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

 

[FlameTail]

Oh wow, three different panel suppliers. Yay for panel lottery? Good thing I won't be getting subjected to that after spending a lot of money. Don't care about anything Apple.

Usually, Apple is very careful to avoid "lotteries" when they are multi-sourcing. All the suppliers must meet Apple's standards or they won't get to supply for Apple. BOE once got slapped in the wrist for cost cutting on the copper wire thickness in iPhone display panels.

 

[Doug S]

For those following this, what's so special about these OLED displays over regular OLED that people will just "have to" have one.

Reportedly the main reason Apple wants them is they have a lifetime 4x longer than single stack OLEDs. They feel that's important for tablets versus phones due to the relative amount of screen time and expected device usage life. Lifetime not talking about "burn in" but talking about color accuracy. Apple is super picky about that in a way no one else is, they want something to have the same color accuracy at the end of its useful life that it had at the beginning.

Obviously they have better stats than me but I would figure the useful life of an iPhone and iPad are about the same. People may replace iPhones more often than tablets/PCs but iPhones get second, third maybe even fourth owners. I suppose an iPad Pro is likely to be have a lot more screen on time on average than an iPhone though (obviously with outliers on both sides so spare me the "I'm on my phone 12 hours a day but only touch my tablet twice a month")

 

[soresu]

Dual-stack OLEDs can go brighter and yet at the same time have lower risk of burn-in. For people like me, dual-stack OLED may be overkill, but it seems these are targeted at the true creative professionals, where this would actually matter. Related to this, without dual-stack OLED, going from a mini-LED iPad Pro to an OLED iPad Pro could be a big step down in terms of brightness. The other aspect to this is that dual-stack OLED actually uses less power than regular single-stack LED for bright images.

So, dual-stack is brighter, longer lasting, and lower power than conventional single-stack, but way, way more expensive and up until now, was only ever produced in low volumes.

Blue PHOLED does this anyway for QD OLED displays that are exclusively blue OLED pixels (with QD's for color conversion to red/green).

Shifting to lower power dual stack + blue PHOLED should yield a pretty meaty increase in lifetime and peak brightness.

Though for either W-OLED or full AMOLED display configurations the benefit brought by blue PHOLED pixels won't be anywhere near as great, because red and green PHOLED pixels have been in use for years already in these display types.

So you are only getting power and/or brightness improvements for 1/3 of the subpixels in each whole pixel.

I wonder if LG will switch from W-OLED to QD OLED as the blue PHOLED comes in, and mix it with their MLA (micro lens array) for premium displays/TVs - that would be ideal, albeit likely expensive to produce.

 

[FlameTail]

Guys... I have a hunch that the Apple M4 will adopt a 4P+6E configuration for the CPU.

 

[Mopetar]

Reportedly the main reason Apple wants them is they have a lifetime 4x longer than single stack OLEDs.

I could definitely use something like that since I have my iPad running most of the day. Mine is an older model from when they still used LCD displays, but I probably average 10 hours a day with it as even if I'm not using it for something specific, I'll just throw a YouTube video or something like that on as background noise while I'm working.

Even at home I'll put on some Netflix show or whatnot while cooking and doing chores.

 

[igor_kavinski]

Even at home I'll put on some Netflix show or whatnot while cooking and doing chores.

Your distractions need an iPad? Talk about high standards!

 

[Doug S]

I could definitely use something like that since I have my iPad running most of the day. Mine is an older model from when they still used LCD displays, but I probably average 10 hours a day with it as even if I'm not using it for something specific, I'll just throw a YouTube video or something like that on as background noise while I'm working.

Even at home I'll put on some Netflix show or whatnot while cooking and doing chores.

But with that usage do you really care if the colors are still as accurate 7 years later as the day you bought it? You might not want to pay the extra to get the iPad Pro with the fancy OLED if you aren't doing something where color accuracy or Pro level performance is super important.

 

[eek2121]

Sounds like I may need to get a new iPad when they come out.

I wish they would create a pro variant of the mini.

