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

 

[Jan Olšan]

That 9% is for total PC shipments, not laptops.

As far as notebooks go, there was a switch from Arm to x86 at some point, so pre-2020 data should be considered with care.

Of course, but given the weight of notebooks (70 %) in the PC mix, it's impossible to have 17% laptop share with 9% overall share.

 

[oak8292]

Of course, but given the weight of notebooks (70 %) in the PC mix, it's impossible to have 17% laptop share with 9% overall share.

Be careful with the definition of a ‘laptop’. This varies among the reporters and there isn’t a bright line between tablets and laptops. Where does a Microsoft tablet fit? Where does an iPad Pro fit? Where does a Chromebook fit.

 

[Nothingness]

Of course, but given the weight of notebooks (70 %) in the PC mix, it's impossible to have 17% laptop share with 9% overall share.

Obtaining data is so difficult, this is all I could find (OK, didn't look very long) and to say the least it's hard to know what laptop vs desktop is

But yeah I agree the 17% looks too high. It's somewhere between 9 and 17, both bounds being excluded.

 

[Mopetar]

They've always talked about "long term" supply agreements, six months is not long term to me so I'd question their wording if that was really the case.

For electronics manufacturing six months is practically an eternity in today's world. Everything is designed around a Just In Time philosophy where nothing is kept around any longer than necessary to save on warehousing costs.

When everything is running smoothly this works great, but as we saw during COVID it's easily disrupted. Even the f it weren't, anything beyond six months has too much uncertainty to reliably plan around unless you manufacture all of your own components on existing production lines.

Most of the time not having to have a warehouse full of memory sitting on shelves because the SoC, screens, or some other component got delayed for a few months is more valuable than eating added costs due to market fluctuations. Even those aren't really a problem if everyone has to eat them at the same time or they can just be passed on to consumers.

 

[oak8292]

For electronics manufacturing six months is practically an eternity in today's world. Everything is designed around a Just In Time philosophy where nothing is kept around any longer than necessary to save on warehousing costs.

When everything is running smoothly this works great, but as we saw during COVID it's easily disrupted. Even the f it weren't, anything beyond six months has too much uncertainty to reliably plan around unless you manufacture all of your own components on existing production lines.

Most of the time not having to have a warehouse full of memory sitting on shelves because the SoC, screens, or some other component got delayed for a few months is more valuable than eating added costs due to market fluctuations. Even those aren't really a problem if everyone has to eat them at the same time or they can just be passed on to consumers.

This was the practice that was essential with Moore’s Law due to obsolescence. It is good practice even without obsolescence. However, it is less essential as performance differences decrease. If Compaq or Apple were caught with a warehouse full of computers in a downturn it wouldn’t be an existential crisis.

If Micron has a lot of DDR4 in a warehouse it is increasing in value at this point versus decreasing. Memory and Moore’s Law was a killer as bits typically decrease in value so rapidly.

The market for old electronics may have a renaissance here in 2026 as they were built with ‘cheap’ parts.

 

[Doug S]

For electronics manufacturing six months is practically an eternity in today's world. Everything is designed around a Just In Time philosophy where nothing is kept around any longer than necessary to save on warehousing costs.

When everything is running smoothly this works great, but as we saw during COVID it's easily disrupted. Even the f it weren't, anything beyond six months has too much uncertainty to reliably plan around unless you manufacture all of your own components on existing production lines.

Most of the time not having to have a warehouse full of memory sitting on shelves because the SoC, screens, or some other component got delayed for a few months is more valuable than eating added costs due to market fluctuations. Even those aren't really a problem if everyone has to eat them at the same time or they can just be passed on to consumers.

Apple wouldn't be buying two years worth of memory for immediate delivery, they'd have it delivered over time! They couldn't have it delivered up front even if they wanted to fill up a giant warehouse, because Micron et al wouldn't have enough to ship them two years worth all at once. The reason Apple started doing these multiyear contracts was to guarantee supply. e.g. back when the iPod switched from 1.8" HDDs to flash, they got prepaid deals with multiple suppliers that amounted to nearly all the NAND capacity at first. Even ignoring black swans like Covid and the AI bubble, having guaranteed supply at a guaranteed price makes costing for upcoming products much easier. Cook is an operations guy, after all, this is the kind of stuff he nerds out over (there was an interview not long after he became CEO where a good chunk of it was him doing exactly that lol)

