Discussion Apple Silicon SoC thread

Page 151 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Eug

Lifer
Mar 11, 2000
23,579
992
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:

Eug

Lifer
Mar 11, 2000
23,579
992
126
More benchmarks, from Tom's Hardware:


The single core is pretty consistent with previous leaked scores. Not quite 1900.

Screen Shot 2022-09-15 at 10.54.31 PM.png

3D is incremental vs A15, and maintains big lead over Android competition.

Screen Shot 2022-09-15 at 10.54.47 PM.png

Rush shows no improvement from A15. Is Rush hardware accelerated?

Screen Shot 2022-09-15 at 10.55.12 PM.png

Ultimately though, my main interest with these new phones is with the camera. It seems the new bigger pixel-binned 48 MP sensor is a nice boost for low light shots, but the edge detection for portrait photos remains problematic, despite the updated image processor. Whereas the rest of the camera, particularly video, is better than most of the competition, portrait mode edge detection still suffers the same problems that I had with my iPhone 7 Plus 6 years ago. Samsung is clearly in the lead with this particular metric.
 

Eug

Lifer
Mar 11, 2000
23,579
992
126
Multi core:
View attachment 67351

We should see a significant improvement next year with N3E.
This 1887/5455 score has to be from Apple. There are a bazillion iPhone15,x scores on Geekbench now, and this remains one of the best. (iPhone 15,x is, ironically, iPhone 14 series.)

After sifting through a ton of recent Geekbench scores, the fastest single-core and multi-core scores I have found are 1892 and 5499. Oh so close to 1900/5500.

It seems the multi-core vs single-core multiplier is ~2.9X.

EDIT:

1894 and 5511

BTW, Metal is 15886.
 
Last edited:

soresu

Platinum Member
Dec 19, 2014
2,582
1,778
136
Interesting that Apple are switching to using mountains for their core codenames.

An odd decision to make if they do not feel threatened by ARM's future CPU IP considering they switched to using mountains back in 2020 with Matterhorn for X2/A710.
 
Last edited:

soresu

Platinum Member
Dec 19, 2014
2,582
1,778
136
Rush shows no improvement from A15. Is Rush hardware accelerated?
If I remember correctly it benefitted from Adobe converting their 'Mercury' engine from OpenCL to Vulkan code (at least on Android), so I would assume yes.

I would assume that Adobe created equivalent an Metal backend, or at least use MetalVK for the purpose to reuse the Vulkan work.
 
  • Like
Reactions: Eug

soresu

Platinum Member
Dec 19, 2014
2,582
1,778
136
I still wonder why so many key people left. Money is always an incentive, but usually it's not just the money.
Going by Keller I would assume it is simply both the challenge and a change of scenery.

He's never stuck around in one place for very long and it seems like the same can be said of many in the silicon engineering part of the industry.

Those that stick around seem to be 'corporate ladder climbers', while those who move around seem to only increase their value with experience on diverse work portfolio's.
 
  • Like
Reactions: Eug

soresu

Platinum Member
Dec 19, 2014
2,582
1,778
136
AMD is not using a bog standard TSMC process, why should we assume Apple must?
Because typically Apple have been pushing the leading edge of the new nodes, as in pretty much pushing out SoC's using them as soon as risk production is over and bulk production begins.

The extra tweaks AMD needs for their designs have typically come in 2-3 years after the initial node announcement by TSMC, I would not expect that to change drastically just because Apple are willing to pay more.

In fact I believe part of the reason that Apple can buy up so much capacity of new nodes is because they are basically breaking much of the node R&D burden for TSMC - if Apple start getting finicky about wanting more tweaks then it will take TSMC longer to recoup those costs and TSMC would just give the early capacity to someone else to redress the difference, thus losing Apple their historic advantage.
 
  • Like
Reactions: Tlh97 and moinmoin

name99

Senior member
Sep 11, 2010
404
303
136
This 1887/5455 score has to be from Apple. There are a bazillion iPhone15,x scores on Geekbench now, and this remains one of the best. (iPhone 15,x is, ironically, iPhone 14 series.)

After sifting through a ton of recent Geekbench scores, the fastest single-core and multi-core scores I have found are 1892 and 5499. Oh so close to 1900/5500.

It seems the multi-core vs single-core multiplier is ~2.9X.

EDIT:

1894 and 5511

BTW, Metal is 15886.

You might want to add Jetstream2 to your list...

If we believe preliminary numbers, there has been a MASSIVE improvement in JS performance:

claims
A16: 282
A15: 180
A14: 159
A13: 133

I mentioned two patents that came out around the time of the A15 that were written up primarily in the language of accelerating Objective C msgdispatch, but that to my eye could be applied to any interpreted language with a particular type of structure, like JS (or eg Lua, or Python?)

Certainly looks like something happened with the A16! Either the slight HW generalizations I had in mind were implemented, or Safari was improved [and we will also see this change on A15]. My guess is it's HW; if it were just SW the changes are small enough and the payoff large enough that I suspect it would have been done very soon after release if not before.
 
  • Like
Reactions: Tlh97 and Eug

Eug

Lifer
Mar 11, 2000
23,579
992
126
This research company doing iPhone teardowns claims that A16 costs $110, and A15 cost ~$45 last year.

That doesn't seem right to me, but there ya go.


Fomalhaut estimates that production of the iPhone 14 Pro Max calculated in terms of parts cost $501 -- up more than $60 from that of the iPhone 13 Pro Max that went on sale last year. While parts prices for high-end Max models have hovered between $400 and $450 since they debuted in 2018, the latest model represents a steep production cost rise of over $60.

