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

 

[amosliu137]

here is the spec2017 int test by JHT5132 in baidu tieba with additional cooling (icebag)
整机 means total power
减空载 means total power minus idle power. By the way, my test score of M3 pro is lower than 10

 

[poke01]

I'm sure that they are planning something big for the macbooks since their competition will be Zen 5 and Lunar Lake and even Strix Halo once the M4 Macbooks are released. They wouldn't like their shiny new CPU underperforming against those. I'm guessing they may do something extreme and even desperate (fancy cooling solution to hit high clocks etc.). The macbook M4 may not even be the same die as the iPad one. It may have bigger caches or other tweaks to really push the computing throughput. Starting from June, things are getting very, very interesting.

If they didn't have this mentality of wanting Apple users to own multiple Apple devices, they could've allowed the iPad Pro to at least run the normal MacOS or dual boot it with the iPadOS. All that power is simply wasted if it can't be used for desktop computing. At least the current M1/M2/M3 users would've gotten an immediate and overwhelming urge to buy the M4 iPad Pro and lighten their traveling load as well as enjoy a boost in their workloads. Apple, the company that refuses to do geeky stuff and instead tries to make computing exclusive and stylish, with obvious drawbacks like too much computing power in the hands of a lot of wealthy individuals with no idea of what to do with it other than mundane stuff. Ironically, despite packing the most computing power per square milimeter, Apple CPUs could be the most underutilized silicon in the ENTIRE history of computing. God bless corporate America and their brilliant business types!

The masses will have to wait maybe 5 years for these to be available on the used market and maybe then their true potential will finally be realized and they will most likely be used by people actually grateful for the computing power of these CPUs, using them for productive tasks that matter, rather than reading corporate emails, creating flashy but containing corporate speak gibberish presentations and planning corporate political strategies to one-up their colleagues and competitors and doing their part to make humanity miserable by destroying lives.

I agree with you Igor, most Apple's users and reviewers have been critical of the iPad Pro. It has excellent hardware but lacklusture software. It's got BMW hardware but has Honda software, no wonder iPad's sales are declining(Why upgrade when the last 3 iPad Pros do the same thing). There is no proper software to take advantage of this hardware.
At best its great marketing for their chips, thats it.

For the last 14 years, the iPad has been largely stale because of its iOS roots and while some may like iPadOS when you spend $999+, it's time for the software to also be "Pro" and not have the same Operating System as a $349 iPad.

 

[poke01]

https://twitter.com/x/status/1790625130428907677

I want this display on a Macbook. In the pic above 2018 vs M4.

 

[DavidC1]

I wonder how they achieved such an increase on FP. I don't think there were any changes to the backend?

You can have positive improvements in FP and zero improvements in Integer, but positive improvements in Integer always brings improvements in FP, because uarch changes are aimed at Integer performance which is the hard part. This is CPU architecture basics.

FP has basis on an accelerator because prior to 486 chips they were indeed accelerators and off the chip.

So naturally anything that improves Integer will improve FP by a similar amount(often more), because improving Integer is akin to widening the lanes or increasing speed limit on a highway.

They had it a lot easier a decade ago when they were coming at x86 from way behind and large yearly gains were there for the taking.

Ok, but they also started flatlining after Gerald Williams III and other architects left. CPUs aren't plants and grow all on their own.

 

[roger_k]

You can have positive improvements in FP and zero improvements in Integer, but positive improvements in Integer always brings improvements in FP, because uarch changes are aimed at Integer performance which is the hard part. This is CPU architecture basics.

FP has basis on an accelerator because prior to 486 chips they were indeed accelerators and off the chip.

So naturally anything that improves Integer will improve FP by a similar amount(often more), because improving Integer is akin to widening the lanes or increasing speed limit on a highway.

Thank you for the explanation!

Ok, but they also started flatlining after Gerald Williams III and other architects left. CPUs aren't plants and grow all on their own.

I don't see any evidence of that. As often noted, improvements in Apple Silicon since A10 were linear, not exponential (they only looked exponential because it happed to be in the part of the curve where +300 GB points corresponded to 20% growth). And they were still linear past A14/M1. In fact M3 to M4 was the biggest jump in GB6 in a long while, even if we remove the SME-accelerated tests.

