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

 

[Ajay]

Sorry, but I need to let it out of my chest...
This dieshot/floorplan looks so... ugly!

Hmm, massively integrated features on one SoC + software placement optimization == M1 Pro/Max.

 

[leoneazzurro]

I doubt Intel would be able to match Apple's performance per watt even with the exact same process node. They need a whole new architecture. Apple simply made the most of available die space.

Who spoke about Intel? And well, there are other options with better perf/W.

 

[nxre]

Was L1 Instruction cache double to 128kb on the new efficiency cores, or was it this size too in M1?

 

[StinkyPinky]

Well I just dropped $3400 with tax (edu pricing) for the 16 inch, 1tb ssd, 32gb ram, maxed out CPU. Even with a $1400 trade in that hurts a bit but hey, this performance looks insane and there's always the return window..........right?

....right?


...right guys?

 

[Heartbreaker]

Also, I wonder if the two efficiency cores are just there for compatibility (maybe some OS functions are hard coded to only run on LP cores).

It's still a laptop and you want to extend battery life when doing minor tasks like web surfing and watching video.

 

[igor_kavinski]

Also, I wonder if the two efficiency cores are just there for compatibility (maybe some OS functions are hard coded to only run on LP cores).

There are power optimized transistors and performance optimized transistors. You can't beat a power optimized transistor in a power consumption contest even if you massively downclock a performance optimized transistor because the latter leaks electrons more readily. So CPU designers have to choose between these two types of "libraries". Apple's tests must have shown that two LP cores are more than up to the task of handling normal OS background functions, plus it also allows them to enable maximum battery life.

 

[StinkyPinky]

Looking at the slide footnotes, it appears that the M1 Pro's GPU is straight up faster than a laptop 3050ti ("Discrete PC laptop graphics") while the M1 Max's compares very well with a maxed-out RTX 3080 Mobile ("High-end PC laptop graphics"). In an SoC that probably consumes less than 80w all-in, that's an engineering statement if there ever is one.

I suspect it may not be quite that fast, but even 3070 level performance mobile would be amazing. If only Microsoft could include ARM support in Windows 11 I'd straight up sell my desktop since it would be strong enough at gaming to not need one.

 

[igor_kavinski]

Well I just dropped $3400 with tax (edu pricing) for the 16 inch, 1tb ssd, 32gb ram, maxed out CPU.

I'm envying you right now...

 

[Ajay]

It's still a laptop and you want to extend battery life when doing minor tasks like web surfing and watching video.

The multiple on board encode/decode blocks handle all the heavy lifting for the later - hence the long battery life watching videos.

 

[trivik12]

Apple can destroy x86 business if they are ok selling these chips to other OEMs.

 

[igor_kavinski]

Who spoke about Intel?

Sorry. My brain isn't thinking straight after watching the Apple event

And well, there are other options with better perf/W.

Such as?

 

[Tup3x]

The multiple on board encode/decode blocks handle all the heavy lifting for the later - hence the long battery life watching videos.

Do they...? Still no VP9 nor AV1 support?

 

[StinkyPinky]

Here at the compared models apparently:

Compact pro PC laptop	Razer Blade 15 Advanced (RZ09-0409CE53)	NVIDIA GeForce RTX 3080 Laptop (95 W)
High-end PC laptop	MSI GE76 Raider 11UH-053	NVIDIA GeForce RTX 3080 Laptop (155 W)

So from what I remember of that graph it was slightly above the compact and slightly below the high end. So probably between the two in performance. Of course that is cherry picked. Still amazing performance though.

 

[Ajay]

Well I just dropped $3400 with tax (edu pricing) for the 16 inch, 1tb ssd, 32gb ram, maxed out CPU. Even with a $1400 trade in that hurts a bit but hey, this performance looks insane and there's always the return window..........right?

....right?


...right guys?

Wow! You make a lot more money in college than I did (course, that was about three decades ago).

Hey, just a 1TB SSD with that budget?!

 

[repoman27]

That was with a discrete GPU. They are using "Quad" (or "Eight" on the Max) channel LPDDR5.

LPDDR5 channels are generally 16-bit, so it would actually be 16 or 32-channel. They're using LPDDR5-6400 in x128 packages... which is rather exotic. And not at all what I was expecting seeing as Apple has never previously shipped a product utilizing LPDDR5. Pretty bonkers.

 

[Ajay]

Do they...? Still no VP9 nor AV1 support?

IDK. But, if not, well it is Apple.

 

[Ajay]

LPDDR5 channels are generally 16-bit, so it would actually be 16 or 32-channel. They're using LPDDR5-6400 in x128 packages... which is rather exotic. And not at all what I was expecting seeing as Apple has never previously shipped a product utilizing LPDDR5. Pretty bonkers.

Yeah, when I saw the large squarish DRAM chips, I knew something was up. Pretty bonkers it is!
I wonder who did this custom work??

 

[Roland00Address]

Looking at the slide footnotes, it appears that the M1 Pro's GPU is straight up faster than a laptop 3050ti ("Discrete PC laptop graphics") while the M1 Max's compares very well with a maxed-out RTX 3080 Mobile ("High-end PC laptop graphics"). In an SoC that probably consumes less than 80w all-in, that's an engineering statement if there ever is one.

Per Anandtech Ryan Smith there are more transistors in this 5nm chip than the NVIDIA's Ampere A100 (currently Nvidia’s fastest gpu which is on 7nm.)

Now of course the Nvidia will be faster due to having a much higher TDP for Apple is putting the chip in a laptop, but I would hope the Apple silicon is fast in GPUnsince they are putting so much transistors in everything!

 

[StinkyPinky]

Wow! You make a lot more money in college than I did (course, that was about three decades ago).

Hey, just a 1TB SSD with that budget?!

I'm not a college student, but I work with a college hence I still get the discount.

Honestly I may even sell my desktop, I rarely game these days.

 

[leoneazzurro]

Sorry. My brain isn't thinking straight after watching the Apple event


Such as?

I was speaking about having a better perf/W than Intel, and it's not a mistery that Ryzen mobile atm has a better perf/W in most tasks compared to Intel's offerings.

 

[Panino Manino]

You can put a lot of teraflops on a chip, being able to use them is another matter. Also, there is a process node of advantage.

Apple don't talk about "teraflops".

I believe that Apple could easily support ML upscale, but what about RT? PowerVR used to have RT cores in the past.

 

[jpiniero]

Apple don't talk about "teraflops".

I believe that Apple could easily support ML upscale, but what about RT? PowerVR used to have RT cores in the past.

No RT.

 

[leoneazzurro]

Apple don't talk about "teraflops".

I believe that Apple could easily support ML upscale, but what about RT? PowerVR used to have RT cores in the past.

Well, then it speaks about what?

 

[ultimatebob]

I really want to see some 3rd party benchmarks for the M1 Max, but it sounds like it's going to be the first APU in history where the 3D graphics do not suck. That's a huge milestone!

 

[igor_kavinski]

If there's one bad thing about this launch, Apple still doesn't care about gamers that buy Alienware/Razer sort of high powered gaming laptops. Their focus is still creative professionals. With so much GPU power available, would it have killed them to announce a native AAA title? I mean, everything is there, even a 120Hz display! All they needed was to get a major publisher on board (Take Two comes to mind) and gamers would be rejoicing everywhere right now.