Discussion Apple Silicon SoC thread

Page 173 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Eug

Lifer
Mar 11, 2000
23,586
1,000
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:

 
Last edited:

Doug S

Platinum Member
Feb 8, 2020
2,252
3,482
136
From what I can find M2 is ARMv8.5A whereas the A15 uses the slightly older ARMv8.4A ISA.

I don't know how much practical difference there is between the two, but supposed the M2 has for support for any instructions added in the 8.5 update. The ARM blog page on what v8.5 added doesn't make it seem like a major overhaul. Some instructions for better random numbers and a few things to make it harder for the chip to be exploited by certain types of attacks, but nothing earth shattering.


Making it harder to exploit is something Apple would want, but it is possible they added that stuff to the A15 core and updated the CPU ID to say 8.5A as a result without actually using the A16 core.

I'd want to see multiple things A16 has which A15 does not - which may be difficult as A16 is barely changed from A15. Because there are so few changes it is all academic, but I would find it noteworthy if they took the A16 core and backported it from N4 to the N5P process M2 was made in.
 

Mopetar

Diamond Member
Jan 31, 2011
7,831
5,980
136
Even if it did have new instructions, it adds nothing if the software doesn't make use of them.

Apple's core is already plenty strong. I'd be more interested in the improvements they can make to their graphics and other parts of the SoC.
 

Doug S

Platinum Member
Feb 8, 2020
2,252
3,482
136
Even if it did have new instructions, it adds nothing if the software doesn't make use of them.

Apple's core is already plenty strong. I'd be more interested in the improvements they can make to their graphics and other parts of the SoC.

Well it has been reported Apple had a major GPU update planned for A16 but was not able to include it and had to stick with the A15 GPU. I imagine they would have wanted that in M2, but if it didn't make A16 it sure couldn't make M2. Presumably that will be in A17/M3.

If we don't see a Mac Pro in the M2 cycle I wonder if the GPU issues might be part of the reason. They might have been counting on a big boost to GPU power (no doubt along with driver improvements in dealing with a chiplet GPU since there is some obvious room for improvement based on Mac Studio's results) for the Mac Pro launch, and might consider pushing it back rather than launching something that's underwhelming on the GPU front.
 

Heartbreaker

Diamond Member
Apr 3, 2006
4,226
5,228
136
I don't care much about naming, but economic reality dictates Mac Pro must be built from the same tape-out as the higher volume mid-high parts (Max-Ultra). While I don't care much about naming, but I expect they will use M naming if they are using an M tapeout. The use the M naming in the iPad even though presumable the M designation stands for Mac Chips. So for consistency. All M tapeouts will keep the M naming.

The need to use higher volume parts is why I keep thinking the eventual Mac Pro will be some kind of backplane design with M-Ultra blades.

Hmmm...


The modules' purpose is unclear, but speculation argues that they are designed for the Apple silicon Mac Pro – potentially serving as a solution to enable a modular interface for swappable hardware components or add additional compute power via technologies like Swift Distributed Actors. There is also a chance that the compute modules could be designed for Apple's upcoming mixed-reality headset or something else entirely.
 

Doug S

Platinum Member
Feb 8, 2020
2,252
3,482
136
The code was found in iOS. If it was for the Mac Pro it wouldn't be compiled into iOS just like a lot of macOS features are not compiled into iOS.

The name 'compute module' makes it sound like it is sort of external device that an iPhone could move some of its workload to. That's the sort of thing you'd expect with a VR/AR device that requires an iPhone to operate (which i think is a given for a first gen device) That's also a nice generic name so if they generalized the API enough it could work for other stuff (whether something is in the works or they just want to be ready) that would require similar support.
 

ashFTW

Senior member
Sep 21, 2020
307
231
96
The code was found in iOS. If it was for the Mac Pro it wouldn't be compiled into iOS just like a lot of macOS features are not compiled into iOS.
The compute element could still be a device that’s designed for the Mac (Pro) ecosystem. It itself may not need a full blown OS, and iOS may be sufficient. Kind a like the Mac Studio display. As posited by @guidryp it could be the “backplane” where cartridges running full macOS plug into.
 

