and so on BEFORE you eagerly start running GB6?

 

[mvprod123]

Mac16,9 vs iPhone18,2 - Geekbench

In some ST subtests, A19 Pro outperforms M4 Max

 

[Eug]

There also used to be no cars in existence that supported CarPlay or Wireless CarPlay.
And no-one asking for them.

View attachment 130137

Uh what? People for YEARS asked for a better multimedia experience in cars, myself included. I personally wished I could wirelessly connect from my iPhone to an iPhone-like interface in the car. If you had perused the various forums you would have seen the repeated comments about how car companies should outsource their infotainment interface design, because they simply don't know how to do it themselves. CarPlay and Android Auto then appeared, as a direct response to that. (BTW, I had similar comments with Sony Ericsson and Nokia cellphones, since they had decent hardware but atrocious software, and then Apple changed the game completely a few years later with new hardware and software design.)

Ironically, I will point out the basis for AirPlay for cars actually already does exist in iOS 26, but no existing cars support it, and AFAIK no car manufacturer has announced support for this feature either. Its purpose though is to play video on the car's infotainment screen (not to funnel lossless audio from multiple iPhones through a single connected iPhone), but car manufacturers haven't seemed to have jumped on implementing this feature yet, maybe because it can't function while the car is moving anyway. I suspect the first cars that will support this are those that implement CarPlay Ultra, and brands like Kia which pushes cars based on their included infotainment features rather than reliability.

 

[Mopetar]

I don't either, but not everyone has to want something only enough people. It remains to be seen if that's an "enough people" thing or not.

My concern is that companies decide that the only model they'll make is the foldable phone and I don't get a choice. That's a rather silly worry that's not likely all that realistic, and one could always argue that a foldable phone is just like a regular one if you never fold it.

There's nothing I hate more than having some product that I've liked and that suited me perfectly undergo some kind of forced change for the sake of change and having things I don't want or need shoehorned in. Windows (and even to some degree OS X) are guilty of this.

The general lack of pockets or pockets that are tiny compared to men’s clothing is shocking.

I'll never understand this. Do women just complain about a lack of pockets and then refuse to buy the clothing that has them? The notion that we live in a a world where hundreds of companies will invest billions of dollars into trying to sell people things that no one really needs, but there isn't a single company that's even trying to sell people something that they want is absurd.

I can only conclude that women really don't want pants/clothes with big pockets (or will always wind up choosing small pocket clothing for some other reason if given a choice) or that the women who do want this fall into a small niche below the "enough people" threshold that Doug mentions. I e always just assumed that it was because most women carried purses and therefore pockets weren't as necessary. I don't really pay and mind to the fashion choices of either sex, but it does seem like yoga pants have become more popular and those are even worse for pocket space compared to women's jeans. Maybe purses are out of style and cross body bags are in. It's not unusual for younger generations to want something new in terms of fashion even it's not functionally different or better.

Then again there is the stereotype of the women's fashion industry largely being run by gay men, which does make the whole problem a lot more humorous.

For most of the world, their phones are literally the only computer they have. It's the source of their entertainment, work, social, study. When it's the only computer you have, you go for the biggest screen possible. Just travel to 3rd/2nd world countries and see how people use their phones. A great majority do not have laptops or desktops or even tablets. Heck, just look at younger Gen Zs in 1st world countries. They don't know how to use a Mac or Windows computer. Everything they do is on iOS or Android.

I think the problem with this argument is that for most things a foldable phone just offers a bigger screen for consumption. It's not a better work machine, etc. Anyone who benefits from a bigger display can get an even bigger one by being able to plug into or wirelessly connect to an external display. A physical keyboard is also a lot better as well. The foldable isn't any better or worse at connecting to external peripherals.

The 2nd/3rd world are also a lot less well off and less able to afford the fancy new technology. Apple even makes a new model using older chips or less expensive components for these markets because people still want "new" devices instead of something that's three years old. Spending twice as much money on a foldable seems like a waste, particularly if it's less reliable. In some of these countries a new phone can represent several months or even a year's salary for the average person. Spending more money on something that's less reliable or more prone to break isn't worth it for having a slightly larger screen, at least in my opinion.

Foldable phones feel a lot more like fashion to me. I can appreciate that some people like them for tech nerd reasons. I've purchased plenty of things that were years ahead of their time when they first came out, but weren't all that useful or good when it came down to it. The Apple Newton predated the iPad by over a decade. It was really cool when it first came out, but wasn't really all that useful for most people. I think this was true of PDAs in general. Even the first iPhone is kind of bad when you look at where we are now, but at the time it was completely revolutionary.

Uh what? People for YEARS asked for a better multimedia experience in cars, myself included. I personally wished I could wirelessly connect from my iPhone to an iPhone-like interface in the car.

I recently purchased a new vehicle and was mildly surprised that it couldn't do this. For long trips I'm plugging my phone in anyways so it can charge, but for short trips the idea that I should have to do this is rather annoying.

Maybe some woman is getting back at all of us for the lack of pockets in her pants.

 

[Eug]

I recently purchased a new vehicle and was mildly surprised that it couldn't do this. For long trips I'm plugging my phone in anyways so it can charge, but for short trips the idea that I should have to do this is rather annoying.

Maybe some woman is getting back at all of us for the lack of pockets in her pants.

Which car?

