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

 

[igor_kavinski]

Only to be bought by Qualcomm and return to laptop chips

QC has probably promised them they can do whatever they want if they deliver them a decent laptop chip and probably put other restrictions in place like stock options maturity etc. to prevent them from leaving.

 

[The Hardcard]

Which is going to be a very interesting squeeze. They can't get any more IPC since all the IPC makers apparently left for Nuvia.
We're going to see a very wild switcheroo where QC will provide M1-like chips, for cheaper, and go to M2/3/4 pretty quickly without raising area or frequency too much.
While Apple will have to eat through their margins for area or admit that fans are necessary again because freq has gone up too much.

https://twitter.com/x/status/1787959708156129512

Mmmmmh.

Did IPC makers leave for Nuvia? I haven’t seen evidence that Oryon has significantly higher IPC. The house benchmarks put it about the same.

As far as future chips from anyone, no passing Apple IPC chickens have hatched yet at Qualcomm, AMD, Intel, Mediatek, ARM or Nvidia. I don’t get the urge to count them now.

 

[Nothingness]

Did IPC makers leave for Nuvia? I haven’t seen evidence that Oryon has significantly higher IPC. The house benchmarks put it about the same.

They had to rebuild a chip from scratch, with good ideas to start from of course (except for ideas that were patented during their careers at Apple of course). That takes time. Only with the upcoming generations we'll be able to see if they are able to increase IPC significantly starting from their new design.

But again, I wonder if we have not reached a plateau where every high performance teams will land and from which extracting more IPC will come with a high price in power. That's why I'm very curious to see Zen5; I have high hopes it will be very good and I hope I won't be disappointed (I'm kind of concerned the desktop variant might be power hungry and I'd really like to be wrong).

 

[Mahboi]

Or maybe it's just difficult to make good IPC increases if one already leads the market in IPC by a large margin. I don't think the "lost team" argument has much merit.

Ok then please explain less than 10% IPC growth in 3 years. That makes the Skylake era look good.

Apple is still delivering very good performance improvements, and they had substantial IPC improvements (just not on all possible workloads) since A14. I don't really see any momentum loss here.

Wut?
Every single Apple chip has taken more and more area. They're all more expensive to produce than the last. M4 is just M3 but bigger, M2 is M1 but bigger, etc. There is no strong rework, it's tweaks and bigger area (and higher frequencies), which ofc at the prices they sell, Apple can afford, BUT that's only because there is no competitor on the market.

What is this story about their "huge area cores" though? The A17 Pro P-cores are smaller than AMD's Zen4 and less than half the size of Intel's ores.

Yes I would expect the phone cores to not be the size of a laptop core. How about comparing M4 to Zen 4?

I'm not quite sure what you are saying here. Apple's performance improvements have been solid and on time. And they seem to balance the IPC, performance, clock, and power consumption well. If you argue that they are in trouble unless they can improve the IPC, you would need to demonstrate that there is some other product that can surpass their dominance in this segment.
I am curious about the new Cortex cores for example. If ARM managed to match the A17 pro performance on a similar power budget, that would lend your argument more credibility. Of course, we'd need to get some clear performance figures. ARM's slides have me headache.

Computex is in 5 days.

 

[Mahboi]

Did IPC makers leave for Nuvia?

It sure hasn't been growing at Apple anymore.

 

[FlameTail]

QC has probably promised them they can do whatever they want if they deliver them a decent laptop chip and probably put other restrictions in place like stock options maturity etc. to prevent them from leaving.

There were ominous rumours that some engineers did already leave QC.

 

[The Hardcard]

What he means is that solely depending on ARM is kind of digging a hole for themselves. What if ARM innovation hits a wall? Then they will have to expend extra resources to switch to some other architecture later on. Maintaining two ISAs with x86 models priced lower could've been one way to prevent future issues with ARM IPC scaling.

All IPC advances in modern chips is happening outside of ISAs. ISA gets you virtually nothing today, it also loses you virtually nothing. IPC currently is about conquering the memory wall with branch prediction, cache structure, and keeping instructions in flight.

There was a time when x86 ISA was a liability, but AMD and Intel massively mitigated that years ago. Structures that today handle x86 idiosyncrasies are effective, low power, and tiny in die area.

Apple’s IPC success has zero to do with ARM ISA. It is because they built a cache design and combined it with branch prediction and a massive reorder architecture that keeps more instructions in flight than anyone else, thus doing the most to overcome the memory wall. It is as simple as that.

 

[xiewe3wq]

The narrative that Apple is stagnating has been going around since before the M1 was released. What happened in those 5 years? Did other companies also stagnate? Why is still no company close to surpassing Apple's SoC? The highly anticipated Oryon cores appear to be slower than the M3, while using more power.

 

[roger_k]

Ok then please explain less than 10% IPC growth in 3 years.

8 out of 15 GB6 subtests show 5% or higher iso-clock improvement just from M3 to M4, four subtets over 10% improvements — this is excluding SME. Didn't look at M1 vs M4. I think the improvements are there, and they are decent, it just becomes much more nuanced. Again, Apple is so far ahead in the IPC game that extracting more gains is goign to get more and more difficult.

Wut?
Every single Apple chip has taken more and more area. They're all more expensive to produce than the last. M4 is just M3 but bigger, M2 is M1 but bigger, etc. There is no strong rework, it's tweaks and bigger area (and higher frequencies), which ofc at the prices they sell, Apple can afford, BUT that's only because there is no competitor on the market.

Are you talking about the chip or are you talking about the cores? I though we are discussing CPUs? Because the A17 Pro P-cores were measured around rougly, 2.2mm2 if I remember correctly (that's smaller than previous Apple cores). The die area grows because of GPU cores, caches, and all other IP Apple packs on there.

