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

 

[Eug]

Mac Pro M2 Extreme? No, probably not.

 

[Ajay]

Mac Pro M2 Extreme? No, probably not.

Mac Pro upgrade may need to wait until max memory can be doubled. Or not - who fricking understands Apple.

 

[Doug S]

If it's N3B, then it would make sense to use the A17 cores (reuse same functional units). If, but some strange miracle, Apple is getting enough early wafers of N3E to do a limited launch - then I would still think they would be essentially the same A17 cores (probably bug fixes, maybe tweaks for high clocks), but everything would have had to be resynthesized using the N3E PDK.

Given that N3E is about to enter formal mass production any day now Apple could easily have had enough risk production wafers run if that's what they were using. Since TSMC wants to forget N3B like a bad dream they'd have incentive to cooperate in this plan to avoid keeping N3B around any longer than it has to. But I think if we see anything M3 related tomorrow it will be N3B...that's the path of least resistance.

The real question is, if this is actually M3 then if they're doing MBP they would have to be launching M3, M3 Pro and M3 Max at the same time. That's very different from previous launches where the Pro/Max followed months after the base model. There's nothing stopping them from having them all at the same time but for whatever reason they didn't do so with M1 & M2 so I'd be a little surprised if they change that path for M3.

That departure would be doubly confusing since the yields of N3B are bad enough even making ~100 mm^2 A17s. Imagine how bad the yields must be with a ~400 mm^2 M3 Max! Though I suppose if Apple is paying for "known good die" that's TSMC's problem? But would TSMC make that deal for chips that big? It is one thing to do it for iPhone SoCs, another to do it for chips in the neighborhood of half reticle sized. They might say "sorry you got pay per wafer on those or wait for N3E"

 

[Thibsie]

Interesting that you should mention this.

My two kids both have iPad Airs 2s, and I tried getting Bluetooth case keyboards for them for use as a laptop replacement. This failed miserably, because the keyboards suck, and the experience isn't good. We have "full-size" (18 mm key spacing) Apple Smart Keyboards (so no Bluetooth pairing or battery charging required), on my wife's iPad 7 (2019) and my iPad Pro 10.5" (2017). However, ultimately, the kids wanted their own "real" computers, which work better for some educational apps and other apps. So, now one kid has a 2015 MacBook Pro (running Monterey) I got for cheap off eBay, and the other kid has my old 2010 27" iMac (running High Sierra) which I had previously been using just as an external monitor. Actually, for the iMac, I got a Logitech keyboard that can Bluetooth pair with multiple machines. It has function keys to select which device to use it with and works seamlessly amongst multiple devices. I have it paired with the iMac as the primary device, and the iPad as the secondary device, selectable with a touch of a button. However, it only ever gets used with the iMac. He just uses the on-screen keyboard for the iPad.

C'mon, there hasn't been any acceptable keyboard from Apple for decades.
Any Logitech Apple targeted proper keyboard is a million times better. Well just a PC keyboard with keys maps for Mac.
Hello Apple !
Typical case where shiny is more important than function. I hate that part of Apple.

 

[Thibsie]

What?

Apple just doesn't want Qualcomm to steal the show. Even if it means paper launch.

 

[poke01]

Apple just doesn't want Qualcomm to steal the show. Even if it means paper launch.

Qualcomms announcement is a paper launch. It's not released till mid 2024

 

[Thibsie]

Qualcomms announcement is a paper launch. It's not released till mid 2024

I completely agree but so is Apple's ego....

 

[Ajay]

Given that N3E is about to enter formal mass production any day now Apple could easily have had enough risk production wafers run if that's what they were using. Since TSMC wants to forget N3B like a bad dream they'd have incentive to cooperate in this plan to avoid keeping N3B around any longer than it has to. But I think if we see anything M3 related tomorrow it will be N3B...that's the path of least resistance.

