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

 

[Leeea]

Apple's choice, TODAY, gives you equal single-threaded performance to the best achievable on the x86 side

oh, fantastic!

Does it run Crysis?

 

[kurosaki]

If they go with 2x2 LP-DDR5 in the package for M1X (or whatever it’s called), doesn’t that mean support for up to 64 GB?

32 GB seems like too low a target to aim for, for the Mac mini and iMac.

64 MacGB's, wow! Wouldnt that be like 10 000 usd?

 

[IntelUser2000]

For example I see that the most recent Apple Intel MBA and MBP run their LPDDR4x at 3733, not 4266.

That's just Icelake being crap. Rocketlake-U before it was canned in favor of Tigerlake was slated to also support LPDDR4x-3733. Rocketlake uses the same foundational Sunny Cove as in Icelake and likely design methodologies.

Lakefield, Tigerlake and Renoir all support 4266 speeds.

 

[amrnuke]

(a) Apple's choice, TODAY, gives you equal single-threaded performance to the best achievable on the x86 side, at substantially lower power.
How is this not a superior result?

(b) AMD are doing OK with the hand they are dealt (which includes their customer base). That's not the same thing as saying that they are on the right side of design history.

(c) The issue is not "brute-forcing a bunch of transistors into a wide core and huge L2$". Saying that demonstrates massive ignorance.

Growing resources blindly buys you very little.
Look at the graphs (esp the first) in this paper:

https://arxiv.org/pdf/1906.08170.pdf

Note how blindly quadrupling resources gets you no more than 1.5x IPC; yet Apple are already at higher than than relative to AMD or Intel.

It's all about how you use your transistors; better algorithms not just larger storage. This is the part that x86 people consistently refuse to even understand, let alone believe, the massive impact of things like better prefetching algorithms, better cache placement and replacement algorithms, better branch prediction algorithms (not just TAGE for most branches, but special case handlers for the various difficult branches). x86 folks are happy to talk about the value of fusion on their side, because they know Intel and AMD engage in it; but they cannot see the extent to which ARMv8 provides a very rich pre-fused instruction set, along with Apple adding a very rich set of dynamic fusions. They can't see the value in Load/Store Pair. They refuse to concede the value of CSEL and other ARMv8 predication (presumably because Intel has made a hash of this every time it implements, so they can't imagine anyone else could do it correctly). And so on and so on.

For example:
look at what IBM do with single instruction branches, from POWER8 on:
https://pdfs.semanticscholar.org/a1e4/f4ae16c5a18896fe1718acfe56a26aeca620.pdf
(figure 4, c)
until you understand why this transformation is worth doing, you're not understanding what matters in a modern CPU...
And then remember than it's almost certain that pretty much every good idea you see in any other CPU is also implemented by Apple. I say that not as a fan boy, but as someone who understands what it takes to hit the IPC numbers they achieve across such a wide range of code.

And so on, and so on.

You're talking a lot about Arm's ISA and that's not at all my focus. We're talking about Apple and AMD.

Apple designing a core that leverages and works within the limits of the Arm ISA to achieve their design targets for the market within they work is fantastic. AMD designing a core that leverages and works within the limits of the x86-64 ISA to achieve their design targets for the market within they work is fantastic. Neither should be labeled as smarter.

Do you think Apple could design a better x86 server/gaming/workstation chip than AMD? Do you think AMD could design a better iPhone/iPad/Chromebook Arm chip than Apple?

 

[IntelUser2000]

Do you think Apple could design a better x86 server/gaming/workstation chip than AMD? Do you think AMD could design a better iPhone/iPad/Chromebook Arm chip than Apple?

Former: I do
Latter: Nope

Apple hasn't failed in delivery of new cores for the past ten years. That can't be said for Intel nor AMD.

The fact that the M1 beats Intel/AMD in not just CPU, but graphics tells us its not about the ISA(maybe slightly) but the ability for the teams to execute their vision.

This crazy spat between x86 vs ARM causes people to forget that some people/teams/companies do better than others under equal scenarios.

