Discussion Apple Silicon SoC thread

[Eug]

M1
5 nm
Unified memory architecture - LP-DDR4
16 billion transistors

8-core CPU

4 high-performance cores
192 KB instruction cache
128 KB data cache
Shared 12 MB L2 cache

4 high-efficiency cores
128 KB instruction cache
64 KB data cache
Shared 4 MB L2 cache
(Apple claims the 4 high-effiency cores alone perform like a dual-core Intel MacBook Air)

8-core iGPU (but there is a 7-core variant, likely with one inactive core)
128 execution units
Up to 24576 concurrent threads
2.6 Teraflops
82 Gigatexels/s
41 gigapixels/s

16-core neural engine
Secure Enclave
USB 4

Products:
$999 ($899 edu) 13" MacBook Air (fanless) - 18 hour video playback battery life
$699 Mac mini (with fan)
$1299 ($1199 edu) 13" MacBook Pro (with fan) - 20 hour video playback battery life

Memory options 8 GB and 16 GB. No 32 GB option (unless you go Intel).

It should be noted that the M1 chip in these three Macs is the same (aside from GPU core number). Basically, Apple is taking the same approach which these chips as they do the iPhones and iPads. Just one SKU (excluding the X variants), which is the same across all iDevices (aside from maybe slight clock speed differences occasionally).

EDIT:



M1 Pro 8-core CPU (6+2), 14-core GPU
M1 Pro 10-core CPU (8+2), 14-core GPU
M1 Pro 10-core CPU (8+2), 16-core GPU
M1 Max 10-core CPU (8+2), 24-core GPU
M1 Max 10-core CPU (8+2), 32-core GPU

M1 Pro and M1 Max discussion here:

Page 78 - Discussion - Apple Silicon SoC thread

Page 78 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

M1 Ultra discussion here:

Page 109 - Discussion - Apple Silicon SoC thread

Page 109 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

M2 discussion here:

Page 127 - Discussion - Apple Silicon SoC thread

Page 127 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

Second Generation 5 nm
Unified memory architecture - LPDDR5, up to 24 GB and 100 GB/s
20 billion transistors

8-core CPU

4 high-performance cores
192 KB instruction cache
128 KB data cache
Shared 16 MB L2 cache

4 high-efficiency cores
128 KB instruction cache
64 KB data cache
Shared 4 MB L2 cache

10-core iGPU (but there is an 8-core variant)
3.6 Teraflops

16-core neural engine
Secure Enclave
USB 4

Hardware acceleration for 8K h.264, h.264, ProRes

M3 Family discussion here:

Page 215 - Discussion - Apple Silicon SoC thread

Page 215 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

M4 Family discussion here:

Page 263 - Discussion - Apple Silicon SoC thread

Page 263 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

M5 Family discussion here:

Page 431 - Discussion - Apple Silicon SoC thread

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

forums.anandtech.com

 

[fastandfurious6]

meta is a useless company

If all social media platforms had a price tag (i.e. acquisition by others), Instagram would have by far the most expensive price, with a massive difference from position #2

 

[branch_suggestion]

they have these bursts everytime there is a good product or redesign. The iPhone Fold is coming this year, next year is the 20th anniversary iPhone and next major one should be the all screen iPhone in 2030.

I think after 2030 it will be like the laptop market, completely mature and users will upgrade when major camera and battery advancements arrive.

This burst is a redistribution of the market moreso than the market growing.
Poverty Android does not exist when memory is expensive, that is a big market that needs filling.

 

[mikegg]

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

I'll answer it after you tell me why you think OpenAI is a scam.

 

[mikegg]

This guy explains it very nice.

Tldr, they do not have and will not have the revenue to pay back all the investments needed. That's the biggest one. Also the power needed to run the data centers that they bought all future ram and gpus is not available and will not be available, and even then, there will be backlash from communities where they are set up, for raising utilities for everyone.

One of my favorite Youtubers visited Institute of Advanced Study at Princeton recently. He basically said all the professors there agree that AI is incredibly useful and is completely changing their field, including people like Ed Witten. Worth watching.

I have the same experience as him. AI has already made it such that I no longer write code. I'm a software engineer.

 

[mikegg]

advertising is big money.

So not useless.

 

[The Hardcard]

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

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

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

Are they better? I know certain brands have started promising 7 years of updates, but has it played out yet for anyone?

I got an iPhone 6S in 2016, a 12 PM 2021 and a 17 PM last month, effectively 2026. Always access to the latest software, and outside of compute-heavy Javascript web apps, it’s been day one to day last snappyness, with a still-not-nearly-low-enough-nevertheless-much-better-than-Android exposure to ads, tracking, and outright malware.

Would that have been possible on Android at all, set aside sticking to a particular brand?

The top Chinese brands seem to compete with Apple for the consumers who upgrade every year or two, but for the consumers who buy long term use phones, what Android brand has comparable trust to Apple?

