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

 

[HurleyBird]

albeit less impressive overall now that Apple showed their hand.

In an absolute sense I suppose so. But I think this is actually really good for AMD. M1 makes Intel laptop chips extremely unappealing and brings performance and efficiency to the forefront of everyone's mind. People are primed to compare their Windows laptop purchases against Macbooks now. While Cezanne might not be as impressive of an overall package as M1, it will at least put up a fight and win a lot of benchmarks.

 

[Heartbreaker]

Cezanne marks a more aggressive AMD in the mobile field, not another delay. They aim to compress the gap between server/desktop and mobile, since in the last years they didn't really have the resources to have them ready at the same time. The 5000 series is bound to be very competitive in the x86 ecosystem, albeit less impressive overall now that Apple showed their hand.

Disappointment for me, that has been happy with CPU for a while now, but not iGPU performance.

 

[Doug S]

To put that in to perspective: if M1 was implemented on N7, it would either be a much larger chip, physically, reducing the number of die per wafer greatly. This would adversely affect the per chip profit, and make the die plus ram package larger and more expensive. Alternately, they could have maintained the same die size, and per wafer chips, but it would have likely cost them two of the Firestorm cores and at least some of the L2 and SLC. Power draw in single core scenarios would be notably higher, and multi threaded benches would be much slower.

The A12X/A12Z is a 4+4 design just like the M1, so they clearly wouldn't have had to compromise on the number of Firestorm cores if it was built on N7, just on the L2/SLC sizing. From the A12 generation to the A14 generation the biggest transistor increase was in the NPU, which doubled in size from 8 to 16 cores, they would have stuck with 8.

 

[Eug]

So, the latest cryptic rumour leak from L0vetodream is that alongside new Mx Macs in new form factors next year, Apple will release new Intel-based models.

I think that makes sense. However, my guess is that if that is true, Apple would likely recycle the existing form factors for those new Intel Macs. It'd only be a couple of models though, not a full slate by any means. (I had predicted either this scenario, or else Apple would just keep selling 2020 models for a few years.)

What chips though, if they do get updated with Intel?

 

[Heartbreaker]

So, the latest cryptic rumour leak from L0vetodream is that alongside new Mx Macs in new form factors next year, Apple will release new Intel-based models.

I think that makes sense. However, my guess is that if that is true, Apple would likely recycle the existing form factors for those new Intel Macs. It'd only be a couple of models though, not a full slate by any means. (I had predicted either this scenario, or else Apple would just keep selling 2020 models for a few years.)

What chips though, if they do get updated with Intel?

Waste of time and resources.

The only people buying Intel Macs will be the uninformed, or locked into Intel, neither of which will skip buying if they don't do a new Intel refresh next year. So they may as well just keep selling current Intel Models where appropriate.

They have a VERY good reason to not do new Intel Model refresh next year, and it's not like Apple ever even needed a reason to not update Intel Macs in the past.

Not only that, but it stops them building the new design ARM Macs totally around their new ARM silicon, if they also have to do double duty as Intel Macbooks.

 

[lobz]

Apple's version of what?

Apple's version of anything really I swear I'll never leave the /s away anymore, no matter what.

It was just a small sarcastic reference to how I saw the most fanatic people defending the non-vesa $999 monitor stand as a) being vastly superior to an open standard solution, and b) how you should always be prepared to pay the price if you want stability and quality. Sorry if I derailed the topic a bit, that wasn't my intention, I just wanted a chuckle

 

[lobz]

Next Generation AMD APU, does't appear to be that impressive vs M1:

AMD Ryzen 7 5800U Flagship 8 Core Cezanne 'Zen 3' APU Benchmark Leaks Out, Up To 38% Faster Than 4800U in Single-Core Performance

The first benchmark of AMD's upcoming Ryzen 7 5800U APU which is an 8 core chip and part of the Zen 3 based Cezanne family has leaked out.

wccftech.com

ST performance still behind M1, GPU is still 8 core VEGA.

You're right. Apple's new chip is the undisputed king of Geekbench ST, I don't think anyone has ever tried to challenge this here, but please let's not bury an unreleased APU on its leaked GB score too deep in your disappointment just yet.