 

[moinmoin]

Personally I'm reserving my opinion on whether Apple is capable of designing desktop ARM chips (including expandability through PCIe etc.) for when they actually give that a try. Those would be more comparable to existing x86 chips. M1 isn't it yet, and Apple may well opt to keep it locked down, with USB-C and Thunderbolt/USB4 remaining the only way for expansions.

 

[IntelUser2000]

Apple is going to be a locked garden, and may even kill expandable Mac Pros.

Regardless, it doesn't change the fact they have a far far superior core.

 

[moinmoin]

Apple is going to be a locked garden, and may even kill expandable Mac Pros.

Regardless, it doesn't change the fact they have a far far superior core.

Those factually far far superior cores on a non-expandable Mac Pro will be pretty fun.

 

[jpiniero]

Personally I'm reserving my opinion on whether Apple is capable of designing desktop ARM chips (including expandability through PCIe etc.) for when they actually give that a try. Those would be more comparable to existing x86 chips. M1 isn't it yet, and Apple may well opt to keep it locked down, with USB-C and Thunderbolt/USB4 remaining the only way for expansions.

There may not be much to expand to. You obviously won't get any nVidia GPUs, and even AMD probably won't bother making native drivers unless they get an OEM deal which doesn't seem likely. USB4 might be fast enough for storage.

 

[moinmoin]

There may not be much to expand to. You obviously won't get any nVidia GPUs, and even AMD probably won't bother making native drivers unless they get an OEM deal which doesn't seem likely. USB4 might be fast enough for storage.

With the advantages Apple's ARM chips showcased so far I could easily imagine using high end Mac Pros as render farms. But if there's no way to connect any RAIDs to it, or even worse all storage is hardwired that kind of usage is done for.

 

[Carfax83]

Regardless, it doesn't change the fact they have a far far superior core.

I'm curious, what is your rationale for believing that Apple has a "far far superior core" to Zen 3?

 

[amrnuke]

Former: I do
Latter: Nope

Apple hasn't failed in delivery of new cores for the past ten years. That can't be said for Intel nor AMD.

The fact that the M1 beats Intel/AMD in not just CPU, but graphics tells us its not about the ISA(maybe slightly) but the ability for the teams to execute their vision.

This crazy spat between x86 vs ARM causes people to forget that some people/teams/companies do better than others under equal scenarios.

Do we have any actual evidence in benchmarks that the M1 beats Intel and AMD? We should probably start by seeing if it's even faster in SPEC2006 than Zen3 and then we'll move out to whether it's faster at all the things people actually need and want powerful processors for.

You say Apple could design a better x86 server/desktop chip than AMD and Intel - well, why haven't they? There's a 9-figure market out there waiting! And it's so easy that even the only-a-little-smart AMD and Intel can do it! And the majority of servers and supercomputers and the like are already on Linux/Unix, should be cake!

 

[jpiniero]

You say Apple could design a better x86 server/desktop chip than AMD and Intel - well, why haven't they? There's a 9-figure market out there waiting! And it's so easy that even the only-a-little-smart AMD and Intel can do it! And the majority of servers and supercomputers and the like are already on Linux/Unix, should be cake!

Apple's never really had that much success with servers. I just don't think they are interested.

 

[moinmoin]

Apple's never really had that much success with servers. I just don't think they are interested.

Apple is a big cloud service provider to its consumers (and that part of its business will only become more and more important, it's already bigger than its Mac business) but instead increasing its own data centers it instead has become a huge customer of AWS, Google Cloud and Azure (though reports on the latter are contradicting). Apple leaving server business has made it more dependent on those, with efforts for own data centers often reported to be not as fruitful as planned.

 

[name99]

With the advantages Apple's ARM chips showcased so far I could easily imagine using high end Mac Pros as render farms. But if there's no way to connect any RAIDs to it, or even worse all storage is hardwired that kind of usage is done for.

Over and over again you guys keep living in the past. Skate to where the puck is going!
Stop insisting that problem be solved the way they were 5 years ago. The problems don't go away, but the solutions change substantially.