The real question is, if this is actually M3 then if they're doing MBP they would have to be launching M3, M3 Pro and M3 Max at the same time. That's very different from previous launches where the Pro/Max followed months after the base model. There's nothing stopping them from having them all at the same time but for whatever reason they didn't do so with M1 & M2 so I'd be a little surprised if they change that path for M3.

That departure would be doubly confusing since the yields of N3B are bad enough even making ~100 mm^2 A17s. Imagine how bad the yields must be with a ~400 mm^2 M3 Max! Though I suppose if Apple is paying for "known good die" that's TSMC's problem? But would TSMC make that deal for chips that big? It is one thing to do it for iPhone SoCs, another to do it for chips in the neighborhood of half reticle sized. They might say "sorry you got pay per wafer on those or wait for N3E"

Apple has used risk production to start building up a chip inventory b/4 - only with phones, still there is a precedent. If they are using large M3 SoCs built on N3E, that would explain the projected initial low volumes. I have a curious thought, what if Apple is only producing Pro and Max M3s and will introduce the low end latter. Given the 'scary fast' title of the presser - that might make sense.

 

[Eug]

Apple has used risk production to start building up a chip inventory b/4 - only with phones, still there is a precedent. If they are using large M3 SoCs built on N3E, that would explain the projected initial low volumes. I have a curious thought, what if Apple is only producing Pro and Max M3s and will introduce the low end latter. Given the 'scary fast' title of the presser - that might make sense.

Haha, if true, that would go back to my contention a long time ago that N3E M3 series chips could come out Q3 2023 for a Q4 Mac release, but everyone here convinced me otherwise.

However, while Kuo and Gurman are both (now) saying the M3 Pro / Max MacBook Pros are slated to be released next week, they are also suggesting that the 13" M3 MBP is also in the pipeline, and the M3 iMac is also slated for release soon.

OTOH, both of those guys just last week said there were no Macs coming at all in Q4 2023. Everyone is flying by the seat of their pants with regards to M3 series Mac releases predictions.

P.S. I'll bring up this post again:

Page 185 - Discussion - Apple Silicon SoC thread

Page 185 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

This is ancient history, but it still bugs me.

Back in 2000 I told everyone that I believed Apple would release a 12" G4 PowerBook. Nobody believed me. And then in November 2000 Apple released... yet another 15" G4 PowerBook. No 12" G4 PowerBook to be seen. I was very disappointed because I needed a G4 to upgrade from my 12" G3 iBook, and I loved that 12" form factor. However, since I needed the G4, I bought the 15" PowerBook.

Two months later in January 2001, they released the 12" PowerBook. Arrggh!

Apple's releases are hard to predict, no matter how much we think we know about them.

 

[Doug S]

However, while Kuo and Gurman are both (now) saying the M3 Pro / Max MacBook Pros are slated to be released next week, they are also suggesting that the 13" M3 MBP is also in the pipeline, and the M3 iMac is also slated for release soon.

OTOH, both of those guys just last week said there were no Macs coming at all in Q4 2023. Everyone is flying by the seat of their pants with regards to M3 series Mac releases.

Well I guess that shows that Apple is still able to keep some secrets - if the two guys who have been most "in the know" about Apple's plans in advance over the last few years were caught so flat footed by Apple's event announcement. They've been forced to scramble and guess based on what little information Apple provided, just like the rest of us.

It would be doubly funny if it was something completely unexpected, like if "scary fast" referred to adding wifi 7 or TB 5/USB 4v2 rather than CPU speed to updated models (though IMHO they would need something else like brand new displays or revamped design to justify a product reveal event)

I'm not predicting that, M3 is the most obvious explanation. I just think it would be funny if Apple started messing with people deliberately trying to throw them off with their announcements and make the guys who make a living off "predictions" look foolish.

 

[Mopetar]

I don't think Apple wants them to look foolish. They probably want them under thumb enough to "leak" things that Apple might want them to for whatever reason.

Being able to subvert independent journalism and make it a PR apparatus of the company is probably beneficial.

