Discussion Apple Silicon SoC thread

Page 9 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Eug

Lifer
Mar 11, 2000
23,714
1,250
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:


M4 Family discussion here:

 
Last edited:

name99

Senior member
Sep 11, 2010
420
312
136
https://www.google.com/search?q=flaws+in+synthetic+benchmarks&oq=flaws+in+synthetic+benchmarks

https://www.spec.org/retired.html

I assume now you have lost all your respect for Spec.org? They decided spec2006 had "become to out-of-date" to be relevant and replaced it with spec2017. The last official spec2006 submission was 2018.

Yes, I have indeed lost respect for you.
Look at WHY Andrei runs 2006 rather than 2017. His reasoning is not secret and he's explained multiple times.

There have also been multiple publications showing that 2006 and 2017 results are extremely correlated. Not perfectly of course, it's good to have some of the more problematic codes removed. But good enough for practical purposes.

Would you prefer no numbers rather than 2006 numbers? I expect you would because then you could complain about how Apple and AnandTech are deliberately refusing to run even SPEC2006 because they are scared of the results.
 

Eug

Lifer
Mar 11, 2000
23,714
1,250
126
And Cinebench is a multicore benchmark. That's fine for adolescents who primarily care about boasting, but the hard problem in CPU design is single-threaded performance. As soon as someone prioritizes MT scores over ST scores I know they're a fool, someone who cares about these scores as a way to measure dick-length not a way to understand the engineering.
(Yes, there ARE important uses for MT scores. But believe me, I can tell from your tone whether or not you understand what they are...)
? Cinebench is both a single-core and a multi-core benchmark.
 
  • Like
Reactions: Tlh97

Leeea

Diamond Member
Apr 3, 2020
3,645
5,379
136
Sorry lads, could barely find room to post with this guys ego getting in the way.

We all like to focus on metrics, but chief this ain’t it.

I used to be a Nvidia crusader two decades ago. No other brand could be better then my deity Nvidia, and its divine creation, the Geforce 256!

Nothing quite like looking into the mirror 20 years later. The flame wars rage unabated, for there will always be true believers.
 
  • Like
Reactions: Tlh97

Doug S

Platinum Member
Feb 8, 2020
2,355
3,731
136
Looks like it probably is the Dev box, hopefully M1 is significantly better.

There is a native version:
"Cinebench R23 now supports Apple’s M1-powered computing systems "


Well it should be at least as better as the difference between A12 and A14, or more if they are clocked higher in the actively cooled Macs.

So is Cinebench single core or multicore? Seems to be some disagreement here, and it can do both which numbers are the ones people are typically quoting? For single core it is a fair test for the M1, for multicore obviously it can't compete against Intel and AMD CPUs with dozens of cores and SMT.

Since the M1 has 4 big and 4 little cores, it is probably roughly equivalent to a 4 core SMT chip so its multithreaded scores would most comparable with Intel and AMD quad core SMT CPUs. Comparing with the high end beasts will have to wait a couple years for the ARM based Mac Pro to arrive.
 

moinmoin

Diamond Member
Jun 1, 2017
4,985
7,751
136
I'm hearing that Geekbench 5 is now universal with native support for M1. I can't seem to find a link for that though.
Shouldn't that be the same as available on App Store already? So https://apps.apple.com/us/app/geekbench-5/id1435082259

The 9880H's TDP is technically 45 W. I guess it would depend on what your definition of DTR is.

They may of course be talking about single thread performance.
That's Intel's definition of TDP (which is more an "at least" for achieving the advertised base frequency on all cores), which can mean PL1/PL2 being set to 145W.
 

Heartbreaker

Diamond Member
Apr 3, 2006
4,243
5,245
136
Well it should be at least as better as the difference between A12 and A14, or more if they are clocked higher in the actively cooled Macs.

So is Cinebench single core or multicore? Seems to be some disagreement here, and it can do both which numbers are the ones people are typically quoting? For single core it is a fair test for the M1, for multicore obviously it can't compete against Intel and AMD CPUs with dozens of cores and SMT.

Since the M1 has 4 big and 4 little cores, it is probably roughly equivalent to a 4 core SMT chip so its multithreaded scores would most comparable with Intel and AMD quad core SMT CPUs. Comparing with the high end beasts will have to wait a couple years for the ARM based Mac Pro to arrive.

Cinebench has both scores, and the picture shown before had both.

Single: 987
Multi: 4530

MT Scaling Ratio: 4.59x - This is less scaling that Intel/AMD 4 Core with SMT, which is usually a bit over 5x.

