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

 

[Panino Manino]

Apple’s M1 is a fast CPU—but M1 Macs feel even faster due to QoS

Howard Oakley did an excellent deep dive on M1 scheduling and performance.

arstechnica.com

Remembers me Anandtech's investigations on Exynos failures.

 

[Thala]

Apple’s M1 is a fast CPU—but M1 Macs feel even faster due to QoS

Howard Oakley did an excellent deep dive on M1 scheduling and performance.

arstechnica.com

Remembers me Anandtech's investigations on Exynos failures.

I am not convinced that the Arstechnica author did understand half of what he was writing.

 

[Eug]

Bloomberg says:

Jade C-Chop: 8+2 CPU cores, 16 GPU cores, up to 64 GB RAM
Jade C-Die: 8+2 CPU cores, 32 GPU cores, up to 64 GB RAM

Jade 2C-Die: 16+4 CPU cores, 64 GPU cores
Jade 4C-Die: 32+8 CPU cores, 128 GPU cores

Basically it seems C-Die is being aggregated in 1X, 2X, and 4X configurations, for 10 CPU / 32 GPU, 20 CPU / 64 GPU, and 40 CPU / 128 GPU configurations respectively.

Perhaps Jade C-Chop is the binned variant of C-Die, at 10 CPU / 16 GPU, although that seems like a lot of GPU cores to "Chop".

Also, does that mean Jade 4C-Die will support 256 GB RAM? My guess is yes.

 

[B-Riz]

I can stomach this price for a luxury Chromebox, I join up tomorrow!

 

[defferoo]

Bloomberg says:

Jade C-Chop: 8+2 CPU cores, 16 GPU cores, up to 64 GB RAM
Jade C-Die: 8+2 CPU cores, 32 GPU cores, up to 64 GB RAM

Jade 2C-Die: 16+4 CPU cores, 64 GPU cores
Jade 4C-Die: 32+8 CPU cores, 128 GPU cores

Basically it seems C-Die is being aggregated in 1X, 2X, and 4X configurations, for 10 CPU / 32 GPU, 20 CPU / 64 GPU, and 40 CPU / 128 GPU configurations respectively.

Perhaps Jade C-Chop is the binned variant of C-Die, at 10 CPU / 16 GPU, although that seems like a lot of GPU cores to "Chop".

Also, does that mean Jade 4C-Die will support 256 GB RAM? My guess is yes.

The current Mac Pro supports 1.5TB of RAM, I don't see them reducing that amount in a new Mac Pro. The Jade 2C and 4C probably have a different memory controller that allow for much more memory and ECC.

 

[Eug]

The current Mac Pro supports 1.5TB of RAM, I don't see them reducing that amount in a new Mac Pro. The Jade 2C and 4C probably have a different memory controller that allow for much more memory and ECC.

You're probably right. I crossed out that line in my post (but left it in place, as a record of my idiocy ).

However, could that still mean that the 'base' die is Jade-C and they are packaging 4 of them together with a separate memory controller for the Mac Pro?

That could mean a vastly simplified Mac Pro lineup, with just two (or three) CPU options: 4C, 2C, and maybe C. Memory would be separately upgradable, offering anywhere from 32 GB and up.

 

[Doug S]

Those code names confirm they're using chiplets like many of us speculated, assuming this rumor is more true than the one from six months ago claiming they were designing monolithic chips for the Mac Pro.

 

[Eug]

Those code names confirm they're using chiplets like many of us speculated, assuming this rumor is more true than the one from six months ago claiming they were designing monolithic chips for the Mac Pro.

Where was that rumour about monolithic designs from? I remember some people here mentioning it, but I don't recall a legit leaker saying that.

 

[jpiniero]

Where was that rumour about monolithic designs from? I remember some people here mentioning it, but I don't recall a legit leaker saying that.

Possibly confusing with the Higher End Mini? That product could be called Mac Pro Mini.

 

[moinmoin]

Bloomberg says:

Jade C-Chop: 8+2 CPU cores, 16 GPU cores, up to 64 GB RAM
Jade C-Die: 8+2 CPU cores, 32 GPU cores, up to 64 GB RAM

Jade 2C-Die: 16+4 CPU cores, 64 GPU cores
Jade 4C-Die: 32+8 CPU cores, 128 GPU cores

Basically it seems C-Die is being aggregated in 1X, 2X, and 4X configurations, for 10 CPU / 32 GPU, 20 CPU / 64 GPU, and 40 CPU / 128 GPU configurations respectively.