It also helps a bit with getting better terms. Saying "we want this two year deal and we'll pay for everything up front and you ship it as we need it" is gonna get you better pricing than if you do a two year deal and you pay over time. Especially in capital intensive industries like DRAM/NAND (and merchant foundries like TSMC) having that cash up front so you aren't relying on lines of credit for building out future capacity saves Micron/TSMC/etc. money and makes their financials look better (although prepaid orders show up as a liability on the balance sheet just like a line of credit does, it is evaluated much more positively by Wall Street)

I suppose the one downside is that you won't necessarily get as good of pricing during periods where prices are rapidly falling as might if you bought it in smaller quantities, but Apple isn't exactly pricing their RAM/NAND upgrades at thin margins so they don't care about that. They don't want lowest possible price, they want a long term predictable price. Since DRAM/NAND is always sort of a boom/bust cycle with prices that trend down I imagine some of that low term downward price trend gets built into the contracts, just like a contract being signed now would have the upward spike built into it.

The other thing with long term contracts is that you are probably going to trail the market tech wise, and that's proven very true for Apple as they always lag a bit on LPDDR technology.

 

[fkoehler]

Most of the time not having to have a warehouse full of memory sitting on shelves because the SoC, screens, or some other component got delayed for a few months is more valuable than eating added costs due to market fluctuations.

Can't disagree that this is how companies ran until Covid made an impact.
Now with RAM?

Any CEO in the future who tries to use this an excuse is likely to get dinged hard as his Stock price gets hammered and investors scream for his head.

Expect most industries to return to having something like at least 1QTR worth of inventory in the future. The potential for lost business, stock hammering and business discontinuity makes some warehousing expense a penny save-pound foolish excuse no one will buy.

 

[LightningZ71]

JIT is losing it's luster, but, it's too late for much of the industry as whole supply chains are heavily capacity restricted as manufacturers are loathe to build extra capacity.

 

[name99]

Apple wouldn't be buying two years worth of memory for immediate delivery, they'd have it delivered over time! They couldn't have it delivered up front even if they wanted to fill up a giant warehouse, because Micron et al wouldn't have enough to ship them two years worth all at once. The reason Apple started doing these multiyear contracts was to guarantee supply. e.g. back when the iPod switched from 1.8" HDDs to flash, they got prepaid deals with multiple suppliers that amounted to nearly all the NAND capacity at first. Even ignoring black swans like Covid and the AI bubble, having guaranteed supply at a guaranteed price makes costing for upcoming products much easier. Cook is an operations guy, after all, this is the kind of stuff he nerds out over (there was an interview not long after he became CEO where a good chunk of it was him doing exactly that lol)

It also helps a bit with getting better terms. Saying "we want this two year deal and we'll pay for everything up front and you ship it as we need it" is gonna get you better pricing than if you do a two year deal and you pay over time. Especially in capital intensive industries like DRAM/NAND (and merchant foundries like TSMC) having that cash up front so you aren't relying on lines of credit for building out future capacity saves Micron/TSMC/etc. money and makes their financials look better (although prepaid orders show up as a liability on the balance sheet just like a line of credit does, it is evaluated much more positively by Wall Street)

I suppose the one downside is that you won't necessarily get as good of pricing during periods where prices are rapidly falling as might if you bought it in smaller quantities, but Apple isn't exactly pricing their RAM/NAND upgrades at thin margins so they don't care about that. They don't want lowest possible price, they want a long term predictable price. Since DRAM/NAND is always sort of a boom/bust cycle with prices that trend down I imagine some of that low term downward price trend gets built into the contracts, just like a contract being signed now would have the upward spike built into it.

The other thing with long term contracts is that you are probably going to trail the market tech wise, and that's proven very true for Apple as they always lag a bit on LPDDR technology.

Here's another way to look at this that non-finance people might understand more easily:

Option 1. You pre-arrange contracts for DRAM (and flash, etc) at a fixed price a long time in advance.
That is a fixed price, and the price might go down over the next two years or whatever.

Option 2. You pay the spot DRAM price and buy as you need. This SEEMS like a good idea, especially for high tech items where the price is usually falling.
BUT look at the downside. The downside is not that you pay an extra $10 for some DRAM, it's that you cannot ship an iPhone and thus lose the $400 (or whatever) of profit on that iPhone.
$400>>$10!!!
The fixed-price contract doesn't just avoid price fluctuations; the more important point is that it guarantees you can ship as many iPhones as you need to...