On the other hand I don't see GWIII and others doing any better outside Apple either. Talent is one thing, culture and resources is another one. Apple appears as agile as any startup when it comes to developing their IP, and they have many more resources.

 

[DavidC1]

On the other hand I don't see GWIII and others doing any better outside Apple either. Talent is one thing, culture and resources is another one. Apple appears as agile as any startup when it comes to developing their IP, and they have many more resources.

It takes more than brainpower for sure. The culture of the company has to be right and the teams have to work in harmony, because bunch of geniuses fighting won't do any better than average that don't.

But we know for a fact their big gains simply stopped after he left and have been coasting on the leftover since.

 

[eek2121]

https://twitter.com/x/status/1790625130428907677

I want this display on a Macbook. In the pic above 2018 vs M4.

Good news, you somewhat get your wish: https://support.apple.com/en-us/102597

New Macbooks with the OLED tech are expected in the future.

 

[roger_k]

But we know for a fact their big gains simply stopped after he left and have been coasting on the leftover since.

Again, I don't see that as a fact at all. From A10 to A12 Apple gained approximately 700 points in GB6. From A12 to A14 it was 800. From M1 to M3 it's again 800 points. From M2 to M4 it's a whopping 1100 points (let's say 900 if we remove SME).

Where do you see the gains slowing down?

 

[poke01]

Nice to nice the thermal improvement vs M2 iPP. Apple really did deliver more performance while making it thinner. Amazing, these would be the best computer on the market if it dual booted macOS. Ahh!

https://twitter.com/x/status/1790695989940265221

 

[Eug]

Nice to nice the thermal improvement vs M2 iPP. Apple really did deliver more performance while making it thinner. Amazing, these would be the best computer on the market if it dual booted macOS. Ahh!

https://twitter.com/x/status/1790695989940265221

Yes, quite impressive. Dave2D got the same result.

Page 274 - Discussion - Apple Silicon SoC thread

Page 274 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

PS. The new iPad Pro has a metal piece inside that I believe they are using both as a heat spreader and as a cross brace to increase rigidity (since the thing is so crazy thin).

 

[amosliu137]

Finally, Geekerwan break the wall of 4000 .

iPad16,3 - Geekbench

Benchmark results for an iPad16,3 with an ARM processor.

browser.geekbench.com

https://browser.geekbench.com/v6/cpu/6107586
It is 9 core M4

 

[Eug]

Finally, Geekerwan break the wall of 4000 .

iPad16,3 - Geekbench

Benchmark results for an iPad16,3 with an ARM processor.

browser.geekbench.com

View attachment 98980
https://browser.geekbench.com/v6/cpu/6107586
It is 9 core M4

They did it! And in an iPad!

And no overclocking involved of course. 9-core model too.

 

[poke01]

https://twitter.com/x/status/1788175880138977621

That’s actually a good use case.

 

[Nothingness]

here is the spec2017 int test by JHT5132 in baidu tieba with additional cooling (icebag)View attachment 98977
整机 means total power
减空载 means total power minus idle power. By the way, my test score of M3 pro is lower than 10

Thanks for this. Assuming this is using the same binaries as previously, here we go:

 

[Eug]

Finally, Geekerwan break the wall of 4000 .

iPad16,3 - Geekbench

Benchmark results for an iPad16,3 with an ARM processor.

browser.geekbench.com

View attachment 98980
https://browser.geekbench.com/v6/cpu/6107586
It is 9 core M4

Could you provide a link to the where the pic was posted? I’d love to read the description of their method, and the news sites might be interested as well. By the sounds of it, they have a few different models they are testing.

BTW, 13595 is the fastest M4 9-core multi-core to date. Freezer guy’s 14924 score is the fastest 10-core score to date, which is 10% faster.

 

[GC2:CS]

While the performance of M4 seems great I am a bit concerned that the specint pulls some 10 W for a single core. I rememver when A13 was concerning when pulling 6W yet it was the best perf/batt balance by far. Even today I would say it is cimpetitive. How is possible fire power to go up but the battery life to keep up ?

