Discussion Apple Silicon SoC thread

Eug · Nov 10, 2020

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

johnsonwax · Aug 23, 2024

Glo. said:
If M3 Pro was stop gap product on a stop gap node - it does not have to be.

I think the original point still holds. M1 Pro was overengineered for the market it serves and made the M1 Max less attractive to customers. Nerfing the Pro a bit improves Apple's upsell potential for the Max.

poke01 · Aug 23, 2024

Pardon my language mods but ***** yes

The four machines have base-level M4 chips, according to developer logs. Three of the Macs have a 10-core CPU and 10-core GPU. The fourth machine has an 8-core CPU and an 8-core GPU, which is not an M4 configuration that we've seen so far. All four of the M4 Macs have either 16GB or 32GB of Unified Memory.

16GB ram base! Shaming companies works.

Apple Testing Four New M4 Macs Ahead of Fall Launch

Apple has ramped up testing of four new Mac models equipped with an M4 chip, according to Bloomberg's Mark Gurman. Apple is planning to refresh...

www.macrumors.com

Doug S · Aug 23, 2024

poke01 said:
16GB ram base! Shaming companies works.

More like they had their schedule based on future contract cost when they'd go 16 GB and reached it. You should start "shaming" them now to make 32 GB the new base, I imagine in 3 or 4 years you'll "succeed"

igor_kavinski · Aug 23, 2024

Surprising that they went for 16GB instead of 12GB. But it's not launched yet so maybe there IS a 12GB base model...

johnsonwax · Aug 23, 2024

poke01 said:
Pardon my language mods but ***** yes

The four machines have base-level M4 chips, according to developer logs. Three of the Macs have a 10-core CPU and 10-core GPU. The fourth machine has an 8-core CPU and an 8-core GPU, which is not an M4 configuration that we've seen so far. All four of the M4 Macs have either 16GB or 32GB of Unified Memory.

16GB ram base! Shaming companies works.

Apple Testing Four New M4 Macs Ahead of Fall Launch

Apple has ramped up testing of four new Mac models equipped with an M4 chip, according to Bloomberg's Mark Gurman. Apple is planning to refresh...

www.macrumors.com

Hmm. A Mac with fewer cores than an iPad. That's a product strategy designed to move certain users off the Mac.

I wouldn't attribute the 16GB to shaming, but to AI. We know RAM is one of the main constraining factors for AI. Maybe not for the models rolling out this year, but it limits what Apple can roll out in future years.

poke01 · Aug 23, 2024

Doug S said:
More like they had their schedule based on future contract cost when they'd go 16 GB and reached it. You should start "shaming" them now to make 32 GB the new base, I imagine in 3 or 4 years you'll "succeed"

True but that Apple interview about RAM last year went viral in western and Chinese social media, marketing probably thought it wasn't worth the hassle this time.

poke01 · Aug 23, 2024

johnsonwax said:
Hmm. A Mac with fewer cores than an iPad. That's a product strategy designed to move certain users off the Mac.

Its 4p + 4e instead of the 3p+6e like on the base iPad pro. I think this 8 core config will be seen in the Mac mini or base iMac.

The 14" MacBook pro m4 will get the full 10 core.

johnsonwax · Aug 23, 2024

igor_kavinski said:
Surprising that they went for 16GB instead of 12GB. But it's not launched yet so maybe there IS a 12GB base model...

It's not a cost consideration, but an upsell one. 8GB RAM costs Apple, what, $6? Tops. I suspect 12GB kills upsell as badly as 16GB and constrains them down the road, so might as well go to 16GB.

johnsonwax · Aug 23, 2024

poke01 said:
Its 4p + 4e instead of the 3p+6e like on the base iPad pro. I think this 8 core config will be seen in the Mac mini or base iMac.

The 14" MacBook pro m4 will get the full 10 core.

But the bigger iPad is 4P + 6E, which is more than the Mac would be. The iPad has been creeping into the Mac space in terms of features for a while - better camera, better screen tech, etc. This would mark the first time it had better CPU, even if just marginally with the extra efficiency cores.

poke01 · Aug 23, 2024

johnsonwax said:
But the bigger iPad is 4P + 6E, which is more than the Mac would be. The iPad has been creeping into the Mac space in terms of features for a while - better camera, better screen tech, etc. This would mark the first time it had better CPU, even if just marginally with the extra efficiency cores.

The MacBook Pros M4 will likely get get the same 4P+6E. Its just segmentation, newer nodes are expensive and its not entirely surprising a Mac that's costs less than a iPad pro will get binned CPUs.

Its also interesting that the iPad pro with M4 is still capped at 16GB. If M4 is capable of up to 32GB of RAM but only for Macs with M4. This just tells me they see Macs as the power platform for AI as RAM is an important factor here too.

jdubs03 · Aug 23, 2024

poke01 said:
The MacBook Pros M4 will likely get get the same 4P+6E. Its just segmentation, newer nodes are expensive and its not entirely surprising a Mac that's costs less than a iPad pro will get binned CPUs.