In terms of RAID, WTF? Why on earth would RAID not be supported? I'm guessing you believe that RAID has to be connected by a PCI card, to which the only response is WHY? Thunderbolt RAID has been a thing for years. Or build your own RAID using whatever disks you want to connect to the machine and Apple's SoftRAID, the way I have done for 20 years.

In terms of hardwired storage generally: the current Mac Pro has a T2. It also has 2 SATA ports and multiple USB/TB ports. The SATA ports may go away in favor of whatever the PCIe SSD standard is next year, but why will the rest go away? Clearly they're compatible with the secure enclave/encrypted storage regime Apple imposes on the boot disk.

What is your mental model here? You all seem to have no clue how either Apple views Macs, or how people use them; but feel convinced that this is no barrier to absolute certainty about where Apple is headed with these machines.

 

[name99]

Do we have any actual evidence in benchmarks that the M1 beats Intel and AMD? We should probably start by seeing if it's even faster in SPEC2006 than Zen3 and then we'll move out to whether it's faster at all the things people actually need and want powerful processors for.

You say Apple could design a better x86 server/desktop chip than AMD and Intel - well, why haven't they? There's a 9-figure market out there waiting! And it's so easy that even the only-a-little-smart AMD and Intel can do it! And the majority of servers and supercomputers and the like are already on Linux/Unix, should be cake!

“Focus,” Jobs told Parker, “means saying no to the hundred other good ideas.” Jobs added, “I'm actually as proud of the things we haven't done as the things I have done. Innovation is saying 'no' to 1,000 things. You have to pick carefully.”

Apple plays the long game. The long game is not picking up some cash for a few years in the dying x86 market.

 

[amrnuke]

“Focus,” Jobs told Parker, “means saying no to the hundred other good ideas.” Jobs added, “I'm actually as proud of the things we haven't done as the things I have done. Innovation is saying 'no' to 1,000 things. You have to pick carefully.”

Apple plays the long game. The long game is not picking up some cash for a few years in the dying x86 market.

Parroting the "x86 is dying" line.

Back to the original point, can you point out what exactly in the Apple core makes it smarter than the Zen3 core?

 

[AmericanLocomotive]

Apple hasn't failed in delivery of new cores for the past ten years. That can't be said for Intel nor AMD.

The fact that the M1 beats Intel/AMD in not just CPU, but graphics tells us its not about the ISA(maybe slightly) but the ability for the teams to execute their vision.

This crazy spat between x86 vs ARM causes people to forget that some people/teams/companies do better than others under equal scenarios.

It also beats Intel/AMD in graphics because it's a ~1.5 years newer design than anything AMD or Intel (well at least AMD) has in the mobile space. The tightly coupled and controlled nature of the design lets them optimize immensely.

They don't have to design a CPU core and iGPU that has to work "well enough" with everything from DDR4 2400 to LPDDR4 4266. They don't have to worry about using a lot of the transistor budget to support the massive IO needs of PCs. They don't have to worry about keeping transistor counts and die-sizes down to be affordable for <$500 notebooks. I mean just look at the new Mac Mini with this chip. You get 4 USB/Thunderbolt ports, Ethernet, HDMI ...and that's it. You don't get to upgrade your RAM, you don't get to change out your SSD, you don't get any expandability. You get what you get, and that's it.

It's a great chip for people who treat computers like phones - use it for a couple years and then throw it away. AMD and Intel will never be able to achieve that level of integration just by the very nature of their products. At best you may see things like giant local caches to reduce the demands on the memory bus (Navi 2 style), but I can't imagine things like non-upgradeable storage or on-package RAM with no upgrade path will be very popular with traditional PC buyers.

I am really interested to see what Apple plans to do with the "Mac Pro" market. They can't stay x86 forever and support two different lines. But the very tight coupled nature of this M1 chip seems inherently incompatible with the ideals of a Mac Pro. Especially the latest one that has lots of storage, RAM and IO upgradability.

 

[jpiniero]

Back to the original point, can you point out what exactly in the Apple core makes it smarter than the Zen3 core?

It's a lot wider.