It makes sense for the core to be able to go high power in desktop but why allow it in a tablet/phone ? Who has an icepack phone acessory ?

What about someone running Gfx bench ?

Regarding oled it is nice but I am turned off by PWM and supposedly higher reflectivity of the tandem panel. And call me crazy but I think that the lower, softer contrast feels easier on the eyes especially at night.

 

[SarahKerrigan]

Thanks for this. Assuming this is using the same binaries as previously, here we go:


View attachment 98984

Seeing a big boost there on Perlbench but not on gcc is really odd. They usually track each other a little bit.

 

[roger_k]

While the performance of M4 seems great I am a bit concerned that the specint pulls some 10 W for a single core. I rememver when A13 was concerning when pulling 6W yet it was the best perf/batt balance by far. Even today I would say it is cimpetitive. How is possible fire power to go up but the battery life to keep up ?

Keep in mind that these scores are obtained under extreme cooling. That is not how the chip is designed to operate in every day condition (unless you are a researcher in the Antarctic). They are still useful since they allow us to pin the operating frequency for more accurate iso-clock gain estimates. But a regular iPad will likely run at a slightly lower frequency and much more manageable wattage.

 

[Eug]

Keep in mind that these scores are obtained under extreme cooling. That is not how the chip is designed to operate in every day condition (unless you are a researcher in the Antarctic). They are still useful since they allow us to pin the operating frequency for more accurate iso-clock gain estimates. But a regular iPad will likely run at a slightly lower frequency and much more manageable wattage.

It’s gonna be sooooo smooth watching Netflix and surfing AnandTech on this tablet.

 

[amosliu137]

Could you provide a link to the where the pic was posted? I’d love to read the description of their method, and the news sites might be interested as well. By the sounds of it, they have a few different models they are testing.

BTW, 13595 is the fastest M4 9-core multi-core to date. Freezer guy’s 14924 score is the fastest 10-core score to date, which is 10% faster.

微博

随时随地发现新鲜事！微博带你欣赏世界上每一个精彩瞬间，了解每一个幕后故事。分享你想表达的，让全世界都能听到你的心声！

m.weibo.cn

 

[Eug]

微博

随时随地发现新鲜事！微博带你欣赏世界上每一个精彩瞬间，了解每一个幕后故事。分享你想表达的，让全世界都能听到你的心声！

m.weibo.cn

Thanks. Is that Geekerwan's Weibo, or someone else's? The reason I ask is because I don't see any mention of Geekerwan in that post.

 

[SpudLobby]

For a CPU that's only 8 months further along the development timeline, and getting a bigger clock boost than the process improvement allowed for I don't think its that bad. I'm honestly mystified that Apple released the fastest CPU on the planet for single thread in a freaking 5mm tablet and people are underwhelmed. I guess they set the bar for improvement too high in the past, but diminishing returns are a thing. They had it a lot easier a decade ago when they were coming at x86 from way behind and large yearly gains were there for the taking.

Honestly I was more underwhelmed at first with the 3500-3600 stuff and SME, now we’re seeing 4000, which implies some very different stuff.

 

[amosliu137]

Thanks. Is that Geekerwan's Weibo, or someone else's? The reason I ask is because I don't see any mention of Geekerwan in that post.

That is his personal account.

 

[Nothingness]

Seeing a big boost there on Perlbench but not on gcc is really odd. They usually track each other a little bit.

Yes, that's strange. Could it be that the ice bag started melting?

Joke aside we don't know if that was enough to maintain clock speed during all the run of each test. I don't have data handily available, but it might be worth comparing the run time of each test.

 

[Nothingness]

Keep in mind that these scores are obtained under extreme cooling. That is not how the chip is designed to operate in every day condition (unless you are a researcher in the Antarctic). They are still useful since they allow us to pin the operating frequency for more accurate iso-clock gain estimates. But a regular iPad will likely run at a slightly lower frequency and much more manageable wattage.

As I wrote above, I'm not sure it's known frequency throttling didn't happen anyway.