Perhaps Jade C-Chop is the binned variant of C-Die, at 10 CPU / 16 GPU, although that seems like a lot of GPU cores to "Chop".

Also, does that mean Jade 4C-Die will support 256 GB RAM? My guess is yes.

Sounds to me like Apple may be repeating AMD's approach with Zeppelin (the Zen 1 server/desktop die).

 

[Eug]

Possibly confusing with the Higher End Mini? That product could be called Mac Pro Mini.

Yeah, the Bloomberg article says an updated Mac mini would just use the same higher end MacBook Pro chips, so 10 CPU / 32 (or 16) GPU.

As you suggest, this would likely represent a new (more expensive) higher end tier, not a replacement for the current M1 Mac mini.

I'm not sure I buy the idea of a "Mac Pro mini" naming scheme though. I'd guess it'd just be called "Mac mini" again.

BTW, I hope they finally update the form factor.

 

[Doug S]

Yeah, the Bloomberg article says an updated Mac mini would just use the same higher end MacBook Pro chips, so 10 CPU / 32 (or 16) GPU.

As you suggest, this would likely represent a new (more expensive) higher end tier, not a replacement for the current M1 Mac mini.

I'm not sure I buy the idea of a "Mac Pro mini" naming scheme though. I'd guess it'd just be called "Mac mini" again.

BTW, I hope they finally update the form factor.

I was thinking about today and it made sense for them not to update the designs of the products that got the M1, since that was the first gen / low end version. They want to give the people who are waiting to buy the higher end version something so maybe this is when you see redesigns for stuff like less bezel on the laptops.

Not sure what you want from a redesign of the Mac Mini exactly though. It is a Mac and it is Mini. It could be minier I suppose, but what's the point? What benefit is there to customers if they made it the size of a deck of playing cards instead of its current size?

 

[defferoo]

I was thinking about today and it made sense for them not to update the designs of the products that got the M1, since that was the first gen / low end version. They want to give the people who are waiting to buy the higher end version something so maybe this is when you see redesigns for stuff like less bezel on the laptops.

Not sure what you want from a redesign of the Mac Mini exactly though. It is a Mac and it is Mini. It could be minier I suppose, but what's the point? What benefit is there to customers if they made it the size of a deck of playing cards instead of its current size?

They could make the low end one a Mac nano considering the amount of unused space left in that enclosure

I think they could even fit everything in an Apple TV 4K sized box if they wanted to, basically out-NUC Intel's NUC.

 

[B-Riz]

They could make the low end one a Mac nano considering the amount of unused space left in that enclosure

I think they could even fit everything in an Apple TV 4K sized box if they wanted to, basically out-NUC Intel's NUC.

Maybe, but the mini form factor is used in server racks, lol.

5U HyperShelf for 16 Apple Mac Mini

Rack Mount up to 16 Mac minis per shelf Integrated airflow ducting for optimal cooling and thermal efficiency Supports more Mac minis per rack than alternatives in the industry Increases operation efficiency with limited space HyperShelf for Apple Mac mini Features Rack mount 16 Mac minis...

www.racksolutions.com

 

[defferoo]

Maybe, but the mini form factor is used in server racks, lol.

5U HyperShelf for 16 Apple Mac Mini

Rack Mount up to 16 Mac minis per shelf Integrated airflow ducting for optimal cooling and thermal efficiency Supports more Mac minis per rack than alternatives in the industry Increases operation efficiency with limited space HyperShelf for Apple Mac mini Features Rack mount 16 Mac minis...

www.racksolutions.com

Yeah, that's true. Not saying they will, but they could if they wanted to.

 

[defferoo]

Regarding the "Chop" having only 16 GPU cores, the GPU seems like it will make up a significant amount of die space of the new chip. Going by M1 die shots, the 8-core GPU takes up ~25% of the space? In a theoretical Jade C a 32 core GPU would be as big as the M1. The rest of the components would probably be about the same size as the M1 (+4 Firestorm cores -2 Icestorm cores + additional Neural Engine cores +updated memory controller), making it approximately 2x the size of the M1. Having the "Chop" version only require 16 GPU cores gives Apple a lot of room for error so that they can maximize yields. Up to 25% of the die could have problems before the chip is unusable. Most people would be happy with the performance of the 16 core version of the chip, so it makes sense for Apple. They probably also don't want to have too many SKUs and want to keep their lineup simple as well.