 

[mikegg]

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

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

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

The same reason why most Americans prefer iPhone.

It works better. Easier to use than Android. My parents know how to use an iPhone. Hardware is better. More updates. Longer lasting. For some people, it signals more wealth. More resale value too.

 

[jpiniero]

The same reason why most Americans prefer iPhone.

Sure it's not people being shamed for being one of those dirty Green Bubble people?

 

[misuspita]

One of my favorite Youtubers visited Institute of Advanced Study at Princeton recently. He basically said all the professors there agree that AI is incredibly useful and is completely changing their field, including people like Ed Witten. Worth watching.

I have the same experience as him. AI has already made it such that I no longer write code. I'm a software engineer.

I've listened to about half of it and I don't think the opinions are canceling each other. Ok, for science, AI is a boon and it already discovers a lot of new stuff, because it has a much larger memory than any human and can do intelligent associations based on that data it has been trained on. But the problem that he touches at some point is exactly of cost. The 20€ subscription doesn't get enough revenue to recoup investments. The deployment of gigawatts of AI data centers is not possible at the levels needed for the general population. There was a nice calculation that I have seen in the video I posted, it was something like Nvidia already sold a number of gpus that would consume many GW of power, GW that are already more than what the infrastructure can support. And there will be more produced, and the question is about sustainability of the AI model. Just as the dotcom bubble, the ideas were great, the implementation was botched.

 

[mikegg]

I've listened to about half of it and I don't think the opinions are canceling each other. Ok, for science, AI is a boon and it already discovers a lot of new stuff, because it has a much larger memory than any human and can do intelligent associations based on that data it has been trained on. But the problem that he touches at some point is exactly of cost. The 20€ subscription doesn't get enough revenue to recoup investments. The deployment of gigawatts of AI data centers is not possible at the levels needed for the general population. There was a nice calculation that I have seen in the video I posted, it was something like Nvidia already sold a number of gpus that would consume many GW of power, GW that are already more than what the infrastructure can support. And there will be more produced, and the question is about sustainability of the AI model. Just as the dotcom bubble, the ideas were great, the implementation was botched.

I'm paying $20 for Copilot + $20 for OpenAI Codex today and it writes 90% of my code. For $40 total.

I'd gladly pay $1,000 for it. Easily. As many people would. Companies wouldn't hesitate to pay that much given the insane productivity boost.

Don't worry about AI companies making revenue. It'll come.

OpenAI might end up at $80b ARR this year if they grow 4x again.

 

[misuspita]

I have no use for AI except menial tasks, so I don't pay. But I will watch the development, but I still agree that while the tech is useful, it can't overcome the explosive expanditure needed and some expensive bubbles will burst

 

[mikegg]

I have no use for AI except menial tasks, so I don't pay. But I will watch the development, but I still agree that while the tech is useful, it can't overcome the explosive expanditure needed and some expensive bubbles will burst

It's weird how you're so confident that the AI returns won't justified investment when you don't have any use for AI.

Meanwhile, our entire software dev team no longer writes code.

 

[poke01]

So not useless.

I don’t use meta apps. So unlike Google I find their advertising business useless.

Hey if you like what Zuck does and how he influences world elections and the corruption go for it. Meta is the one company I never brought hardware from and will never do.

 

[Grabo]

I'm paying $20 for Copilot + $20 for OpenAI Codex today and it writes 90% of my code. For $40 total.

I'd gladly pay $1,000 for it. Easily. As many people would. Companies wouldn't hesitate to pay that much given the insane productivity boost.

Don't worry about AI companies making revenue. It'll come.

OpenAI might end up at $80b ARR this year if they grow 4x again.

Do you review code 90% of the time then?
And if Codex writes unit tests and executes and seldom needs reviewing at all, what do you do? (I find that a bit unlikely though)

Personally reviewing others' code is my duty sometimes but it isn't the most fun part of development

I look forward to the Mac Mini m5. I just might buy one mostly for a local llm for dev work. It's cool that some of the models are publicly available for that purpose.
Am still trying to find a balance, but it does help when they can actually grasp a framework from a codebase. Autocomplete is all but useless to me, disrupts the train of thought.

 

[name99]

I've listened to about half of it and I don't think the opinions are canceling each other. Ok, for science, AI is a boon and it already discovers a lot of new stuff, because it has a much larger memory than any human and can do intelligent associations based on that data it has been trained on. But the problem that he touches at some point is exactly of cost. The 20€ subscription doesn't get enough revenue to recoup investments. The deployment of gigawatts of AI data centers is not possible at the levels needed for the general population. There was a nice calculation that I have seen in the video I posted, it was something like Nvidia already sold a number of gpus that would consume many GW of power, GW that are already more than what the infrastructure can support. And there will be more produced, and the question is about sustainability of the AI model. Just as the dotcom bubble, the ideas were great, the implementation was botched.