 

[ultimatebob]

It also beats Intel/AMD in graphics because it's a ~1.5 years newer design than anything AMD or Intel (well at least AMD) has in the mobile space. The tightly coupled and controlled nature of the design lets them optimize immensely.

They don't have to design a CPU core and iGPU that has to work "well enough" with everything from DDR4 2400 to LPDDR4 4266. They don't have to worry about using a lot of the transistor budget to support the massive IO needs of PCs. They don't have to worry about keeping transistor counts and die-sizes down to be affordable for <$500 notebooks. I mean just look at the new Mac Mini with this chip. You get 4 USB/Thunderbolt ports, Ethernet, HDMI ...and that's it. You don't get to upgrade your RAM, you don't get to change out your SSD, you don't get any expandability. You get what you get, and that's it.

It's a great chip for people who treat computers like phones - use it for a couple years and then throw it away. AMD and Intel will never be able to achieve that level of integration just by the very nature of their products. At best you may see things like giant local caches to reduce the demands on the memory bus (Navi 2 style), but I can't imagine things like non-upgradeable storage or on-package RAM with no upgrade path will be very popular with traditional PC buyers.

I am really interested to see what Apple plans to do with the "Mac Pro" market. They can't stay x86 forever and support two different lines. But the very tight coupled nature of this M1 chip seems inherently incompatible with the ideals of a Mac Pro. Especially the latest one that has lots of storage, RAM and IO upgradability.

I'd imagine that they'll release an "M2 XL" processor for the Mac Pro next year with more cores, faster cores, support for tons of system memory, and support for dedicated GPU's. And I'm sure that it will run GeekBench really REALLY fast

 

[wlee15]

It's a lot wider.

I'll note that on the SIMD floating point/integer side AMD and Intel still has a significant advantage due to vector size and cache bandwidth.

 

[name99]

Parroting the "x86 is dying" line.

Back to the original point, can you point out what exactly in the Apple core makes it smarter than the Zen3 core?

I've said this a thousand times, FFS.
Better branch predictors. Better prefetching. Better cache placement/replacement algorithms. Better ISA. More aggressive instruction fusion. Better fetch engine. Better TLB reach. Better memory controller.

How can you not comprehend this? If Apple substantially exceeds AMD's PPC, doesn't that mean that they do better than AMD on pretty everything that determines CPU performance? How else do you think the numbers work out?

 

[Thala]

I'd imagine that they'll release an "M2 XL" processor for the Mac Pro next year with more cores, faster cores, support for tons of system memory, and support for dedicated GPU's. And I'm sure that it will run GeekBench really REALLY fast

Precisely, the M1 is just the lowest end entry level SoC for TDP up to 15W. In particular it is apparently not a solution for the Mac Pro.

 

[Eug]

..

I am really interested to see what Apple plans to do with the "Mac Pro" market. They can't stay x86 forever and support two different lines. But the very tight coupled nature of this M1 chip seems inherently incompatible with the ideals of a Mac Pro. Especially the latest one that has lots of storage, RAM and IO upgradability.

Apple has already said they plan on transitioning the entire line within a two-year time frame.

It is possible even after they transition everything over to ARM, they’ll still keep a couple of legacy Intel Macs for a while longer, but it won’t be for too long because Apple doesn’t like to support old cruft for longer than it absolutely needs to, and by that time those machines will be slow anyway. They just would keep them around for mission critical Intel replacements, until those last customers either leave the platform or get with the ARM program.

As for the Mac Pro, I don’t think its chips have to directly reflect M1’s SoC design philosophy. Apple is more than a one-trick pony.

 

[insertcarehere]

You say Apple could design a better x86 server/desktop chip than AMD and Intel - well, why haven't they? There's a 9-figure market out there waiting! And it's so easy that even the only-a-little-smart AMD and Intel can do it! And the majority of servers and supercomputers and the like are already on Linux/Unix, should be cake!

A 9-figure market wouldn't get Apple out of bed, Airpods alone are $12 billion in revenues per year.

Mac pro/iMac volumes alone can't support the R&D necessary for a bespoke cpu design, x86 or ARM, and Apple, not being Intel and AMD, don't care to design and manufacture CPUs only to sell to third parties.