Discussion Apple Silicon SoC thread

Page 6 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Eug

Lifer
Mar 11, 2000
23,807
1,385
126
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:

Screen-Shot-2021-10-18-at-1.20.47-PM.jpg

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:


M1 Ultra discussion here:


M2 discussion here:


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:


M4 Family discussion here:

 
Last edited:

IvanKaramazov

Member
Jun 29, 2020
56
102
66
I wouldn't be at all surprised if the M1 and A14X are the same silicon, just with a fuse or two blown to cause it to identity itself differently. The M1 has an ISP, which probably doesn't have much function on a laptop that has a single camera of comparatively low quality (versus what you get on modern phones)
Worth noting that they are using the ISP to attempt to sharpen the image on the webcam. I can't imagine you can do much to improve that terrible 720p camera, but I suppose we'll see.
 

Doug S

Platinum Member
Feb 8, 2020
2,742
4,664
136
Interesting. If they want to continue with the single SoC design for the 15" macbook nad imac, they probably need to use HBM2 there (at least i hope they do), otherwise the graphics perf will take a dive compared to last gen

I suspect the rumored 8+4 core chip will slide in those (and maybe be offered as an optional upgrade for the 13" MBP) and it will also have more GPU cores.

I would also not be surprised if that 8+4 chip was able to operate as a chiplet so 2 or 4 of them could be used to create a 16 or 32 core MCM for the high end stuff. But they said the transition would take two years (like I had previously guessed) so they have plenty of time to do a monolithic design for those should they choose to do so.

I wonder how a Mac Pro with 32 cores taken from a 3nm A16 would do against contemporary Intel and AMD based workstations?
 

Doug S

Platinum Member
Feb 8, 2020
2,742
4,664
136
Worth noting that they are using the ISP to attempt to sharpen the image on the webcam. I can't imagine you can do much to improve that terrible 720p camera, but I suppose we'll see.

That's hardly a computationally demanding process, certainly not worth using an ISP for unless you already have one laying around.
 

Entropyq3

Junior Member
Jan 24, 2005
22
22
81
The RAM isn't integrated into the SoC, it would be integrated into the package via stacking. Just like it is in the iPhone/iPad SoCs. Or so he's assuming.
The CGI assembly videos can be interpreted. I was basically asking the OP, Eug, where he got memory type, and cache size from. Didn’t see either in the presentation.
 

IvanKaramazov

Member
Jun 29, 2020
56
102
66
That's hardly a computationally demanding process, certainly not worth using an ISP for unless you already have one laying around.
For sure, I expect they're using it because it's there, essentially. Not a bad side benefit but I do wonder if they'll use it for anything else or if it's mostly wasted space on the Mac?
 

Eug

Lifer
Mar 11, 2000
23,807
1,385
126
The CGI assembly videos can be interpreted. I was basically asking the OP, Eug, where he got memory type, and cache size from. Didn’t see either in the presentation.
Hmm... I didn't realize that Apple had listed even more cache, separate for the high-efficiency cores. I'll adjust the first post.

f1605031779.jpg

f1605031791.jpg

As for the DDR4 memory, they didn't actually say that for the M1 chip, but they were describing a non-M1 logic board design like this:

Screen Shot 2020-11-10 at 4.26.30 PM.png

But then they go on to say all of that has been combined into M1.

f1605031716.jpg

I think it's reasonable to believe M1 uses DDR4 and not DDR5.
 
Last edited:
  • Like
Reactions: amd6502

name99

Senior member
Sep 11, 2010
496
382
136
I can’t find that anywhere. Where’s this info from?
Likewise for the cache-size data.
Delete this, My bad. I misinterpreted something that slipped by during the presentation.

I'm no longer certain the RAM is LPDDR4, could go either way.
Anyone have better evidence.
 

Tuna-Fish

Golden Member
Mar 4, 2011
1,475
1,976
136
Did they say 128KiB+192KiB L1 cache?
That's huge, Zen and Skylake are 32KiB+32KiB L1.

Implies low clock speed. Fetch from L1 is one of the major timing-critical circuits in modern CPUs. High clocked CPUs have small L1:s, backed by medium-sized fast L2:s, because otherwise the delay (in cycles) of fetching from L1 would get too high to reasonably hide. If you reduce clocks, you can push up L1 sizes and not have an intermediate level between L1 and the LLC. To keep performance up, this of course also implies that their design is very wide.

Regarding the non-upgradeable RAM: This has been a long time coming. In reasonably specced laptops, the energy needed to signaling to DRAM is a significant proportion of total energy budget. Going with soldered LPDDR cuts that in half. Everyone knows it would help, but so far no-one has been willing to take the leap. I suspect that now that Apple gets to be the one who sells this concept to people and makes it palatable, ultraportables will mostly switch over in just a few hardware cycles. Other laptops will follow suit later.