So it's all about the performance core count. The efficiency cores don't add much.

If that's the locked Devkit with 2.5GHz, then I guess it will depend how fast they can run M1.

Any info about clock speeds for M1?
 
  • Like
Reactions: Tlh97

Eug

Lifer
Mar 11, 2000
23,714
1,250
126
Shouldn't that be the same as available on App Store already? So https://apps.apple.com/us/app/geekbench-5/id1435082259
No, that’s the old Geekbench version. Apparently the new version should be coming soon, unless it’s just an unsubstantiated rumour.


Cinebench has both scores, and the picture shown before had both.

Single: 987
Multi: 4530

MT Scaling Ratio: 4.59x - This is less scaling that Intel/AMD 4 Core with SMT, which is usually a bit over 5x.

So it's all about the performance core count. The efficiency cores don't add much.

If that's the locked Devkit with 2.5GHz, then I guess it will depend how fast they can run M1.

Any info about clock speeds for M1?
Unsubstantiated rumour from CPU-Monkey is 3.1 GHz.
 
Apr 30, 2020
68
170
76
What EXACTLY are the statements that Apple made that have you so incensed?
You have never actually quoted any of them. All you do is insist that other people (not you, you are too smart)
I literally have quoted their statements in my posts. That's what " " marks mean - they indicate I am quoting someone or something.
Want to tell us where your objections lie?
I've already been over every single one of these points, but for the last time before I give up:

Here are their actual statements:
" Our high‑performance core is the world’s fastest CPU core in low‑power silicon. "
- What does this statement mean? Are they saying it's the highest performing "low power silicon" CPU? Is it saying it's the highest performing CPU core period, and it just happens to be made using low-power silicon? Who knows? It's purposely ambiguous.
"Comparison made against the highest-performing CPUs for notebooks commercially available at the time of testing. "
- Notice it doesn't name a specific CPU or say "The highest performing notebook CPU". Them having "CPUs" as plural is pretty suspect. "Highest performing CPUs" could just mean "i7", and still technically be the truth. Them not listing the notebook models is also concerning. Maybe they're talking about the highest performing Chromebook CPU? We don't know.


“World’s best CPU performance per watt”
Qualified with this statement:
"Comparison made against high-performing CPUs for notebooks and desktops, commercially available at the time of testing.
- What does "high performing" mean? A 4700U offers much better perf/watt than a 3800X. Is a 4700u a "high performing" CPU? It will only get about 1/2 the R20 score of a 3800X, but consumes 1/9th the power. Why don't they say "against the highest perf/watt competing processor" ?

"2x faster CPU performance/Matches peak PC performance at 25% of the power"
Qualified with:
"Comparison made against latest‑generation high‑performance notebooks commercially available at the time of testing."
- Once again, what are "high-performance" notebooks? Is it a DTR running a desktop chip at 105w TDP? Is it an "H-Class" notebook chip? Could it be certain "U" class parts like a 4800U? They don't say. Obviously a 4800U offers tremendously better perf-watt than a notebook running a desktop chip. Furthermore, the phrase "latest-generation" is suspect because "latest generation" doesn't mean best. Intel's TGL CPUs are technically the "latest generation" of mobile CPUs in general (since they're newer than Renoir), but still do not offer anywhere near the same MT performance OR efficiency as a 4800U.

Or how about this gem:

"Up to 15 hrs of wireless web browsing"
Qualified with:
"The wireless web test measures battery life by wirelessly browsing 25 popular websites with display brightness set to 8 clicks from bottom."
-
Which websites? What defines "popular?" Why not didn't they use an industry standard website browsing benchmark? What does "8 clicks from the bottom" even mean!? Is "8 clicks from the bottom" the same brightness on a MacBook Air and MacBook Pro? What is it in an industry-standard unit like nits, so I can compare it to another brand's systems?

All of their phrases and qualifications have been chosen very, very carefully. Intel, AMD and nVidia typically call out exact specifications of systems they are comparing against. The only systems Apple called out rather specifically were their own previous gen units, using quite lackluster CPUs. There's a reason why they aren't "naming names" with competing Windows-based systems. Compare that to AMD's recent launches - the Ryzen 5000 series and upcoming Radeon RX 6000 series where they offered hard numbers of their chips, and competitors chips right off the bat, they were later backed up and verified by 3rd party review. That's confidence. Apple has a lot of very wishy-washy qualifications for this new chip that screams to me they don't have the confidence to back those numbers up.
 