 

[Roland00Address]

His main point is that someone used desktop Ryzen power numbers to point to how much more efficient M1 is, but desktop Ryzen isn't the right comparison as that architecture is essentially designed to scale up to 64 cores and so it makes some compromises in order to be able to do that (one major one is power efficiency at low core counts). It is a much better comparison to use Renoir which has a more similar target market and is designed for that market. Renoir's efficiency, especially at low core counts, will be much higher than desktop Ryzen.

I am sorry this is nonsense where someone wants to make it fair. Things are different and that is okay. Ryzen may be superior at server work loads (as of now) and that is okay but people who buy laptops do not care. How many buyers have even heard of the word "Zen" I guarantee less than 5% of Laptop buyers, more like 1% maybe 2%.

 

[Hitman928]

I am sorry this is nonsense where someone wants to make it fair. Things are different and that is okay. Ryzen may be superior at server work loads (as of now) and that is okay but people who buy laptops do not care. How many buyers have even heard of the word "Zen" I guarantee less than 5% of Laptop buyers, more like 1% maybe 2%.

Why is it nonsense? Should we be comparing M1 to Threadripper and laughing at how badly it performs as a workstation CPU?

 

[Roland00Address]

Why is it nonsense? Should we be comparing M1 to Threadripper and laughing at how badly it performs as a workstation CPU?

Compare the things that are similar and not the extremes. Comaring the extremes is an intellectual exercise not how people actually buy devices.

It makes sense to compare a 15w laptop to 45w laptop for sometimes they are the same customer, it does not make sense to compare a 15w laptop to a EPYC 7742 , a 64 core Zen 2 cpu that uses 225w of power (or the future EPYC 7H12 which is 280w but does not have a "list price.")

(Metaphor time) One does not compare a speedboat to a yacht to a cruise ship. That is so far removed from reality that only 1 in 1000 people would care.

-----

Yes Renoir is a better comparison than Ryzen / Epyc "but saying you are not being fair" is trying to move the discussion from comparing to similar to similar.

 

[shady28]

I am sorry this is nonsense where someone wants to make it fair. Things are different and that is okay. Ryzen may be superior at server work loads (as of now) and that is okay but people who buy laptops do not care. How many buyers have even heard of the word "Zen" I guarantee less than 5% of Laptop buyers, more like 1% maybe 2%.

Most of these sites ability to compare like parts to like is a hot mess. Laptop chips should be compared to laptop chips. Desktop chips to desktop chips. If they want to compare laptops to desktops it should be a in a separate chart labelled as such. If they want to project into the future of what "might be" by interpolating power, performance, and so on for a theoretical desktop version of a laptop chip they should separate that from factual data and put it in a section labelled "Guesswork".

 

[amrnuke]

Compare the things that are similar and not the extremes. Comaring the extremes is an intellectual exercise not how people actually buy devices.

It makes sense to compare a 15w laptop to 45w laptop for sometimes they are the same customer, it does not make sense to compare a 15w laptop to a EPYC 7742 , a 64 core Zen 2 cpu that uses 225w of power (or the future EPYC 7H12 which is 280w but does not have a "list price.")

(Metaphor time) One does not compare a speedboat to a yacht to a cruise ship. That is so far removed from reality that only 1 in 1000 people would care.

-----

Yes Renoir is a better comparison than Ryzen / Epyc "but saying you are not being fair" is trying to move the discussion from comparing to similar to similar.

Well, why not compare a "15W laptop" to... a "15W laptop"? Why compare it to a "45W laptop"?

 

[Roland00Address]

Well, why not compare a "15W laptop" to... a "15W laptop"? Why compare it to a "45W laptop"?

Yes you should compare the 15w laptop to a 15w laptops. Like 5 different ones of various OEMs and various intel and amd chips. (Likewise compare it to past Apple Laptops.)

I am also saying some people who buy airs are also sometimes* "pro users" who are happy buying 45w chip laptops like macbook pros so might as well include those relevant AMD and Intel silicon there too. (*not all air users are pro users, but sometimes they are.)