Up until last year, the Toyota Camry was only wired CarPlay. That was one reason why I didn't buy a used Camry. Other reasons were no Hybrid AWD option, and because the used pricing here is stupid high. Instead, I got a new 2025 Toyota Camry Hybrid AWD and it's a nicer car, now with wireless CarPlay.

Anyhow, you can buy a cheap dongle that turns wired CarPlay into wireless CarPlay.

General Motors has actually stopped offering CarPlay too, presumably because they don't want to pay Apple for it. Tesla never offered CarPlay in the first place. And nobody is offering CarPlay Ultra aside from a few boutique brands like Aston Martin. Supposedly Kia / Hyundai / Genesis eventually will support CarPlay Ultra, but it hasn't happened yet. Personally I don't care about CarPlay Ultra though.

 

[poke01]

the general reception is great. The 17s sell boat loads. The base 17 at $799 is the best flagship phone imo for most people.

Makes the base S25 and Pixel 10 look meh

 

[poke01]

Tim really outsmarted everyone, I think no one was expected same prices and more value. They assumed $50-$100 price increases.

perks of a good suply chain

 

[adroc_thurston]

The base 17 at $799 is the best flagship phone imo for most people.

Yeah for the first time in years the base iphone isn't hopelessly behind its peers.
But that makes Pro SKUs extremely niche (byebye promotion upsell).

 

[Eug]

Tim really outsmarted everyone, I think no one was expected same prices and more value. They assumed $50-$100 price increases.

perks of a good suply chain

iPhones got exempted from most of the tariffs... at the last minute.

https://www.cnn.com/2025/08/07/tech/trump-india-tariffs-apple-iphone-ai

Apple still raised the prices of some iPhones though, but it was done in a way that made it more palatable for many buyers.

US$799 - iPhone 16
US$799 - iPhone 17 <-- Price didn't change

US$899 - iPhone 16 Plus
US$999 - iPhone Air <-- Price increased $100. It's a different product, but in some ways it's inferior. Its storage doubled to 256 GB which is good, but it has a smaller screen and lost a camera which is bad.

$999 - iPhone 16 Pro
$1099 - iPhone 17 Pro <-- Price increased $100, but storage doubled to 256 GB.

$1199 - iPhone 16 Pro Max
$1199 - iPhone 17 Pro Max <-- Price didn't change. However, the price did increase $100 when the storage doubled to 256 GB in 2023 with the 15 Pro Max.

IOW, Apple has been gradually increasing base pricing, but doing it when it coincides with increased storage so people don't complain as much.

 

[poke01]

US$999 - iPhone Air <-- Price increased $100. It's a different product, but in some ways it's inferior. Its storage doubled to 256 GB which is good, but it has a smaller screen and lost a camera which is bad.

This is such an odd product. It’s weaker in those things you mentioned but then it has 12GB ram and A19 Pro.

No stereo speakers too. It’s only good if you want the lightest and thinnest iPhone.

 

[Eug]

This is such an odd product. It’s weaker in those things you mentioned but then it has 12GB ram and A19 Pro.

No stereo speakers too. It’s only good if you want the lightest and thinnest iPhone.

A19 Pro in the Air is binned though. Same number of GPU cores as A19 non-Pro.

 

[poke01]

A19 Pro in the Air is binned though. Same number of GPU cores as A19 non-Pro.

There is lot more cache in the A19 Pro, should help with efficiency more than A19.

 

[okoroezenwa]

It’s only good if you want the lightest and thinnest iPhone.

Well that seems to be why it was made so the people who want that should be able to deal with the compromises to get there.

 

[smalM]

Nobody was getting 4000 single-core and most were well under 3900, and only a minority were hitting just 10000 multi-core, so a much higher than average score would be around 3900/10000.

Then all of a sudden we have a single outlier at over 4000/11000, so I'd say there's a strong chance cooling is involved.

Nobody...
Of how many entries in the data base?

 

[poke01]

Apple Announces A19 and A19 Pro SoCs

Latest TSMC Process, GPU Tensor Cores, Memory Tagging for Security, & More

morethanmoore.substack.com

general Ryan article about A19

 

[poke01]

I missed this graph style. Glad to see back in morethanmoore

 

[Eug]

Nobody...
Of how many entries in the data base?

About 80 or so. In the last dozen there is a cluster of readings which are significantly higher than previous ones, but they all appear to be coming from the same unit which achieved the 4019/11054 score.

You guys are right to think this isn’t direct evidence to indicate someone is using cooling to get these scores. Nonetheless it makes it seem like someone is trying very hard to push the results higher, and I personally suspect it may involve cooling. I’m not suggesting geekerwan’s liquid nitrogen method, just something like putting it in a freezer, or even just putting a caseless iPhone on a cold stone block to better wick away heat.

 

[poke01]

https://twitter.com/x/status/1966578273845936155

finally a use case for SME2

 

[okoroezenwa]

https://twitter.com/x/status/1966578273845936155

finally a use case for SME2

Do you think AMX/SME weren’t used before?

 

[poke01]

Do you think AMX/SME weren’t used before?

AMX has been used before. SME, I don’t think has been used by any app yet. Could be wrong though

 

[ashFTW]

I ordered the base orange PM at 5:02 PT and got 9/25.

The delivery date has changed to 9/19

 

[smalM]

About 80 or so.

Use iPhone18,2 only and then compare the entries again.