The higher production cost is mainly due to the A16 Bionic chips used in the iPhone 14 Pro and Pro Max models. The proprietary chip costs $110 -- over 2.4 times more than the A15 version used in the iPhone 13 Pro Max released last year.
 

Heartbreaker

Diamond Member
Apr 3, 2006
4,219
5,220
136
This research company doing iPhone teardowns claims that A16 costs $110, and A15 cost ~$45 last year.

That doesn't seem right to me, but there ya go.


Fomalhaut estimates that production of the iPhone 14 Pro Max calculated in terms of parts cost $501 -- up more than $60 from that of the iPhone 13 Pro Max that went on sale last year. While parts prices for high-end Max models have hovered between $400 and $450 since they debuted in 2018, the latest model represents a steep production cost rise of over $60.

The higher production cost is mainly due to the A16 Bionic chips used in the iPhone 14 Pro and Pro Max models. The proprietary chip costs $110 -- over 2.4 times more than the A15 version used in the iPhone 13 Pro Max released last year.

Yeah, not believing that. If the price was that much more, they would have stayed on the old process. It's NOT that much better.
 
  • Like
Reactions: Lodix

Doug S

Platinum Member
Feb 8, 2020
2,191
3,378
136
That's completely crazy. N4 is a slightly cheaper process per mm^2 than N5/N5P and there is no indication the die is all that much larger, unless they've really expanded the ISP or something. I don't see how it costs even $5 more than A15, let alone $70 more. They need some pretty extensive justification for such a wild claim. The main price increase is for LPDDR5 vs LPDDR4X, but they would be costing that out separately.

If that was true, there's no way Apple would be pricing iPhone 14 Pro / Pro Max the same as last year. Apple would not swallow a $60 price increase in overall BOM when they have the pricing power they do.

Sometimes people come to a conclusion and don't even consider the reasons they need to revisit it because of how obviously wrong it is in context. This is one of those times.
 
  • Like
Reactions: Lodix

Eug

Lifer
Mar 11, 2000
23,579
992
126
Speaking of pricing, apparently TSMC wanted to charge Apple 6-9% more per wafer going forward. Apple balked, but then later came to an agreement with them.


Note that Apple makes up 25% of TSMC's revenue. I didn't realize it was that high.
 

Tigerick

Senior member
Apr 1, 2022
559
437
96
Because TSMC said it was designed to reduce costs by reducing process steps. Same reason why N6 costs less than N7 and N3E costs less than N3.
I think you misunderstand how the process works. There are two version of N7: DUV and EUV; EUV version has greatly reduced the mask layer to 4 layers compared to DUV version. But after N7+ (official term for EUV), every increment process like N6 will cost more to make because they are adding more mask layers. N5 can add up to 14 layers and N4 will cost more cause the process adding one more layers.

N3E and N3 are different stories. At first, TSMC introduces N3 process which can go up to 25 layers thus too expensive and also the gap is too wide even for Apple. Thus N3E is introduced with 20 layers mask. Therefore N3E is cheaper than N3 but still more expensive than N4
 

Heartbreaker

Diamond Member
Apr 3, 2006
4,219
5,220
136
This research company doing iPhone teardowns claims that A16 costs $110, and A15 cost ~$45 last year.

That doesn't seem right to me, but there ya go.


Fomalhaut estimates that production of the iPhone 14 Pro Max calculated in terms of parts cost $501 -- up more than $60 from that of the iPhone 13 Pro Max that went on sale last year. While parts prices for high-end Max models have hovered between $400 and $450 since they debuted in 2018, the latest model represents a steep production cost rise of over $60.

The higher production cost is mainly due to the A16 Bionic chips used in the iPhone 14 Pro and Pro Max models. The proprietary chip costs $110 -- over 2.4 times more than the A15 version used in the iPhone 13 Pro Max released last year.


Note all references to the A16 are removed from the story now. I'm betting that was just a wild guess like much of this "analysis" and someone sobered up...
 
  • Like
Reactions: Lodix

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,439
14,409
136
The M1 came out in 2020 and Intel and AMD are still stuck and unable to match its performance per watt. Apple is 5 years ahead of the competition.
Not in the same area. Intel and AMD have multicore (not one or 2) consumer products and server products up to 128 cores.

Apple uses much less cores in most products, and most of them are portable/mobile products where battery life is very important. But for example, see my 6850H laptop results efficiency thread here. Its in the Apple "area" of efficiency.
 

poke01

Senior member
Mar 8, 2022
632
601
106
Not in the same area. Intel and AMD have multicore (not one or 2) consumer products and server products up to 128 cores.

Apple uses much less cores in most products, and most of them are portable/mobile products where battery life is very important. But for example, see my 6850H laptop results efficiency thread here. Its in the Apple "area" of efficiency.
In laptop's where efficiency is the most important. Apple, AMD and Intel use 8P cores in high end laptops. But what Intel and AMD do is boost for even for the tiny things to get maximum pref as this is the only way that x86 makers designed it and this greatly effects battery life.

If we compare the M1 Pro vs the 6800U where they both have 8P cores. The 14" M1 Pro gets over 4hrs more than the Asus 6800U. This cannot be only due to node difference. We can limit the peaks on AMD/Intel but this in turn effects the single core.


1665280119108.png
Take a look at this video, it's in English.