The thing everyone is missing with regards to important mac features is the one that matters the most to people who use them in a professional context; they are EXACTLY as performant on battery as they are on the wall, and not in a bad way. If you have a macbook pro, you can expect to tote it around and have full fat performance all day until your battery dies. There isn't a windows laptop around that can compete like that, not until the ARM devices get good.

While Macbooks are most certainly more efficient, you are overstating claims. None of the Mx Macbooks I have used last all day. I get 4-6 hours on a charge. They are much faster, however.

Then again, I actually have a workload that loads many cores.

Also PCs throttling when unplugged is a software thing you can disable.

I am actually going to be experimenting with that and limiting CPU power to 15-35w on my Cezanne-based laptop. (It has a 3070 that I also plan to cap at a lower level, I want to see how low I can go while still getting decent FPS)

Oh and if you use anything that puts a load on the GPU the laptop dies quickly. Same with the iPad. I play games on my iPad and depending on the game the battery will last anywhere from 2-4 hours.

 

[FlameTail]

New chip flaw hits Apple Silicon and steals cryptographic keys from system cache — 'GoFetch' vulnerability attacks Apple M1, M2, M3 processors, can't be fixed in hardware

This vulnerability is stupendously serious, affecting most encryption algorithms including those hardened against quantum computers.

www.tomshardware.com

"Researchers have discovered a massive security vulnerability inside Apple M1, M2, and M3 silicon. The vulnerability, dubbed 'GoFetch,' steals cryptographic information from the CPU cache enabling an attacking program to build a cryptographic key from stolen data, allowing the application to access sensitive encrypted data. Ars Technica first reported on the security flaw. "

 

[Glo.]

New chip flaw hits Apple Silicon and steals cryptographic keys from system cache — 'GoFetch' vulnerability attacks Apple M1, M2, M3 processors, can't be fixed in hardware

This vulnerability is stupendously serious, affecting most encryption algorithms including those hardened against quantum computers.

www.tomshardware.com

"Researchers have discovered a massive security vulnerability inside Apple M1, M2, and M3 silicon. The vulnerability, dubbed 'GoFetch,' steals cryptographic information from the CPU cache enabling an attacking program to build a cryptographic key from stolen data, allowing the application to access sensitive encrypted data. Ars Technica first reported on the security flaw. "

This will also affect iPads with M series chips.

Also - since the CPU shares the architecture with M series, it also should affect iPhone A series chips.

 

[FlameTail]

Apple is doomed

 

[igor_kavinski]

Except who does top secret sensitive stuff on an Apple device?

 

[Doug S]

Except who does top secret sensitive stuff on an Apple device?

It may not be "top secret sensitive" but I'd prefer not to have someone able to access my bank or brokerage accounts!

Since it depends on uncontrolled code running at the same time as code being attacked, sounds like all you need is to be sure you close other apps when doing anything like accessing your bank account. I have always had the habit of flicking closed all running apps anytime I put my phone down (habit from my first iPhone in the 3gs days when RAM was precious I guess?) so at least now I have an excuse if someone sees me doing it and wonders why I'm doing that. So close everything (but ESPECIALLY Safari) when you are doing anything involving big money I suppose is the best advice for now.

A vulnerability like this is a bigger deal on multiuser systems like servers, or even a desktop where you have someone else logged in at the same time. Since all iPhones/iPads and most Macs have only a single user, you can control whether third party code is running in the background or not to mitigate this issue (assuming a practical attack ever materializes)

 

[igor_kavinski]

Since it depends on uncontrolled code running at the same time as code being attacked, sounds like all you need is to be sure you close other apps when doing anything like accessing your bank account. I have always had the habit of flicking closed all running apps anytime I put my phone down

If the MacOS has it, I would choose Safe Mode with Networking (one of the boot options in Windows). That one usually has only essential services running and hopefully a browser will function fine in that mode. Otherwise, boot from a USB into Live mode of a decent Linux distro.