Doug S

Platinum Member
Feb 8, 2020
2,252
3,482
136
The compute element could still be a device that’s designed for the Mac (Pro) ecosystem. It itself may not need a full blown OS, and iOS may be sufficient. Kind a like the Mac Studio display. As posited by @guidryp it could be the “backplane” where cartridges running full macOS plug into.


I think people are expecting way too much from Apple if they think the Mac Pro will be designed in such an absolute modular fashion. It might be what they want in their wet dreams but it will not be what Apple gives them. I think it will be like a larger Studio with 2x the Max dies - Apple patents from several years ago describe chips with three sets of fusion I/Os for fully connecting four Max dies, so we know that's the core of the Mac Pro not an active backplane.

The only drama will be what sort of PCIe support there is. I expect a few x4 slots for storage/network expansion, but I think no slots is more likely than x8 or x16 slots let alone some sort of proprietary backplane.

If there is some sort of "compute module" designed to expand macOS capability they would never design it to be Pro only. You take a market as small as the Mac Pro market and design something only a niche of Mac Pro buyers would want and you are talking something that sells in four digit quantities if you are lucky. They'd never recoup engineering unless it is so insanely priced it makes $699 wheels seem like a Black Friday loss leader special. If they did one they'd make it attach via TB, so it could be used across the entire Mac lineup, and even then I can't imagine such a thing would sell all that well. I just don't see that as a viable product.
 

Heartbreaker

Diamond Member
Apr 3, 2006
4,226
5,228
136
If there is some sort of "compute module" designed to expand macOS capability they would never design it to be Pro only. You take a market as small as the Mac Pro market and design something only a niche of Mac Pro buyers would want and you are talking something that sells in four digit quantities if you are lucky.

Depends what the module components are. I'm not suggesting new silicon just for the module.

My thoughts on what Mac Pro Blades would be is just an M-Ultra on each blade. It provides more of each computing resource (CPU, GPU, AI, Media, etc..).

Downside is obviously much higher latency between the blades than local. But when you need more compute power than a single M-Ultra, you are generally looking at render farm type activities that won't be that latency sensitive.
 

ashFTW

Senior member
Sep 21, 2020
307
231
96
Depends what the module components are. I'm not suggesting new silicon just for the module.

My thoughts on what Mac Pro Blades would be is just an M-Ultra on each blade. It provides more of each computing resource (CPU, GPU, AI, Media, etc..).

Downside is obviously much higher latency between the blades than local. But when you need more compute power than a single M-Ultra, you are generally looking at render farm type activities that won't be that latency sensitive.
It’s like the XServe Cluster of old times.
 

ashFTW

Senior member
Sep 21, 2020
307
231
96
I think people are expecting way too much from Apple if they think the Mac Pro will be designed in such an absolute modular fashion. It might be what they want in their wet dreams but it will not be what Apple gives them. I think it will be like a larger Studio with 2x the Max dies - Apple patents from several years ago describe chips with three sets of fusion I/Os for fully connecting four Max dies, so we know that's the core of the Mac Pro not an active backplane.

The only drama will be what sort of PCIe support there is. I expect a few x4 slots for storage/network expansion, but I think no slots is more likely than x8 or x16 slots let alone some sort of proprietary backplane.

If there is some sort of "compute module" designed to expand macOS capability they would never design it to be Pro only. You take a market as small as the Mac Pro market and design something only a niche of Mac Pro buyers would want and you are talking something that sells in four digit quantities if you are lucky. They'd never recoup engineering unless it is so insanely priced it makes $699 wheels seem like a Black Friday loss leader special. If they did one they'd make it attach via TB, so it could be used across the entire Mac lineup, and even then I can't imagine such a thing would sell all that well. I just don't see that as a viable product.
I just wanted to point out that just because the reference was found in the iOS code, doesn’t rule out its use in the Mac Pro ecosystem.

