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

 

[moinmoin]

Also note that that's the share by revenue. While I'm sure non-Pro Macs will have the majority of the share by units, by revenue the numbers may well flip if not move more toward the Pro Macs.

 

[Eug]

First 8-core (6+2) M1 Pro on Geekbench:

1767 / 9948

Roughly 1/3rd faster than M1.

BTW, it appears to be downclocked, if the listed speed is accurate. So, Apple appears to be possibly binning by clock speed as well, at least for this specific 8-core CPU 14-core GPU part. That's fine by me, if it runs even cooler.



M1 Pro 14-core GPU (OpenCL)



M1 Max 24-core GPU (OpenCL)

 

[Semel]

M1 Max 24-core GPU (OpenCL)

3080 (mobile) can reach 135000 points in this test...

 

[Eug]

M1 Max 24-core GPU (OpenCL)

3080 (mobile) can reach 135000 points in this test...

Yes, Nvidia historically does very well with this test, but I should point out two caveats:

1. The M1 Max you quoted is a cut down version, missing 25% of its compute units.

2. There are many variations of the 3080, even on mobile. What one are you specifically talking about?

 

[Doug S]

iPhone revenue is ginormous, but your 1% claim is off. Mac revenue is sometimes over 20% that of iPhone revenue. Are you suggesting that M1 Pro / Max (or similar future chips) will be less than 5% of Mac sales?

I may have been overstating my case, but from what I understand Macbook Air has a pretty large share of Mac revenue, so M1 Pro / M1 Max will be a low to at most mid single digit percentage versus iPhone.

So my point still stands that Apple's design focus is on the iPhone because that's where they make so much more revenue. Maybe the power/performance combo of the ARM Macs enables them to grow Mac marketshare versus the PC and justifies truly separate designs. From a technical standpoint a design targeted at higher performance would be cool to see, and as a shareholder seeing them choose to do so because of Mac marketshare growing would also be cool to see... I'm skeptical though - I think the Mac vs PC relative share isn't likely to change all that much no matter how good Mac hardware. Both due to its much higher price point versus the PC and because of how ubiquitous Windows is.

It is kind of like iPhone in that respect, which can never get much more than the ~15% of market share it has because Android covers the whole market all the way down to the low end, and Apple covers only the high end with a token midrange SE device.

 

[LightningZ71]

Just keep in mind, while the GPU is largely a scaled up version of their mobile one, the M1 is not just a CPU and memory controller, there ae multiple spe I'll purpose fixed function units to accelerate specific classes of workloar that make it punch well above its weight in those functions.

 

[Eug]

Just keep in mind, while the GPU is largely a scaled up version of their mobile one, the M1 is not just a CPU and memory controller, there ae multiple spe I'll purpose fixed function units to accelerate specific classes of workloar that make it punch well above its weight in those functions.

As I mentioned before (and I don't mean to harp on this but I do think it bears repeating), some of the hardware accelerated stuff is different on the M1 Pro/Max than on the M1. Some of the complaints about pro workflows on M1 have been addressed on M1 Pro/Max. Yes memory of course, but also specifically the addition of ProRes acceleration basically means it now competes against the high end Intel Mac Pro with Afterburner.

Apple's Afterburner card alone costs US$2500 aftermarket, and you need a Mac Pro on top of that just to use that card. (The card is $2000 if you buy it with the Mac Pro.)

Apple Afterburner Card

Created to transform the workflow for film and video professionals, Afterburner accelerates ProRes and ProRes RAW video codecs in Final Cut Pro X, QuickTime Player X and supported third-party applications. A PCI Express card exclusively for Mac Pro, Afterburner can be installed in any...

www.apple.com

M1 Max:
Edit up to 40 streams of 4K ProRes 422
Edit up to 7 streams of 8K ProRes

"That's more streams than on a 28-core Mac Pro with Afterburner"
- Shruti Haldea, Mac Product Line Manager

---

I just spec'd out a 96 GB Mac Pro 28-core, Radeon Pro W5700X 16 GB, 4 TB SSD, and Afterburner. Only $17999! (There is is no 64 GB model. It jumps from 48 GB to 96 GB.)

However, the 16" M1 Max 32-core with 64 GB RAM and 4 TB SSD is $4899.

BTW, it should be noted that M1 Max has two ProRes accelerators that give it this functionality. M1 Pro has "only" one ProRes accelerator. That explains why Apple says M1 Max is twice as fast for encoding as M1 Pro.

I wonder how much chip real estate this took up.

Interestingly, A15 in the iPhone 13 Pro & Pro Max now support ProRes video recording too. Not sure how much that helps iPhone users though, since it takes up 6 GB per minute. However, I guess that means A15/A16 based M2 will also support ProRes acceleration, something M1 doesn't have.

 

[Semel]

Yes, Nvidia historically does very well with this test, but I should point out two caveats:

1. The M1 Max you quoted is a cut down version, missing 25% of its compute units.

2. There are many variations of the 3080, even on mobile. What one are you specifically talking about?

0) Yes, and the same applies to the recent Adobe benchmark where M1 wins.

That's why I'm waiting for some games benchmarks..It will make everything clear.

1) You are right..

2) It's the Lenovo Legion 7 model.

 

[insertcarehere]

Geekbench OpenCL is likely not a good proxy for comparing performance between different architectures. A 6700XT scores the same as a 2070 super, and From here the M1 scores ~18k while the 1650 scores 36-39k, but the 1650 doesn't hold anywhere near that advantage in actual games.

And that's with the 1650 having ~twice the available bandwidth of the M1 SoC, which isn't going to be the case with the M1 Pro/Max vs their competitors.