Last edited:

John Carmack

Member
Sep 10, 2016
156
248
116
Do people still take Geekbench seriously? Isn't it one of those 1ms benchmarks?

It's benchmarking the A12Z, not the M1.

And Cinebench is a multicore benchmark. That's fine for adolescents who primarily care about boasting, but the hard problem in CPU design is single-threaded performance. As soon as someone prioritizes MT scores over ST scores I know they're a fool, someone who cares about these scores as a way to measure dick-length not a way to understand the engineering.
(Yes, there ARE important uses for MT scores. But believe me, I can tell from your tone whether or not you understand what they are...)

Long term, if Cinebench is especially important to you, what will matter is
- M1, along with every other improvement, has a 4th NEON unit
- I would expect Cinebench can make good use of SVE/2, when Apple finally gives it to us! How much do Cinebench numbers change as you disable AVX512, then AVX going to just SSE?
- Cinebench might (I honestly don't know) be able to make use of the AMX extensions. Those also still have not been activated in XCode. LLVM has been doing a lot of work over the past six months in the handling of matrices (work in which a lot of Apple people have been involved). So the hope is that fairly soon this will be released in XCode and AMX on both A13 and A14 will be available to developers.

Point is, once again, dick-measure if you must. But understanding-wise, even apart from future Apple chips, there's work to be done on getting maximum value from the existing chips, work that's still ongoing and not yet complete. It need to get into the compilers, then the apps need to be recompiled (and perhaps even restructured to make optimal use of new functionality). On the positive side this means that the M1 mac you buy today may grow 10% faster over the next two years or so with successive OS and app updates.

Your post oozes the insecurity that you're projecting on to everyone else.
 

Eug

Lifer
Mar 11, 2000
23,714
1,250
126
Though it probably should be different between the Fanless MBA, MBP with fan, and finally higher heat dissipation of Mini.

At least for sustained MT workloads, cooling should play a big part.
If you look at A12 Geekbench scores, the best iPhone scores are similar to the best iPad scores. However, if you look further, you'll notice that the iPad scores are relatively consistent, but the iPhone scores are all over the map, presumably due to throttling. So, even with that relatively short bursty benchmark, throttling appears to be an issue.

I can get my wife's iPhone XR to consistently score pretty well, but in order to do this I have to make sure the case is off and the phone is cool when I start. No battery charging at the same time, as that significantly warms up the phone.

However, that's an iPhone. What about a MacBook Air with A14X / M1? I dunno, but I will mention my 2017 fanless 12" MacBook Core m3-7Y32 with sustained Cinebench testing. For this test, I ran Cinebench R15, and then when it completed, I immediately started it again. And did this for around an hour or so.

MacBook Core m3.png

As you can see, the performance dropped from 264 down to 253 after 10 runs. However, on the 11th run, it went back up again. Why? Because I was rearranging stuff on the counter and moved the laptop to a different area, not thinking what it may do to the test result. Well, it turns out the new area was much cooler, with the part that was below the laptop before quite warm. That alone was enough to get the performance back up a bit. I then intentionally did the same thing after the 22nd run and got the same result.

I then called on some of our internet friends to do the same with the i5 and i7 models from the same year. I chose 10 runs because that's all my data was truly valid for, and because it too tedious to run it for longer. This is what we got:

MacBook2017-CinebenchR15-m3-wood.png

The m3 (granite) represents my scores on that original granite countertop. m3 (wood) was on a pine dinner table, and as you can see, the performance was consistently lower on that insulating surface after the first run.

Over that half-an-hour, my m3 (granite) dropped from 265 to 246, which represents a drop of 7%. m3 (granite) dropped from 264 to 253, which represents only a drop of 4%. It was regular room temp in the room, probably around 21-22C.

The i5 and i7 did better than mine on wood, but the i5 didn't do as well as mine on granite.

Overall though, while there was a definite drop in performance due to throttling, it wasn't bad at all. For my type of usage, it's a great tradeoff. I'd much rather the laptop get a bit warm but be thin and fanless, than have the machine be relatively cool and blaring like a vacuum cleaner at full speed, but then again, I usually only need bursty speed on my machine. Caveat: The machine has hardware 4K 10-bit HDR HEVC decode support, so even such decodes don't stress the machine, and watching HEVC video won't warm it up that much. (For the Sony Nature Camp demo video, my Kaby Lake MacBook uses about 25% CPU, whereas my Kaby Lake iMac i5-7600 uses less than 10% CPU. In contrast, the 2015 Skylake Core i7-6700K iMac uses 100% CPU and still can't consistently decode it cleanly.)