-----

But "no one* " cares about comparing 15w laptops to 225w desktop chips.

* Yes there are some one who cares but they are 1 in 1000 type things. And these people are often not doing it for information or factual perspectives but they want one side to be winning the competition for they are "fans" and atypical people out there.

• I say this with love in my heart for I am a fan too, I wanted to see cinebench 11.5 benches of these m1s for I want to compare it to 8 to 12 year old hardware to **signify** how far we have come. But trying to maintain "fairness" with fans is an absurd thing for when you are comparing 15w to 225w chips you are no longer comparing anything that is relevant in the real world for a person who buys a macbook air to use as a daily driver uses it in different ways than a person who has a 225w workstation / server cpu. Different workflow, different experiences, different software running etc. How you use the software changes when the device is an ultraportable mobile computer vs something you sit down on at a desk, and the "work device" doubles as a space heater for the room.

Most of these sites ability to compare like parts to like is a hot mess. Laptop chips should be compared to laptop chips. Desktop chips to desktop chips. If they want to compare laptops to desktops it should be a in a separate chart labelled as such. If they want to project into the future of what "might be" by interpolating power, performance, and so on for a theoretical desktop version of a laptop chip they should separate that from factual data and put it in a section labelled "Guesswork".

100% agree and this is what I am trying to signify. It is trying to cast an augury effect. To see the future by reading the omen signs of the present.

( I love these things too but it is up to the review site to underlying this is not like vs like. Furthermore it is up to fans to remember no one cares if a 225w chip is "tuned" differently than a 15w chip for they are different devices entirely. It is not fair to compare an ocean liner cruise ship to a speedboat and it will never be fair, so stop using the word fair. They are as different as a mouse vs an elephant. )

 

[Hitman928]

Compare the things that are similar and not the extremes. Comaring the extremes is an intellectual exercise not how people actually buy devices.

It makes sense to compare a 15w laptop to 45w laptop for sometimes they are the same customer, it does not make sense to compare a 15w laptop to a EPYC 7742 , a 64 core Zen 2 cpu that uses 225w of power (or the future EPYC 7H12 which is 280w but does not have a "list price.")

(Metaphor time) One does not compare a speedboat to a yacht to a cruise ship. That is so far removed from reality that only 1 in 1000 people would care.

-----

Yes Renoir is a better comparison than Ryzen / Epyc "but saying you are not being fair" is trying to move the discussion from comparing to similar to similar.

Ok, I think I misunderstood the post I replied to previously, thanks.

As far as M1 comparison go, it would be interesting to find out what the actual sustained power limit in the MBP and MBA are. Andrei showed that the M1 can use up to 21 W on the CPU and 30+W full SOC when both CPU and GPU are fully loaded. I wonder how much MBP and MBA could actually sustain though, even just with a full CPU load. I think comparing MBP to 45W x86 CPUs is fine as there is probably a good cross section of people who would be fine with the higher power consumption and its consequences if the 45W solution provided a significant performance advantage for their applications. Maybe not so much the MBA but it also uses the same chip but will throttle more given serious multi-core workloads.

 

[shady28]

Ok, I think I misunderstood the post I replied to previously, thanks.

As far as M1 comparison go, it would be interesting to find out what the actual sustained power limit in the MBP and MBA are. Andrei showed that the M1 can use up to 21 W on the CPU and 30+W full SOC when both CPU and GPU are fully loaded. I wonder how much MBP and MBA could actually sustain though, even just with a full CPU load. I think comparing MBP to 45W x86 CPUs is fine as there is probably a good cross section of people who would be fine with the higher power consumption and its consequences if the 45W solution provided a significant performance advantage for their applications. Maybe not so much the MBA but it also uses the same chip but will throttle more given serious multi-core workloads.

I think they get lost in the weeds of their analysis. This just makes for a bunch of discussion on things that are irrelevant to actual buyers. I mean really, someone actually buying a laptop should probably look at (not in this order, in their order of preference) :

1 - Performance
2 - Battery life
3 - Portability
4 - Aesthetics / construction / comfort

#4 would be affected by things like heat

If someone makes a laptop chip that uses 85W but it's quiet, cool to the touch, fast, light, small, and lasts 48 hours on a charge - do you care? Does it matter how they did it?