Going with just 8GB on the weak models and at most 16GB is a ballsy move. I think it's just too low.
 

name99

Senior member
Sep 11, 2010
496
382
136

name99

Senior member
Sep 11, 2010
496
382
136
Implies low clock speed. Fetch from L1 is one of the major timing-critical circuits in modern CPUs. High clocked CPUs have small L1:s, backed by medium-sized fast L2:s, because otherwise the delay (in cycles) of fetching from L1 would get too high to reasonably hide. If you reduce clocks, you can push up L1 sizes and not have an intermediate level between L1 and the LLC. To keep performance up, this of course also implies that their design is very wide.

Regarding the non-upgradeable RAM: This has been a long time coming. In reasonably specced laptops, the energy needed to signaling to DRAM is a significant proportion of total energy budget. Going with soldered LPDDR cuts that in half. Everyone knows it would help, but so far no-one has been willing to take the leap. I suspect that now that Apple gets to be the one who sells this concept to people and makes it palatable, ultraportables will mostly switch over in just a few hardware cycles. Other laptops will follow suit later.

Going with just 8GB on the weak models and at most 16GB is a ballsy move. I think it's just too low.

Too low based on what?
Plenty of macbooks are sold to people (like half my family) who only use them for email, browsing, and various forms of chat.
Remember macOS uses RAM compression (which has had many years to be perfected) and will have even faster swapping than the current machines.

Honestly, this carping about RAM feels like people desperate to find SOMETHING to complain about. It does not match the reality of most user experience.
 

Eug

Lifer
Mar 11, 2000
23,807
1,385
126
Regarding the non-upgradeable RAM: This has been a long time coming. In reasonably specced laptops, the energy needed to signaling to DRAM is a significant proportion of total energy budget. Going with soldered LPDDR cuts that in half. Everyone knows it would help, but so far no-one has been willing to take the leap. I suspect that now that Apple gets to be the one who sells this concept to people and makes it palatable, ultraportables will mostly switch over in just a few hardware cycles. Other laptops will follow suit later.

Going with just 8GB on the weak models and at most 16GB is a ballsy move. I think it's just too low.
This is just M1. These replace the low end models of those lines. I can't remember all the specs for all the models, but I do know that at least some of them were limited to 16 GB anyway. For example, the entry level 2020 Intel 13" MacBook Pro could not be spec'd with more than 16 GB RAM. However, that model has now been deleted from their lineup, leaving only the higher end 13" model that can have 32 GB.

M1X or whatever it's called is probably going to support up to 64 GB RAM.
 

Eug

Lifer
Mar 11, 2000
23,807
1,385
126
Regarding the non-upgradeable RAM: This has been a long time coming. In reasonably specced laptops, the energy needed to signaling to DRAM is a significant proportion of total energy budget. Going with soldered LPDDR cuts that in half. Everyone knows it would help, but so far no-one has been willing to take the leap. I suspect that now that Apple gets to be the one who sells this concept to people and makes it palatable, ultraportables will mostly switch over in just a few hardware cycles. Other laptops will follow suit later.
Did I misunderstand this part of your post? Cuz Apple has been using soldered RAM for just about forever. However, it used to be on the mainboard, not inside the SoC package.
 

Gideon

Golden Member
Nov 27, 2007
1,769
4,126
136
Honestly, this carping about RAM feels like people desperate to find SOMETHING to complain about. It does not match the reality of most user experience.
I agree that 3 GHz is plenty for macbook Air type machine. I also agree thet 16GB is plenty for Air, but for Macbook? not as clearly.

As a full stack web-develper in a company where a lot of people (me included) use macs I do really see it as being a major limitation. We consult a number of companies whose projects usually don't run that well at all on a 16GB machines (we usually manage to set up a bit leaner stacks). That's despite memory compression (which btw is not some fairy-dust super tech but is also used in Windows and Linux) and swapping. Once you start running decently complex micro-service projects in docker along with multiple IDE windows and browsers along with all kinds of required electron based chats (slack, teams, skype), terminals spotify and other helper tools, 16 GB of ram runs out really quickly.

I do agree that 32GB is mostly overkill for the usual web-dev, but 16GB can often be limiting, hence my complaint. While I have a 15" 2018 model, quite a few devs (especially female ones) do prefer the 13" models and 16GB would be a dealbreaker for them.

And it is a legitimate conern for me. I was thinking about buying a few during the screencast (and possibly some minis for the office, since they are so cheap) but I'll probably wait for the later versions as it stands. (I certainly wouldn't buy a new laptop in 2021, meant to last for years, with just 16GB)

I also wish you were a bit less agressively-defensive about your favorite company (everyone who complains must exclusevily do it because they are dumb and have biases), especially as it's the one you worked for (if I remember correctly).
 
  • Like
Reactions: Tlh97

Eug

Lifer
Mar 11, 2000
23,807
1,385
126
I agree that 3 GHz is plenty for macbook Air type machine. I also agree thet 16GB is plenty for Air, but for Macbook? not as clearly.

As a full stack web-develper in a company where a lot of people (me included) use macs I do really think this is a major limitatino. We consult a number of companies whose projects usually don't run that well at all on a16GB machine, despite memory compression (which btw is not some fairy-dust super tech but is also used in Windows and Linux) and swapping. Once you start running decently complex micro-service projects in docker along with multiple IDE windows and browsers along with all kinds of electron based chat (slack, teams, skype) and spotify 16 GB of ram runs out really quickly.