However, while my machine did pretty well, not all units will fare so well. For example, Notebookcheck did similar testing on their 2017 Core m3 MacBook, and got much worse results.


Screen Shot 2020-11-11 at 6.54.10 PM.png

Their MacBook's performance dropped a lot quicker, and then from time to time the performance fell into the basement.

With the MacBook Pro you should expect fairly consistent performance at the top of the M1 performance curve. With the Air in some cases you'll get pretty decent performance, I guessing probably relatively close to the Pro, but in other cases the Air may behave erratically.

ie. If you're going to be running Final Cut a lot on it, get the Pro, but if you're going to be taking notes and writing papers, with only very occasional video editing, get the Air, unless you absolutely need the extra battery life of the Pro.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Looking at Apples marketing, it is all "up to" speak and comparisons to old intel units. From what I could see, the old intel CPUs they were comparing against did not even break 2 gigahertz. The game they showed off is said to have been on low detail with noticeable frame jumps.

The comparisons were against Icelake. You know, the chip with notoriously low base clocks?

In Turbo they go far beyond that.

And everyone should know x86 isn't the problem, its the screwups of both x86 companies. If it was just the ISA why are they also 2+ years behind in graphics?
 
Last edited:

IvanKaramazov

Member
Jun 29, 2020
56
102
66
Grain of salt, as always.


EDIT: And here's the Mini. Some variation (which might lend credence) and also the suggestion that the peak power of the MBA and the actively cooled are similar, though expect the fanless version to throttle similarly to an iPad of course.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
EDIT: And here's the Mini. Some variation (which might lend credence) and also the suggestion that the peak power of the MBA and the actively cooled are similar, though expect the fanless version to throttle similarly to an iPad of course.

The MT scores are similar to Ryzen 7 4800U.

Granted, in Geekbench, Intel's Amberlake gets scores close to the 25W U parts in single thread(and only in Geekbench, 30% behind in ST in everything else), so you cannot tell from the score whether the MBA is throttling relative to the Mac Mini.
 
  • Like
Reactions: Tlh97

Eug

Lifer
Mar 11, 2000
23,714
1,250
126
No, that’s the old Geekbench version. Apparently the new version should be coming soon, unless it’s just an unsubstantiated rumour.
The one on the App Store is 5.25. The Geekbench scores out in the wild now are 5.3. I think the scores are legit.

Holy shit. 1687 single-core and 7433 multi-core for a frickin' fanless MacBook Air!

Screen Shot 2020-11-11 at 8.00.32 PM.png

Unsubstantiated rumour from CPU-Monkey is 3.1 GHz.
Geekbench says 3.2 GHz.
 

Eug

Lifer
Mar 11, 2000
23,714
1,250
126
Interesting. So the M1 is clocked 100MHz higher?
Dunno. A14 is 3 GHz, and only CPU-Monkey had the 3.1 GHz spec for A14X, which they then quoted for M1.

However, it would make sense. Apple has binned before based on just a 100 MHz speed difference.

For example, the iPad mini 2 is a 1.3 GHz A7, but the regular iPad Air is a 1.4 GHz A7.
 
Apr 30, 2020
68
170
76
Appending .gb5 shows the Air is sustaining the bench at 3.2GHz, while the Mini is running it at around 3.0 GHz.

However something concerning to me is that GB 5.3 "ARM" is using CLANG 12.0 as it's compiler, while GB5.3 x86 is using CLANG 9.0. CLANG9.0 is pretty old, and the LLVM compiler project has added some significant changes since (especially in regards to x86 and Zen). I am not sure why they wouldn't be using the same compiler version on both.
 
  • Like
Reactions: Tlh97

Eug

Lifer
Mar 11, 2000
23,714
1,250
126
Appending .gb5 shows the Air is sustaining the bench at 3.2GHz, while the Mini is running it at around 3.0 GHz.

However something concerning to me is that GB 5.3 "ARM" is using CLANG 12.0 as it's compiler, while GB5.3 x86 is using CLANG 9.0. CLANG9.0 is pretty old, and the LLVM compiler project has added some significant changes since (especially in regards to x86 and Zen). I am not sure why they wouldn't be using the same compiler version on both.
Dunno about Windows and Linux, but for macOS, my (very) n00b understanding is that they just compile Geekbench with whatever compiler version is included with the latest version of Xcode.
 

shady28

Platinum Member
Apr 11, 2004
2,520
397
126
If you look at A12 Geekbench scores, the best iPhone scores are similar to the best iPad scores. However, if you look further, you'll notice that the iPad scores are relatively consistent, but the iPhone scores are all over the map, presumably due to throttling. So, even with that relatively short bursty benchmark, throttling appears to be an issue.