I know that's hyperbolic, but rather than focus on the things that feed into a result, the focus should be on the result itself. That is what I mean by getting lost in the 'weeds'.

 

[RasCas99]

I found this to be a cute comparison of real world usage of M1 vs 4800U and TGL , and I think this is where the M1 shows its worth the most (not in regards to this video but in general the way it is such a clear win in battery) , note its pushing a higher resolution screen to boot :

TLDR of the video - M1 beats the 4800U and TGL laptops in spades (it drains 20% battery vs ~50% of the Intel/AMD machines , AMD does a bit better then Intel) - allowing another 4 cycles of the same usecases tests vs 1 in the X86 machines , Zoom ran non native on the M1 as well , for an ultra portable machines this a killer results , one that will no doubt echo with consumers of such machines.

Agree with the comparing a CPU in its own weight (TDP ?) class , as we do in boxing.

Note - seems like Apple are getting the "they are on 5nm , so thats why they are better" treatment in some posts , but AMD gets a big PASS when it comes to their process advantage over Intel (and to some extent Nvidia) , in the end of the day , Apple are paying a hefty sum to be on the leading edge TSMC tech (covering a portion of TSMC R&D) so they can get the edge on everyone else , we can only compare what we can buy and thats why Intel being behind on process is a case of "it your problem , deal with it" and not an excuse to explain their products inability to compete , I dont care if Intel were better then AMD in my home computer if they had a better process, what i care is that MY home computer actually runs better on AMD.

 

[moinmoin]

but AMD gets a big PASS when it comes to their process advantage over Intel

That's not really correct. Unlike with fabless companies (so every other company we talk about) it's completely on Intel itself how much or little process advantage it has. Intel enjoyed a huge near half decade process advantage over other companies, now that advantage dried up. Intel has nobody to blame but itself for that. And regarding AMD there are actually more people wanting it to pick up new TSCM nodes faster than it already does. On a technical merit the changes between process nodes warrants a discussion, especially as Apple publicly paints the picture of not changing much about its silicon aside the node and number of cores of the different units, whereas AMD just managed huge gains while staying on the same node, of course that's going to be talked about.

 

[coercitiv]

Note - seems like Apple are getting the "they are on 5nm , so thats why they are better" treatment in some posts , but AMD gets a big PASS when it comes to their process advantage over Intel (and to some extent Nvidia) , in the end of the day , Apple are paying a hefty sum to be on the leading edge TSMC tech

AMD gets a pass on desktop for being on 7nm because Intel is deliberately keeping the desktop on 14 nm at this point. In mobile we have an apples to apples comparison with TSMC 7nm and Intel 10SF, so when comparing Renoir vs. TGL vs. Cezanne/Lucienne we won't have to give a free pass to anyone.

Apple being on 5nm is not a problem at all in my book, it's perfectly fine to gauge perf and perf/watt based on existing products. One caveat though, don't try to turn this around if and when some other platform gets early access to some technological advancement. We have 2 separate competing ecosystems, so let's always view them as such.

 

[lobz]

Note - seems like Apple are getting the "they are on 5nm , so thats why they are better" treatment in some posts , but AMD gets a big PASS when it comes to their process advantage over Intel (and to some extent Nvidia)

Huge results for the M1, so you don't really need to make untrue points. AMD never gets a pass for their process node advantage, because it's 2 things at once. First it's one of the reasons it's better than intel, second, they deserve praise to get over hurdles where Intel couldn't. Just because you can explain something with a reason, it doesn't mean it's an excuse. So please, let us all try to stick to fruitful conversation topics. Apple sourced a lot of wafers on a very expensive, bleeding-edge manufacturing process node, kudos to them. They're undoubtedly proud of it, there's no need to feel attacked when someone says, that is one of the reasons it overshadows other low-power laptop solutions so heavily, that is just a pure fact. Why do you feel like you need to defend a very-very rich (and shamefully so) company just because people like to speculate and theorize on tech forums, how would or how will 2 products using the same node compare?