 

[igor_kavinski]

It would be doubly funny if it was something completely unexpected, like if "scary fast" referred to adding wifi 7 or TB 5/USB 4v2 rather than CPU speed to updated models

My guess, crazy fast IO, specifically PCIe 5.0 storage speeds and 8GB or more of CXL RAM, to use as a cache to reduce direct writes to the SSD. Improving that bottleneck would let the compute part work even faster without any upgrades to M3. Video/photo editing etc. are all more I/O bound.

 

[Mopetar]

If Apple wants really fast memory, they should just jump on the HBM bandwagon. Sure it's more expensive, but they're big enough that they could probably make a big down payment to get a volume discount.

They really only need it for the Pro line of products so regular M3 SoCs can still use the cheaper DDR memory. For the Pro line it probably cuts down on die size and offsets some of the cost of having a massive bus to be able to keep the chip fed without using high-speed GDDRX memory.

 

[Glo.]

If Apple wants really fast memory, they should just jump on the HBM bandwagon. Sure it's more expensive, but they're big enough that they could probably make a big down payment to get a volume discount.

They really only need it for the Pro line of products so regular M3 SoCs can still use the cheaper DDR memory. For the Pro line it probably cuts down on die size and offsets some of the cost of having a massive bus to be able to keep the chip fed without using high-speed GDDRX memory.

12 GB of RAM, 36 and 48 GB of RAM on M3, M3 Pro CTO config and base M3 Max config suggests that Apple will simply increase the memory bus width.

 

[roger_k]

If Apple wants really fast memory, they should just jump on the HBM bandwagon. Sure it's more expensive, but they're big enough that they could probably make a big down payment to get a volume discount.

They also need the low power, and I doubt that they can get that with HBM. Also, doesn't HBM rely on wider bus anyway for performance? It seems to me that Apple build themselves a "poor man's HBM" with their approach anyway.

BTW, I am very confident that we will see ECC in the upcoming prosumer Macs. They have published a bunch of patents related to ECC LPDDR5X in the last few months, and it's very typical for Apple to get new patents our just before a product releases.

 

[mikegg]

As some of us originally said years ago, sale prices don't count. Sale prices on Apple products at Amazon and other retailers have existed for eons. We're talking consumer retail (and consumer edu) pricing here.

Also, using inflation calculations to justify your 2020 argument three years later in 2023 is just a cheat.

Apple to Sell Low-Cost 12-Inch and 13-inch MacBooks for $700 or Less

Apple is actively developing new 12-inch and 13-inch MacBook models for sale at a planned price point of around $700 or less, claims a rumor out...

www.macrumors.com

Still not convinced? My 3 year old prediction is spot on.

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?

New 'Scary Fast' Macs May Launch on Wednesday, November 8

Apple may be planning to launch new Mac models announced at its "Scary Fast" event on Wednesday, November 8, MacRumors has learned. At the...

www.macrumors.com

Wrong again.

 

[Eug]

Apple to Sell Low-Cost 12-Inch and 13-inch MacBooks for $700 or Less

Apple is actively developing new 12-inch and 13-inch MacBook models for sale at a planned price point of around $700 or less, claims a rumor out...

www.macrumors.com

Still not convinced? My 3 year old prediction is spot on.

You’re still grasping. Before you were using sale prices or inflation fudges, and now you’re quoting a sketchy rumour as gospel.

As others have said repeatedly, it doesn’t exist until it does.

New 'Scary Fast' Macs May Launch on Wednesday, November 8

Apple may be planning to launch new Mac models announced at its "Scary Fast" event on Wednesday, November 8, MacRumors has learned. At the...

www.macrumors.com

Wrong again.

Probably, but I already posted about that myself several days ago in this thread. In fact, that’s what most of this thread has been about for days. You’re behind in this thread.

BTW, I decided to look up those old posts from 2020. Here's my statement from back then:

Page 38 - Discussion - Apple Silicon SoC thread

Page 38 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

I will go on record again to say I expect we may get an $899 MacBook Air eventually, and we won't get a $699 MacBook Air in the next few years.