I can get my wife's iPhone XR to consistently score pretty well, but in order to do this I have to make sure the case is off and the phone is cool when I start. No battery charging at the same time, as that significantly warms up the phone.

However, that's an iPhone. What about a MacBook Air with A14X / M1? I dunno, but I will mention my 2017 fanless 12" MacBook Core m3-7Y32 with sustained Cinebench testing. For this test, I ran Cinebench R15, and then when it completed, I immediately started it again. And did this for around an hour or so.

View attachment 33647

As you can see, the performance dropped from 264 down to 253 after 10 runs. However, on the 11th run, it went back up again. Why? Because I was rearranging stuff on the counter and moved the laptop to a different area, not thinking what it may do to the test result. Well, it turns out the new area was much cooler, with the part that was below the laptop before quite warm. That alone was enough to get the performance back up a bit. I then intentionally did the same thing after the 22nd run and got the same result.

I then called on some of our internet friends to do the same with the i5 and i7 models from the same year. I chose 10 runs because that's all my data was truly valid for, and because it too tedious to run it for longer. This is what we got:

View attachment 33646

The m3 (granite) represents my scores on that original granite countertop. m3 (wood) was on a pine dinner table, and as you can see, the performance was consistently lower on that insulating surface after the first run.

Over that half-an-hour, my m3 (granite) dropped from 265 to 246, which represents a drop of 7%. m3 (granite) dropped from 264 to 253, which represents only a drop of 4%. It was regular room temp in the room, probably around 21-22C.

The i5 and i7 did better than mine on wood, but the i5 didn't do as well as mine on granite.

Overall though, while there was a definite drop in performance due to throttling, it wasn't bad at all. For my type of usage, it's a great tradeoff. I'd much rather the laptop get a bit warm but be thin and fanless, than have the machine be relatively cool and blaring like a vacuum cleaner at full speed, but then again, I usually only need bursty speed on my machine. Caveat: The machine has hardware 4K 10-bit HDR HEVC decode support, so even such decodes don't stress the machine, and watching HEVC video won't warm it up that much. (For the Sony Nature Camp demo video, my Kaby Lake MacBook uses about 25% CPU, whereas my Kaby Lake iMac i5-7600 uses less than 10% CPU. In contrast, the 2015 Skylake Core i7-6700K iMac uses 100% CPU and still can't consistently decode it cleanly.)

However, while my machine did pretty well, not all units will fare so well. For example, Notebookcheck did similar testing on their 2017 Core m3 MacBook, and got much worse results.


View attachment 33648

Their MacBook's performance dropped a lot quicker, and then from time to time the performance fell into the basement.

With the MacBook Pro you should expect fairly consistent performance at the top of the M1 performance curve. With the Air in some cases you'll get pretty decent performance, I guessing probably relatively close to the Pro, but in other cases the Air may behave erratically.

ie. If you're going to be running Final Cut a lot on it, get the Pro, but if you're going to be taking notes and writing papers, with only very occasional video editing, get the Air, unless you absolutely need the extra battery life of the Pro.

Nice. Esp the part about benchmarking a laptop on a granite counter vs a wood table.

Chaos Theory:

"Small differences in initial conditions, such as those due to errors in measurements or due to rounding errors in numerical computation, can yield widely diverging outcomes for such dynamical systems, rendering long-term prediction of their behavior impossible in general.
[6] This can happen even though these systems are deterministic, meaning that their future behavior follows a unique evolution[7] and is fully determined by their initial conditions, with no random elements involved.[8] In other words, the deterministic nature of these systems does not make them predictable. "
 

Roland00Address

Platinum Member
Dec 17, 2008
2,196
260
126
It's official folks! 1700+ single-core for the MacBook Pro. However, the multi-core is actually noticeably worse in this one test. :p


Geekbench 5.3: 1714 / 6802

Spectacular, spectacular
No words in the vernacular
Can describe this great event
You'll be dumb with wonderment
...
So EX-CIT-ING, the audience will stomp and cheer!
 
  • Haha
Reactions: Tlh97 and Eug