1. People pay $100 a month for a cell phone subscription, or for streaming video. (Not all people, sure, but plenty.)
If AI becomes as essential as "Google" (ie the combination of being able to look up data whenever you want, and be entertained as you want, and be kept "up to date" as you want) there's no reason to believe that it won't be worth $100 to "normal" people, let alone professionals.

2. What "is" "AI"? Right now for most people it's chatbots, and if you don't need a chatbot (that's not how you like to learn things, you don't feel a desire for computer emotionality, etc) you don't the value.
But AI is already on the trajectory to do much more.
One huge example that's barely explored is the matching problem. I want to buy things and people want to sell me things. I want to watch movies, and production companies want me to watch their movies. I want a job, and someone wants to hire me. etc etc. Right now matching is mostly done poorly via crude instruments like broadcast advertising, or politicized HR departments using lousy proxies for capabilities. People pay substantial money for personalized high quality matching (eg in love life, or headhunters for C-suite jobs, or buying a house). If AI can give that personalized high quality matching, it will also be able to extract a percent or two or the economic value of every such match performed.

3. Agents. Most people are currently (and justifiably!) scared of handing over any sort of serious control of tasks to AI. Just like they were scared of handing over their credit cards to random internet companies in the early 1990s. But as experience grows and the technology matures, this will change.
Right now the idea of having to book a hotel fills me with dread; yes it's not hard, but it will be 30 to 60 minutes of time dealing with Kayak. I certainly don't want to give a random AI "agent" that task today, just like I wouldn't hire a random person to book me a hotel – both have no idea what I want. BUT an agent that lives on my phone and learns my preferences is more like a trusted secretary who will know that I care about WiFi but not about a gym or a pool, who knows my price range, who knows I want to be near the airport or train station, etc.
Think of an even simpler version of this - smart form fill-in. How many hours a year do you waste just filling in stupid material like frequent flyer numbers or addresses or next of kin? The Semantic Web was supposed to fix this but that seems to have bogged down in theological disputes and once certain people concluded it couldn't be made perfect, they decided they didn't want to make it merely better than what we already have. So we have a few well-known fields (phone number, address) that browsers (usually...) auto-fill and most that they don't. An AI agent could handle this sort of nonsense.

Just like with the internet, just like with mobile, there's a certain class of person that just loves to project negativity onto everything. This seems to be equal parts personality, lack of vision, and lack of knowledge of history. (Of course these are the exact same people who, ten years later, consider it horribly unfair that others got rich from the gold rush whereas they did not...)
OF COURSE there will be plenty of dumb ideas in the early days – just like there were with the internet, just like there were with mobile. That's how we collectively learn! By trying out multiple ideas.

 

[dr1337]

Local AIs are the end of the line and the cloud companies know this, thats why they're in a mad rush to expand as much as possible right now.

Even if DDR6 is only 50% more bandwidth than DDR5, thats still a massive speedup for AI workloads. M3 ultra already has enough bandwidth for 27b models to absolutely sing. And while that might be a $10k machine today, 4 people paying $40 a month for coding AIs collectively spend $2k a year on their subscriptions. Anyone concerned with long term economics sees the obvious writing on the wall here.

Between all the context length and token limits I don't think anyone using the cheaper plans was actually doing much work to begin with, and will likely be replaced by AI themselves once their bosses learn how to prompt. Currently these cheaper plans are super easy to over-saturate even on a simpler project like a browser based snake clone.

You could fuss around with cloud based AI services, or you could buy a piece of hardware that does the same thing and the context limits, system prompts, hand tunings, ect. are decided by you. More expensive up front but cheaper over time, as hardware gets better, this accelerates.

Speaking of writing on the wall, I think this is partly why apple has sided with openAI as their provider. Cloud AI is in a bubble until consumer hardware goes to next generation or two. Its a bet that apple can ride out the current bubble and still wind up on top once it pops. Not gonna make the mobile.me mistake again

 

[Doug S]

Are they better? I know certain brands have started promising 7 years of updates, but has it played out yet for anyone?

I got an iPhone 6S in 2016, a 12 PM 2021 and a 17 PM last month, effectively 2026. Always access to the latest software, and outside of compute-heavy Javascript web apps, it’s been day one to day last snappyness, with a still-not-nearly-low-enough-nevertheless-much-better-than-Android exposure to ads, tracking, and outright malware.

Would that have been possible on Android at all, set aside sticking to a particular brand?

The top Chinese brands seem to compete with Apple for the consumers who upgrade every year or two, but for the consumers who buy long term use phones, what Android brand has comparable trust to Apple?