I also think $799 is a remote possibility, but that's a whopping 14% more than your original $700 statement. That's quite the revised goalpost there if now all of a sudden you're saying you actually mean $800.

And this one before that:

Page 38 - Discussion - Apple Silicon SoC thread

Page 38 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

I am 100% sure Apple will not have Mac laptops at regular retail pricing of $699 or less in 2022 or earlier.

If you're talking 2023 or later, I'm more hesitant to guess, although I do see them dropping prices (and I've said it was a possibility before too). However, I still very skeptical of a $699 price point. My personal guess is $899 regular retail.

$699 did not happen by 2022, and it turns out we are still at $999 regular retail in 2023 for that same M1 MacBook Air 8 GB / 256 GB , which is now 3 years old. Maybe prices will change next week, but not to $699.

 

[SteinFG]

Didn't read the whole topic, but can the rumored cheap macbook use A17 Pro or similar? It already has 4 performance cores, 8GB of ram, and at least one USB-C.

 

[Thibsie]

USB-C means nothing, it's just a plug.

 

[Doug S]

Didn't read the whole topic, but can the rumored cheap macbook use A17 Pro or similar? It already has 4 performance cores, 8GB of ram, and at least one USB-C.

If they wanted to build a "cheap" Macbook using an iPhone SoC they'd probably rather wait for N3E.

 

[ashFTW]

Didn't read the whole topic, but can the rumored cheap macbook use A17 Pro or similar? It already has 4 performance cores, 8GB of ram, and at least one USB-C.

A17 pro has 4 efficiency, and only 2 performance cores. But that could still be plenty for a low end MacBook. Or maybe Apple will prefer M1, once rest of the lineup is on M3+

 

[Eug]

If they wanted to build a "cheap" Macbook using an iPhone SoC they'd probably rather wait for N3E.

An iPhone class SoC is what I predicted for the entry level MacBook back in 2020, since even back then it was fast enough. However, as we all know, I was wrong.

Meanwhile, even the iPad Air gets M series chips.

 

[eek2121]

An iPhone class SoC is what I predicted for the entry level MacBook back in 2020, since even back then it was fast enough. However, as we all know, I was wrong.

Meanwhile, even the iPad Air gets M series chips.

Which is definitely an odd move by Apple. The iPhone SoCs certainly cost less to make. I can see economies of scale helping drive down M1 costs, but product segmentation is a successful marketing strategy.

Based on benchmarks, I suspect that for typical users, the chips would perform too similarly, so Apple decided to use the M* chips to drive up ASP.

 

[Doug S]

Which is definitely an odd move by Apple. The iPhone SoCs certainly cost less to make. I can see economies of scale helping drive down M1 costs, but product segmentation is a successful marketing strategy.

Based on benchmarks, I suspect that for typical users, the chips would perform too similarly, so Apple decided to use the M* chips to drive up ASP.

I doubt there's all that much difference cost wise. If the die is 50% larger then it is a bit more than 50% more expensive at standard TSMC yields. Based on N5 wafer pricing at least (we have no idea what they are paying for N3B under the KGD deal) that's less than $30 more for the CPU. Obviously in both cases the LPDDR5 and packaging are extra.

I didn't realize they were using M1 for the iPad Air, I thought use of the M in the iPad line was Pro only. M2 is made on a fairly mature node so despite being larger it may be a wash price wise versus A17P (whether under N3B KGD pricing or future by the wafer N3E pricing) so M2 probably makes more sense than A17P.

 

[ashFTW]

An iPhone class SoC is what I predicted for the entry level MacBook back in 2020, since even back then it was fast enough. However, as we all know, I was wrong.

Meanwhile, even the iPad Air gets M series chips.

My only concern with A class chips in low-end laptops is whether it has the needed I/O and display capabilities. For example can it drive an additional external monitor? Or support 2 USB ports (1 for power)? of course, Apple could decide not to support these ..