 

[ashFTW]

Regarding the "Chop" having only 16 GPU cores, the GPU seems like it will make up a significant amount of die space of the new chip. Going by M1 die shots, the 8-core GPU takes up ~25% of the space? In a theoretical Jade C a 32 core GPU would be as big as the M1. The rest of the components would probably be about the same size as the M1 (+4 Firestorm cores -2 Icestorm cores + additional Neural Engine cores +updated memory controller), making it approximately 2x the size of the M1. Having the "Chop" version only require 16 GPU cores gives Apple a lot of room for error so that they can maximize yields. Up to 25% of the die could have problems before the chip is unusable. Most people would be happy with the performance of the 16 core version of the chip, so it makes sense for Apple. They probably also don't want to have too many SKUs and want to keep their lineup simple as well.

I agree. What about memory technology? I think M1 is 2 channel LPDDRx. This 2xM1 chip needs access to up to 64GB memory. What about 4 channels of LPDDR5, with about 3x the bandwidth of M1.

Will this be on TSMC 5nm or 4nm?

 

[defferoo]

I agree. What about memory technology? I think M1 is 2 channel LPDDRx. This 2xM1 chip needs access to up to 64GB memory. What about 4 channels of LPDDR5, with about 3x the bandwidth of M1.

Will this be on TSMC 5nm or 4nm?

Most likely still TSMC 5nm given the size of the chips, I don't expect they would try a new node with these chips. Memory could be LPDDR4x again with additional channels for addressing more memory, but they could surprise us with LPDDR5.

 

[Eug]

I was thinking about today and it made sense for them not to update the designs of the products that got the M1, since that was the first gen / low end version. They want to give the people who are waiting to buy the higher end version something so maybe this is when you see redesigns for stuff like less bezel on the laptops.

Not sure what you want from a redesign of the Mac Mini exactly though. It is a Mac and it is Mini. It could be minier I suppose, but what's the point? What benefit is there to customers if they made it the size of a deck of playing cards instead of its current size?

While I'm usually not one to obsess over thinness, I think the Mac mini could go thinner. This would also help the Mac-mini-as-server types too. Even with a faster SoC, I suspect they could go thinner.

Unlike older models, the M1 Mac mini is largely empty space inside.

Right now the Mac mini is 1.4 in / 3.6 cm tall.

 

[ashFTW]

Most likely still TSMC 5nm given the size of the chips, I don't expect they would try a new node with these chips. Memory could be LPDDR4x again with additional channels for addressing more memory, but they could surprise us with LPDDR5.

Given 4x GPU and 2x CPU, it would need at least 3x the bandwidth of M1– so either 6 channels of LPDDR4x or 4 channel of LPDDR5. I think 6 channels is excessive, and also hard to arrange, (assuming two dram chips to a side), when combing 4 such chiplets for the MacPro.

 

[Gideon]

I really wish they'd use HBM instead of LPDDR on highest end SoCs. The discreet GPUs used to have it after all

 

[Eug]

I really wish they'd use HBM instead of LPDDR on highest end SoCs. The discreet GPUs used to have it after all

How would that work with Apple's shared memory setup, vs cost?

 

[Mopetar]

How would that work with Apple's shared memory setup, vs cost?

I don't really know, but if there's one company that could invest the money into production to get it to scale to the point where the cost is better for them it would be Apple.

 

[dr1337]

How would that work with Apple's shared memory setup, vs cost?

Unified memory should be independent of controller design, but in terms of cost Apple was paying AMD for navi 12, essentially just a 5700xt with HBM memory. If they could budget and afford such GPUs before I see no reason why they wouldn't start sourcing and using HBM for their own GPUs as well.

 

[Eug]

Unified memory should be independent of controller design, but in terms of cost Apple was paying AMD for navi 12, essentially just a 5700xt with HBM memory. If they could budget and afford such GPUs before I see no reason why they wouldn't start sourcing and using HBM for their own GPUs as well.

Well, that was the point of my question. Are you suggesting they will have dedicated GPU memory and separate system memory?