Yes I would expect the phone cores to not be the size of a laptop core. How about comparing M4 to Zen 4?

Apple uses the same CPU core across the entire family. A17 Pro cores are the same as M3 cores. I haven't seen an M4 die analysis yet.

Computex is in 5 days.

Looking forward!

 

[The Hardcard]

There were ominous rumours that some engineers did already leave QC.

It may not be ominous. Am I the only one noticing that these visionary-level chip architects have been bopping around companies for decades now?

They are hired to set the path and often rarely stay to see the company complete that path. When is the last time Keller was at a still at a company to see his design path even hit the market?

A lot of these guys go back and forth between AMD, Intel, Nvidia, IBM and Apple, with some startup work in between. It doesn’t appear these are the guys to keep.

 

[Eug]

.

It may not be ominous. Am I the only one noticing that these visionary-level chip architects have been bopping around companies for decades now?

They are hired to set the path and often rarely stay to see the company stay on that path. When is the last time Keller was at a still at a company to see his design path even hit the market?

A lot of these guys go back and forth between AMD, Intel, Nvidia, IBM and Apple, with some startup work in between. It doesn’t appear these are the guys to keep.

I read somewhere that even if you ignore the “fun” aspect of going to a new company, the best way for some engineers to make bank is to leave every several (five to ten?) years to a new company, to get the new signing bonuses and after the stock options have matured.

 

[The Hardcard]

It sure hasn't been growing at Apple anymore.

Point being, they didn’t grow it at their new location either. They, as the architects, know exactly where and why Apple’s IPC is what it is - five years after they left they made no advances. Where’s the evidence that Apple’s IPC was limited by them leaving? Since there’s no sign of it in Oryon, do you know about a different chip they are also working on?

 

[FlameTail]

Point being, they didn’t grow it at their new location either. They, as the architects, know exactly where and why Apple’s IPC is what it is - five years after they left they made no advances. Where’s the evidence that Apple’s IPC was limited by them leaving? Since there’s no sign of it in Oryon, do you know about a different chip they are also working on?

Oryon V2

 

[Mopetar]

Considering that Apple released M4 only 8 months after M3 and it's more than just a port over to a different node, I think any statements about their chip design team being incapable are vastly over-exaggerated.

 

[Nothingness]

QC has probably promised them they can do whatever they want if they deliver them a decent laptop chip and probably put other restrictions in place like stock options maturity etc. to prevent them from leaving.

Given the fate of Falkor, I doubt the Apple guys would have believed Qualcomm promise.

 

[FlameTail]

Considering that Apple released M4 only 8 months after M3 and it's more than just a port over to a different node, I think any statements about their chip design team being incapable are vastly over-exaggerated.

Exactly.

And the people having consternation are only talking about the P-core.

Why will they not talk about the E-core, where Apple has been making advances every year?

It is the world's most efficient E-core. Nothing is going to touch it anytime soon.

 

[igor_kavinski]

Another possibility is Apple just being lazy. Why bother leapfrogging when the competition is lacking?

 

[FlameTail]

Another possibility is Apple just being lazy. Why bother leapfrogging when the competition is lacking?

They swapped the coffee for smoothies!

 

[igor_kavinski]

Considering that Apple released M4 only 8 months after M3 and it's more than just a port over to a different node, I think any statements about their chip design team being incapable are vastly over-exaggerated.

Without more information, that tells me M3 was delayed

 

[igor_kavinski]

They swapped the coffee for smoothies!

Killing their architects with love

 

[The Hardcard]

They had to rebuild a chip from scratch, with good ideas to start from of course (except for ideas that were patented during their careers at Apple of course). That takes time. Only with the upcoming generations we'll be able to see if they are able to increase IPC significantly starting from their new design.

But again, I wonder if we have not reached a plateau where every high performance teams will land and from which extracting more IPC will come with a high price in power. That's why I'm very curious to see Zen5; I have high hopes it will be very good and I hope I won't be disappointed (I'm kind of concerned the desktop variant might be power hungry and I'd really like to be wrong).

Exactly, so why a discussion about a problem at Apple when we don’t know what it is or even if there is one? It is weird to worry about what remains the best.

Plus, I still have the question of whether Apple is even done with this first design. Not only does analysis from Andrei and Geekerwan (to the extent they can be relied on) indicate constant changes in Apple microarchitecture, Geekerwan’s measurements show back and forth sizing of reorder structures.

Question being, is the first Apple architecture really done yet? I see a possible no to that, at least as of M3.

 

[FlameTail]

This Semianalysis article came out 3 years ago. Amazing how well it aged.

Apple CPU Gains Grind To A Halt And The Future Looks Dim As The Impact From The CPU Engineer Exodus To Nuvia And Rivos Starts To Bleed In

Apple's leading CPU team looks to be struggling as they have now delivered no CPU gains for the first generation ever. Apple likely had to delay their next core architecture due due to personnel overturn from the Nuvia exodus. Furthermore, with a new exodus surrounding Rivos, the future doesn't...

www.semianalysis.com

 

[Mopetar]

Another possibility is Apple just being lazy. Why bother leapfrogging when the competition is lacking?

How well did that work out for Intel?

If the competition is that far back, see if you can lap them.

 

[igor_kavinski]

Why will they not talk about the E-core

E-core is like the maid. Essential but do you really care how skinny it is?

 

[The Hardcard]

Oryon V2

You’ve seen IPC measurements on that? What indicates it will have a significant IPC bump outside of your foundationless hopes and dreams? Are there actual current facts that establish any chip as having higher IPC than Apple?