I do agree that 32GB is mostly overkill for the usual web-dev, but 16GB can often be limiting, hence my complaint. While I have a 15" 2018 model, quite a few devs actually run 13" models and 16GB would be a dealbreaker for them. I was thinking about buying a few during the screencast (and possibly some minis for the office, since they are so cheap) but I'll probably wait for the later versions as it stands.

I also wish you were a bit less agressively-defensive about your favorite company (everyone who complains must exclusevily do it because of their biases), especially as it's the one you worked for (if I remember correctly).
I'm 99% sure Apple will introduce a 14" Pro and 16" Pro next year, both with mini-LED, and with support for at least 32 GB RAM max and possibly up to 64 GB RAM max. And they will be LPDDR5.

Will the RAM be on the mainboard again though?
 
  • Like
Reactions: Tlh97 and Gideon

amrnuke

Golden Member
Apr 24, 2019
1,181
1,772
136
I for one think this is going to be a watershed moment and will surely be one of the biggest highlights for Tim Cook's reign, no matter what comes. This will formally erase that blurry line between phones and computers.

As for the performance, using just SPEC and GB5 it would be easy to get excited. This will be a stellar laptop and entry level desktop Mac for most users. Undoubtedly, it'll handle word processing, light to medium spreadsheet work, presentations, browsing, and so on with ease.

The big question will be whether this thing can scale up, and what the power/speed ramp looks like on a real chip. Do we compare this 4 big/4 little chip to a 4/8 Renoir/ICL/TGL chip or to an 8/16 chip? What did Apple use in their comparison? And in what applications? Lots of questions. Hopefully we get some good benchmarks next week.
 
  • Like
Reactions: Tlh97

name99

Senior member
Sep 11, 2010
496
382
136
I agree that 3 GHz is plenty for macbook Air type machine. I also agree thet 16GB is plenty for Air, but for Macbook? not as clearly.

As a full stack web-develper in a company where a lot of people (me included) use macs I do really see it as being a major limitation. We consult a number of companies whose projects usually don't run that well at all on a 16GB machines (we usually manage to set up a bit leaner stacks). That's despite memory compression (which btw is not some fairy-dust super tech but is also used in Windows and Linux) and swapping. Once you start running decently complex micro-service projects in docker along with multiple IDE windows and browsers along with all kinds of required electron based chats (slack, teams, skype), terminals spotify and other helper tools, 16 GB of ram runs out really quickly.

I do agree that 32GB is mostly overkill for the usual web-dev, but 16GB can often be limiting, hence my complaint. While I have a 15" 2018 model, quite a few devs (especially female ones) do prefer the 13" models and 16GB would be a dealbreaker for them.

And it is a legitimate conern for me. I was thinking about buying a few during the screencast (and possibly some minis for the office, since they are so cheap) but I'll probably wait for the later versions as it stands. (I certainly wouldn't buy a new laptop in 2021, meant to last for years, with just 16GB)

I also wish you were a bit less agressively-defensive about your favorite company (everyone who complains must exclusevily do it because they are dumb and have biases), especially as it's the one you worked for (if I remember correctly).

Two points.
(a) Mocking a technology as "fairy-dust super tech" is not a way to endear yourself to others. I made no claims as to Apple's exclusive use of this tech.

(b) The 13" MBP is the low-end model. Which part of "low end is low end" did you not understand. Only idiots (of which, OMG, there are so many. Reading Twitter right now makes you weep for humanity) assume that these specs are the end point for the MBP range as a whole.

If these specs don't meet your needs, then wait. The claim is not "no-one needs more than 16GB", it is "plenty of people will find these machines meet their needs". If you're not one, you're not one.

(c) I am not "agressively-defensive about [my] favorite company"; I am happy to (and frequently do) criticize Apple. I am aggressively against stupidity and dishonesty.
 

Roland00Address

Platinum Member
Dec 17, 2008
2,196
260
126
I suspect the rumored 8+4 core chip will slide in those (and maybe be offered as an optional upgrade for the 13" MBP) and it will also have more GPU cores.

I would also not be surprised if that 8+4 chip was able to operate as a chiplet so 2 or 4 of them could be used to create a 16 or 32 core MCM for the high end stuff. But they said the transition would take two years (like I had previously guessed) so they have plenty of time to do a monolithic design for those should they choose to do so.

I wonder how a Mac Pro with 32 cores taken from a 3nm A16 would do against contemporary Intel and AMD based workstations?
Serious question

A) could we do M1s in chiplets of 2 and 4?
B) or would it make more sense to make a larger chip as a separate die?
C) or is Apple just going to wait 6 months or a year and do the successor of the firestorm (big) and icestorm (little) before doing 8 or more big cores?

A, B, or C? I have had time to see if it is technically impossible / not feasible to see if the M1 could be done as chiplets today with the info Apple dropped.