Its also interesting that the iPad pro with M4 is still capped at 16GB. If M4 is capable of up to 32GB of RAM but only for Macs with M4. This just tells me they see Macs as the power platform for AI as RAM is an important factor here too.

To me their product segmentation with iPad Pro and the low end MacBook Pros just seems weird. Also with the timing of the M4 in the iPad. I know this has probably been hashed out already but it just seems strange how their power platform for work gets the state-of-the-art chip after. Don’t get me wrong, the iPad Pro is a great device, but we all know iPadOS has significant limitations compared to MacOS. Seems like a waste of computing power, relatively. To me it makes more sense to release the most capable devices with their most recent CPU versions, then a few months later release the iPad Pro, along with refreshes of the Air and standard.

poke01 · Aug 23, 2024

jdubs03 said:
To me their product segmentation with iPad Pro and the low end MacBook Pros just seems weird. Also with the timing of the M4 in the iPad. I know this has probably been hashed out already but it just seems strange how their power platform for work gets the state-of-the-art chip after. Don’t get me wrong, the iPad Pro is a great device, but we all know iPadOS has significant limitations compared to MacOS. Seems like a waste of computing power, relatively. To me it makes more sense to release the most capable devices with their most recent CPU versions, then a few months later release the iPad Pro, along with refreshes of the Air and standard.

The M4 debut was strange indeed, all other M chips came out on Mac first. Apple did say they needed the new display engine in the M4 to drive the Tandem OLEDs and I see this being the only reason as too why the m4 came to iPad first and oh iPad sales were down a lot. No iPads were laucnhed in 2023 so they probably released it early.

Doug S · Aug 23, 2024

poke01 said:
The MacBook Pros M4 will likely get get the same 4P+6E. Its just segmentation, newer nodes are expensive and its not entirely surprising a Mac that's costs less than a iPad pro will get binned CPUs.

Its also interesting that the iPad pro with M4 is still capped at 16GB. If M4 is capable of up to 32GB of RAM but only for Macs with M4. This just tells me they see Macs as the power platform for AI as RAM is an important factor here too.

N3E yields are normal for TSMC (i.e. at least 90%) so there's no way Apple needs to bin out M4s going into Macbook Pros. I could see them saving up the few that do have bad cores to go into Macbook Air. That's where they'd want to segment - to keep MBP performance ahead of the less expensive MBA.

name99 · Aug 24, 2024

johnsonwax said:
I think the original point still holds. M1 Pro was overengineered for the market it serves and made the M1 Max less attractive to customers. Nerfing the Pro a bit improves Apple's upsell potential for the Max.

Or to put it in simpler terms, Apple realized that the product split it (and its customers) want is more like i5, i7, and Xeon; not i5, i9KS, and Xeon...

DZero · Aug 24, 2024

I am thinking... I am lost in Apple chips and the M1 in which tier is? I mean in U tier Intel/AMD laptops.
Is on par of Ryzen 7 6th generation U tier Laptops? 12th generation U tier Intel?

poke01 · Aug 24, 2024

DZero said:
I am thinking... I am lost in Apple chips and the M1 in which tier is? I mean in U tier Intel/AMD laptops.
Is on par of Ryzen 7 6th generation U tier Laptops? 12th generation U tier Intel?

Base M chips are i7/i9 U class chips, M Pro chips are H class i7/i9 chips, the M Max chips are i9 HX class chips but much more efficient while delivering the same performance.

Note this is just comparing the CPU performance only. It does not account for efficiency nor the iGPU or the media engines or the IO.

FlameTail · Aug 25, 2024

I came across this interesting comment about x86 vs ARM, SMT and many other topics;

https://forums.macrumors.com/thread...elites-benchmark-scores.2408596/post-33230475

The comment itself is huge, with detailed explanation. I found this section interesting (emphasis mine):

Basically microarchitecture does indeed matter more, much more than architecture (ISA). But this gets at the heart as to why Apple able to design the microarchitecture of a Firestorm core and beat x86 designs so handily in both performance and performance per watt when it came out 4 years ago. The particular reason I'm going to focus on (but it isn't the only one) is that ARM is easier to decode in parallel than x86. Prior to the introduction of Firestorm, x86 was pretty much stuck on 4-wide decode, I even remember an AMD engineer quoting back n the day that they didn't think it was possible to go past that (it is). With an 8-wide decode, Apple could make everything else wider too, caches, number of parallel pipes (the number of different parallel instructions a CPU can perform at once), Reorder Buffer, etc ... similar to what we were just discussing in the GPU section a wider core processing more instruction in parallel, operating at lower frequency and with shorter pipes, that is still smaller in silicon, uses much less power for the same or better performance especially on a single thread. To get more out of longer pipes, x86 designs have to speed up the CPU, in doing so though much of the time parts of an individual pipeline is empty - wasting watts. That's why SMT works so well in current x86 designs, you can interleave instructions from another thread into the same pipe and suddenly the core is now filled with work. But it's also why SMT (currently) doesn't add much for ARM. The cores are designed around getting as much out of shorter pipes as possible but more of them. The architecture aided in the natural development of the microarchitecture and indeed on Twitter, deleted now I think, Shac Ron, basically said ARM and Apple collaborated quite closely on ARM v8's design with future Firestorm-like cores in mind. In other words, the ISA was designed around getting to where we are today.