 

[nxre]

On the topic of node advantages, I think Apple and AMD don't use the same cells even on the same node. AFAIK apple uses the high density cells while AMD uses the high performance ones, so even ISO-node comparisons would be complicated.

 

[Qwertilot]

Better place to ask this here I suppose - has anyone mamaged to bench mark the NPU portion of the M1?

Apple put a lot of area and resources into upping its performance for A14 & M1, so they presumably have a considerable end goal in mind.

 

[moinmoin]

On the topic of node advantages, I think Apple and AMD don't use the same cells even on the same node. AFAIK apple uses the high density cells while AMD uses the high performance ones, so even ISO-node comparisons would be complicated.

For Zen 2 on N7 this is completely wrong and a common(ly repeated) mistake, AMD did use the high density cells as well. Considering the talks about Zen 3 using the same improvements as the Zen 2 XT chips that likely applies to Zen 3 too.

Interesting! The slides state that Zen 2 uses the N7 HD (6T) process variant (high density). It has been widely presumed it would use N7 HP (7.5T) for high performance.

Perhaps Zen 2 could have hit that 5 Ghz mark, if that was the priority. Now it seems power and area were more important.

Edit: 6T is also in this slide, so it is very unlikely to be a mistake.

 

[DrMrLordX]

Still a disappointment. It means no movement on x86 integrated graphics for yet another generation.

It's so disappointing that AMD will probably sell out of Cezanne, too. Wait, isn't this supposed to be an M1 thread?

But "no one* " cares about comparing 15w laptops to 225w desktop chips.

I do, if I think the 15w laptop SoC may be the future basis for a 45-105w desktop/workstation chip.

 

[scannall]

I'm rather late to this conversation. Though the squabbling in it is disappointing.

I think they hit at least a standing triple if not a home run with this SOC. That they are able to hit the top of the ST benchmarks despite clocking 30 to 35 percent slower is outstanding all by itself. Plus top of the heap graphics all in a 15 watt package. This is not a knock on AMD, they have an excellent product. And what I use in both my desktop and laptop. But what a lot of folks here fail to understand is Windows and OSX are not interchangeable. If you live in the OSX world you get a lot of other benefits aside from just your computer, like continuity between all your iDevices. Software for OSX and iOS that you prefer. Better support, and prefer the OSX workflow. Windows 10 is fine, but so is OSX.

What Apple has done here isn't a miracle. It's been obvious for quite a long time that the way forward is to go wider. At the end of the day, throughput is what matters most. Not IPC, not clocks. But how much you can get done in a given amount of time. Higher clocks has been bled dry and then some. So that's out. Only leaves getting more out of the cycles you have. By keeping clocks down around the silicon's sweet spot they get great energy savings. And by going so wide they get a lot of throughput.

x86 will have to go wider as well at some point, though because of the variable length instructions that is vastly more difficult.

Love Apple or hate them. But at least appreciate that they are pushing the boundaries further and hopefully pushing both x86 and other ARM builders forward as well. It's a good product that is past due, and I for one am glad to see it.

 

[Doug S]

So, the latest cryptic rumour leak from L0vetodream is that alongside new Mx Macs in new form factors next year, Apple will release new Intel-based models.

I think that makes sense. However, my guess is that if that is true, Apple would likely recycle the existing form factors for those new Intel Macs. It'd only be a couple of models though, not a full slate by any means. (I had predicted either this scenario, or else Apple would just keep selling 2020 models for a few years.)

What chips though, if they do get updated with Intel?

If the next round of ARM Macs aren't due until Q2/Q3 that makes my earlier prediction that the 8+4 chip will be made using next year's cores that will ship in the A15, and N5P process, even more likely.

As for what Intel chips they'll use, presumably the latest ones assuming Apple can get them in sufficient quantity. They obviously have nothing to fear from performance testing going head to head against Intel's best.

All Apple needs to do is swap out the Intel CPU and chipset, it keeps the Intel line current for those who are more cautious or consider x86 Windows compatibility very important - and makes them as the "last x86 Mac of their type" useful for more years. Giving those customers more time to wait for ARM Mac software that does what they need to arrive and/or Windows/ARM to become a viable option for them.