I think Mac Pro might be a compute cluster. That’s a better way to scale the compute instead of making each node bigger and bigger. The large media houses are used to this model anyways. There was even a tech talk/workshop demonstrating this at last year‘s WWDC, though I’m too lazy to dig up the link. Mac Pro might be a 4x or 8x cluster of compute and graphics blades, on your desk. The machine could be configured with varying number of these blades. Larger clusters over the net are of course always there as well. Apple did sell something like this in the Xserve days -- a small rack with 4 Xserve nodes. I did buy two Xserves but was tempted to buy the mini cluster instead.
 
Last edited:

mikegg

Golden Member
Jan 30, 2010
1,755
411
136
Digitimes (unreliable) reports that 15" MBA is releasing in April/May. Mark Gurman thinks it's releasing in Summer.

Why M3 could be released in Summer at WWDC:
  • Assuming that Apple is skipping A16 cores for the M series, then Apple would have had extra resources to create M3 using A17. This could explain why M3 using A17 can come out before the iPhone 15 Pro. In "normal years" (which we haven't had due to covid, supply chain issues, 3nm delay), I would assume that the base M always comes out 1-2 months after A series.
  • At around April of each year, Apple should already be starting to mass produce the next A series that will be launched in September. So the A17 design has long been finished.
  • Apple would probably like to launch a new Mac computer with a new M series. Launching M3 with MBA 15" would garner a lot of interest.
  • If 15" MBA uses M2 in May, and then a few months later, all the other laptops get M3, it'd make the 15" MBA look pretty bad. Apple wouldn't want to do this because the 15" laptop market is by far, the biggest, which means the 15" MBA will likely become its top seller.
  • Mark Gurman did report that M2 will be a shortlived generation, which I agree based on the timing of 3nm mass production, A17, and M3. Even so, at WWDC 2023, it would be one year after M2 already.
My other speculation is that the M2 Ultra is also canceled along with M2 Extreme. I believe Apple would like to go all in on M3. It's also possible that the base M, M Pro, and M Max get updated once a year, but the Ultra and Extreme are updated every 2 years due to the effort, complexity, and niche market.
 

Eug

Lifer
Mar 11, 2000
23,586
1,000
126
Digitimes (unreliable) reports that 15" MBA is releasing in April/May. Mark Gurman thinks it's releasing in Summer.

Why M3 could be released in Summer at WWDC:
  • Assuming that Apple is skipping A16 cores for the M series, then Apple would have had extra resources to create M3 using A17. This could explain why M3 using A17 can come out before the iPhone 15 Pro. In "normal years" (which we haven't had due to covid, supply chain issues, 3nm delay), I would assume that the base M always comes out 1-2 months after A series.
  • At around April of each year, Apple should already be starting to mass produce the next A series that will be launched in September. So the A17 design has long been finished.
  • Apple would probably like to launch a new Mac computer with a new M series. Launching M3 with MBA 15" would garner a lot of interest.
  • If 15" MBA uses M2 in May, and then a few months later, all the other laptops get M3, it'd make the 15" MBA look pretty bad. Apple wouldn't want to do this because the 15" laptop market is by far, the biggest, which means the 15" MBA will likely become its top seller.
  • Mark Gurman did report that M2 will be a shortlived generation, which I agree based on the timing of 3nm mass production, A17, and M3. Even so, at WWDC 2023, it would be one year after M2 already.
My other speculation is that the M2 Ultra is also canceled along with M2 Extreme. I believe Apple would like to go all in on M3. It's also possible that the base M, M Pro, and M Max get updated once a year, but the Ultra and Extreme are updated every 2 years due to the effort, complexity, and niche market.
I disagree with this.

I suspect if a 15" MacBook Air is released relatively soon (ie. in the next several months), it will be M2.

I also don't think M2 Ultra will get cancelled. M3 Ultra will not get released this year.
 

Doug S

Platinum Member
Feb 8, 2020
2,252
3,482
136
I disagree with this.

I suspect if a 15" MacBook Air is released relatively soon (ie. in the next several months), it will be M2.

I also don't think M2 Ultra will get cancelled. M3 Ultra will not get released this year.


The only products that could use M2 Ultra are Studio and Pro. If they aren't doing an M2 based Pro I wouldn't be shocked if there is no M2 Ultra.