Heck Apple just released a patch for iOS 12, the last version for the 5S that was released in 2013! Sure people buying new phones don't care about that (no one who is buying new is planning on keeping their phone for 13 years, heck very few do that with cars) but it influences your trade in value because that makes it worth more to the second owner, because that makes it worth more to the third owner. Maybe even a fourth owner when you're talking 13 year old phones.

Having a vibrant second/third/fourth hand market means iPhone's share of the installed base is significantly higher than its market share. Which means there are more iPhones around which makes it a more attractive target for developers which makes people more likely to choose them when buying new.

Other than a few brand name flagships like Galaxy S, Android phones have little value in the second hand market - because those second hand phones are competing with brand new cheaper models. So unless you buy one of those few brand names having little value for trade in means used Androids get stuck in a drawer or recycled rather than living a long life with multiple owners.

 

[fastandfurious6]

-

so many words without substance

1. It works better.
2. Easier to use than Android.
3. My parents know how to use an iPhone.
4. Hardware is better.
5. More updates.
6. Longer lasting.
7. For some people, it signals more wealth. More resale value too.

1. Many claim the opposite / subjective
2. Many claim the opposite / subjective
3. Many claim the opposite / subjective
4. Almost same hardware tbh
5. What updates? Who cares?
6. Many claim the opposite / subjective
7. Can signal more 'wealth' , more resale


The only meaningful answer I got here is "money"... but what is the reason?

Still no answers

 

[poke01]

The only meaningful answer I got here is "money"... but what is the reason?

because there is only one company selling a phone that runs iOS. That’s it and the 17 lineup was pretty good too

 

[jdubs03]

Just today they released updates for products that have been legacy for years. As of now, they still have the latest iOS on a phone from six years prior. As a point of reference my old galaxy S8 plus only had updates for about four years. That additional time does matter to people. I have multiple family members and friends that have iPhones that are before the 13 and they’ll probably keep it until they can’t update their operating system any longer.

 

[oak8292]

so many words without substance

1. Many claim the opposite / subjective
2. Many claim the opposite / subjective
3. Many claim the opposite / subjective
4. Almost same hardware tbh
5. What updates? Who cares?
6. Many claim the opposite / subjective
7. Can signal more 'wealth' , more resale


The only meaningful answer I got here is "money"... but what is the reason?

Still no answers

Business model is an advantage in owning the whole stack. The CPU, hardware and OS are making money and there is margin to optimize the OS to the hardware and keep it updated.

Many Android phone vendors have very low margins and don’t have the money or desire to keep updating to the latest OS for compatibility. Google is not going to test their Android updates against every vendors hardware. Security slips. The more you pay in margin the more likely you will get updates. Some Samsung phones aren’t cheap but the number of models increases the amount of testing required for each update.

Many in this forum will say that the ‘walled garden’ is a problem but there are a lot of users that may view it as a positive.

 

[fastandfurious6]

Umm have you seen iPhone 17 sales? Sold 17 million in 4 months just in China alone.
View attachment 137761

This graph is very misleading.

1. 2025 sales in mainland China, apple was #3:
huawei+oppo+vivo+xiaomi = ~180million
apple iphone = ~45.9million

Just from this, >75% prefer android/huawei

2. iphone 17 has only minor upgrades to iphone 16

Then why is iphone selling so higher?

Business model is an advantage in owning the whole stack. The CPU, hardware and OS are making money and there is margin to optimize the OS to the hardware and keep it updated.

Fully agree and makes sense for real computers but what optimizations are we talking about exactly for phones? Phone CPUs are x100 stronger than needed, even if OS is unoptimized it will still be snappy

Many Android phone vendors have very low margins and don’t have the money or desire to keep updating to the latest OS for compatibility. Google is not going to test their Android updates against every vendors hardware. Security slips. The more you pay in margin the more likely you will get updates. Some Samsung phones aren’t cheap but the number of models increases the amount of testing required for each update.

Literally 99% of phone users in the entire world don't care and don't even know anything about this topic

For 99% of people, Android 11 is as good as Android 18, can't tell the difference

 

[poke01]

This graph is very misleading.

1. 2025 sales in mainland China, apple was #3:
huawei+oppo+vivo+xiaomi = ~180million
apple iphone = ~45.9million

Just from this, >75% prefer android/huawei

2. iphone 17 has only minor upgrades to iphone 16

Then why is iphone selling so higher?

It’s not misleading. Ice Universe is post about high end flagship phones. Huawei sells mid range phones which made it the number one overall in China. But the in the premium space Apple is way ahead

 

[poke01]

Literally 99% of phone users in the entire world don't care and don't even know anything about this topic

They do care especially in the EU.

 

[fastandfurious6]

So basically iphone sells as a status symbol and has millions of instant purchases of newest gen by rich people who are just 'buying the new iphone', $1000 is pocket change

The reason is money + no reason at all

They do care especially in the EU.

Specify? Realistically 95% of people in world really don't care and don't have a reason to care about android updates