This would explain why AMD/Intel laptop CPUs have much better multi-thread efficiency relative to their single-thread efficiency, where they get clobbered by Apple/Qualcomm. (Other than the fact that Intel/AMD also have worse uncore/idle power which hurts them in ST efficiency).

Eug · Aug 25, 2024

16 GB RAM base surprises me if true. However, if true, I wouldn’t be surprised if Apple increases base pricing by US$100.

Also, the next step up being 32 GB instead of 24 GB could end up being a disappointment to many, as I’m sure it would be pricey. Apple historically charges US$400 for the 16 to 32 GB upgrade, whereas it’s only $200 for the 16 to 24 GB upgrade.

Here’s hoping Apple will reduce pricing for the 16 to 32 GB upgrade, if 24 GB is not an option.

poke01 · Aug 25, 2024

Eug said:
16 GB RAM base surprises me if true. However, if true, I wouldn’t be surprised if Apple increases base pricing by US$100.

Also, the next step up being 32 GB instead of 24 GB could end up being a disappointment to many, as I’m sure it would be pricey. Apple historically charges US$400 for the 16 to 32 GB upgrade, whereas it’s only $200 for the 16 to 24 GB upgrade.

Here’s hoping Apple will reduce pricing for the 16 to 32 GB upgrade, if 24 GB is not an option.

Its important to see how Apple will market the M4 Macs, if AI the focus. I see the base prices being the same and the 16GB to 32GB upgrade path becoming "cheaper", from $400 to $300.

Who knows its Apple?! And they go bonkers with the pricing lol

name99 · Aug 25, 2024

DZero said:
I am thinking... I am lost in Apple chips and the M1 in which tier is? I mean in U tier Intel/AMD laptops.
Is on par of Ryzen 7 6th generation U tier Laptops? 12th generation U tier Intel?

For what purpose?
IMHO the best match FOR MOST USERS is to look at GB6 single-threaded performance and compare that across your chips of interest:
Here's Apple models:

Mac Benchmarks - Geekbench

Here's x86:

Processor Benchmarks - Geekbench

But if your PRIMARY concern is games you will look at this differently. Likewise if your PRIMARY concern is running Blender.

Just remember that plenty of people will try to persuade you that it's vitally important that you get a machine with, eg, a high Cinebench number, and that is basically PURE redacted!
Utterly irrelevant to the way most technically naive people use their computers.

No profanity allowed in tech.

esquared
Anandtech Forum Director

Glo. · Aug 26, 2024

Eug said:
16 GB RAM base surprises me if true. However, if true, I wouldn’t be surprised if Apple increases base pricing by US$100.

Also, the next step up being 32 GB instead of 24 GB could end up being a disappointment to many, as I’m sure it would be pricey. Apple historically charges US$400 for the 16 to 32 GB upgrade, whereas it’s only $200 for the 16 to 24 GB upgrade.

Here’s hoping Apple will reduce pricing for the 16 to 32 GB upgrade, if 24 GB is not an option.

If 16 GB is base now, then most likely 32 moves to become 200$ upgrade option.

P.S. If 16GB is stadard, then M3 Max moves to 64 GB base if it maintains 512 bit bus.

FlameTail · Aug 26, 2024

Apple Event

Mopetar · Aug 26, 2024

Glowtime huh? Are they having a rave? Do you think they'll hand out glow sticks and ecstasy to attendees or should they plan on bringing their own?

shiznit · Aug 26, 2024

Glo. said:
If 16 GB is base now, then most likely 32 moves to become 200$ upgrade option.

P.S. If 16GB is stadard, then M3 Max moves to 64 GB base if it maintains 512 bit bus.

M3 Max has a 3/4 memory bus option.

I'm looking forward to the M4 Pro, the cut down M3 Pro was initially a letdown but the battery life savings are pretty good.

The Hardcard · Aug 26, 2024

CouncilorIrissa said:
I think that's the price you pay for having a low-power fabric. All laptop SoCs seem to suffer from this no matter the manufacturer (well aside from MTL, but I wouldn't be surprised to see the same in LNL).

Is there a fabric though? Aren’t those cross cluster latencies about exactly double the DRAM latency? I thought the DRAM latency was 90 ns thereabouts.

Discussion Apple Silicon SoC thread

Lifer

Member

Golden Member

Platinum Member

Lifer

Member

Golden Member

Golden Member

Member

Member

Golden Member

Senior member

Golden Member

Platinum Member

Senior member

Member

Golden Member

Diamond Member

Lifer

Golden Member

Senior member

Diamond Member

Diamond Member

Diamond Member

Senior member

Member