 

[naukkis]

New chip flaw hits Apple Silicon and steals cryptographic keys from system cache — 'GoFetch' vulnerability attacks Apple M1, M2, M3 processors, can't be fixed in hardware

This vulnerability is stupendously serious, affecting most encryption algorithms including those hardened against quantum computers.

www.tomshardware.com

"Researchers have discovered a massive security vulnerability inside Apple M1, M2, and M3 silicon. The vulnerability, dubbed 'GoFetch,' steals cryptographic information from the CPU cache enabling an attacking program to build a cryptographic key from stolen data, allowing the application to access sensitive encrypted data. Ars Technica first reported on the security flaw. "

The most interesting part of that side-channel is that Apple is using pointer finder in their prefetchers and prefetching suspected pointer locations.

 

[Doug S]

The most interesting part of that side-channel is that Apple is using pointer finder in their prefetchers and prefetching suspected pointer locations.

The other thing I find interesting is that since the E cores don't have that behavior, they aren't vulnerable. For the kind of stuff a banking or brokerage app would do, that's not really an issue because they aren't encrypting a whole lot of data. The app would need to be updated of course, but fortunately for me the ones I use get regular updates.

What I wonder is whether it affects the hardware encryption built into the SoC, not stuff running as instructions on the CPU cores. That's where a lot of the heavy lifting for stuff like bulk data transfer to/from iCloud and (I think?) TLS encryption of web content is done, and going to software would be a hit on both performance and power efficiency.

Definitely waiting to hear more about this and see exactly what the real world effect is. Given the timing I have to wonder if the release of iOS 17.4.1 as well as an iOS 16 and Mac update (without any notes about what was fixed security wise) was not a coincidence. Maybe they are able to mitigate it at least partway, or they have disabled certain encryption algorithms that are vulnerable while others are not (or at least less so)

 

[Doug S]

Saw more information on this, supposedly M3 (and therefore presumably A17P) can disable DMP which eliminates the issue, at the cost of some (likely pretty small) amount of performance. Otherwise software implementations would need to be rewritten to work around this. Apple fixing its own implementations would fix a lot of it, since presumably few apps are rolling their own encryption but preferring to call Apple's APIs. But there are always those who insist on doing it themselves, and you'd have no way of determining which apps those are.

I wonder why Apple added the ability to disable DMP? Did it cause performance issues in some cases or did they become aware that this attack was at least possible?

 

[naukkis]

I wonder why Apple added the ability to disable DMP? Did it cause performance issues in some cases or did they become aware that this attack was at least possible?

Everything can be buggy or just doesn't work so usually cpu's have ability to switch off almost all performance options. But loading data with speculated pointers - why the hell everything has to be general purpose on todays cpus - why ISA doesn't implement different registers for data and addressing - like 68k did. If doing so would make hardware implementations easier and also make hardware differentation between data and pointers. I really don't understand making everything general purpose - it might be good thing sometimes when hardware was very resource limited but why they are still afraid of making hardware segmentation on ISA level?

 

[FlameTail]

Apple A18 Pro Geekbench score leak:

A solid 20% jump in Single-core.

 

[Doug S]

Yay more worthless claims from random nobodies on the internet about silicon that's six months from release. Or is someone going to claim "user 937840" has a track record of accurate predictions?

 

[Apokalupt0]

Apple A18 Pro Geekbench score leak:
View attachment 95865
A solid 20% jump in Single-core.

I stopped believing any such leaks. Last year some similar leak said 30% single core gains.

 

[poke01]

I stopped believing any such leaks. Last year some similar leak said 30% single core gains.

Apple won’t get 10+% improvements for CPU until they move from that A16 cores.
Right now they are in A16+ core architecture.

 

[FlameTail]

They need to make 16 wide core with a 1000+ ROB

 

[SiliconFly]

Yay more worthless claims from random nobodies on the internet about silicon that's six months from release. Or is someone going to claim "user 937840" has a track record of accurate predictions?

If we can't even believe random nobodies on the internet, where is the world headed!

 

[okoroezenwa]

Aren’t the A17 cores a new architecture?