 

[igor_kavinski]

It is impossible to put into words the misery of waiting for AT's M1 Max review.

 

[bigggggggg]

It seems like the M1 Pro CPU crown as the king of mobile CPUs lasted less than a week.
If i'm not wrong, anyway, these CPU will come out in Q1 2022.
However, the extraordinary efficiency of the M1 Pro would make me prefer it above anything in the market right now.

The only thing is that, despite being impressive chips, looks like apple didn't try too hard for the cpu subsystem with the M1 Pro. Giving the M1 Pro the A15 cores and maybe two more extra-cores, it would have assured the best absolute performance until the M2/M2 Pro came out.

Anyway, still an impressive chip, with an incredbile GPU subsystem (never thought the GPU would be on the 3080M level of performance)

 

[Eug]

It seems like the M1 Pro CPU crown as the king of mobile CPUs lasted less than a week.
If i'm not wrong, anyway, these CPU will come out in Q1 2022.

Core i9-12900HK? What is the TDP? And what should the real TDP be?

These are the heatsinks in my two Mac Pros for my 80 W and 150 W E5345 and X5365 Xeons…

 

[bigggggggg]

Core i9-12900HK? What is the TDP? And what should the real TDP be?

These are the heatsinks in my 2 Mac Pros for my 80 to 150 W E5345 and X5365 Xeons.

View attachment 51844

Probably the system under full load will consume some 60 watt ca.
Anyway i think it is better to lose only on the efficiency side than badly losing on the efficiency side and the performance side. Something I'm curious is how alder lake will perform in a real working day, and see if low-power cores will be beneficial as we have seen for the M1 machines.

And I'm not saying i would buy a 12900HK instead of an M1 Pro, obviously.

 

[Eug]

Probably the system under full load will consume some 60 watt ca.
Anyway i think it is better to lose only on the efficiency side than badly losing on the efficiency side and the performance side. Something I'm curious is how alder lake will perform in a real working day, and see if low-power cores will be beneficial as we have seen for the M1 machines.

And I'm not saying i would buy a 12900HK instead of an M1 Pro, obviously.

I just read online it’s supposed to be 65 Watts. However these days those numbers seem meaningless to me. It seems 65 Watt TDP of today is a totally different animal vs a 65 Watt TDP some years ago.

 

[bigggggggg]

I just read online it’s supposed to be 65 Watts. However these days those numbers seem meaningless to me. It seems 65 Watt TDP of today is a totally different animal vs a 65 Watt TDP some years ago.

AT gives 65 watt for the past 11900-H under full load. I think it will be in that spot (you have 2 low-efficiency core less but 8 more high-efficiency cores).

In any case it will not be in any sense as efficient as the M1 Pro is.

 

[nxre]

I think Apple is fine with not having the absolute performance crown. Macs never sold based on that, and Apple is investing a lot of die space on accelerators that make the Mac the best machine, desktop or mobile, for certain workloads, instead of a one-size-fits all approach.
Still, the results are extremely competitive for a year-old architecture with no clock increases versus Intel biggest architectural change in a decade. I was actually expecting an even bigger gap, but I guess these are engineering samples that will still get tweaked and ADL will score higher when it hits consumers.

 

[bigggggggg]

Still, the results are extremely competitive for a year-old architecture with no clock increases versus Intel biggest architectural change in a decade. I was actually expecting an even bigger gap, but I guess these are engineering samples that will still get tweaked and ADL will score higher when it hits consumers.

Of course.
I hope M2 chips will bring the speedup we haven't seen with the M1 Pro.
The GPU however is very impressive, if only the software i use could exploit it, i would have bought a MBP like 8 seconds after the keynote

 

[darkswordsman17]

I genuinely do not know why you people keep trying to compare. Apple users are Apple users for reasons other than benchmark performance, and there's almost never a real situation where it even makes sense to compare Apple to PC/Android 1:1 in a benchmark. Apple has had pretty good relative market share throughout, mostly regardless of where performance really is (although, having strong hardware helps the software side as well, and if Apple hadn't been making these moves its possible they might have relinquished quite a bit of market share as Android appears to now have gotten to a point where they're good enough that its not a major limiting factor; like 5 years ago, every other Android release promised to fix all the responsiveness and sluggishness).

For the people that have been wondering why Apple is putting such a beefy GPU in, when they don't care much about games (and they're using so much other dedicated silicon for AI and video/image processing), its likely for VR/AR, where they'll be leveraging the GPU for high resolution and low latency (high framerate/refresh rate) user interface stuff (not so much for game rendering). If their rumored 8K headset is legit, then they'll be wanting a GPU that can do 8K res GUI layovers of probably dual high framerate 4K camera sensors video feed (for high quality AR so that they can pipe the real world in with good responsiveness), hence the strong dedicated video processing. Since its just a GUI, they'll be prioritizing just pixel handling and bandwidth over graphics rendering capabilities (and they can probably get better lighting data from real world image processing than futzing with ray-tracing right now).

 

[Pix12]

AT gives 65 watt for the past 11900-H under full load. I think it will be in that spot (you have 2 low-efficiency core less but 8 more high-efficiency cores).

In any case it will not be in any sense as efficient as the M1 Pro is.

Some laptops with 11900H definitely draw more power tho, running in 107 W Short Burst, 90 W sustained. ~100W in bursts is especially common. If I should guess, given GB5's short bursts, I would say it was running at around 100W.

 

[Roland00Address]

View attachment 51822

Good Graph but I just want to point out that in some ways the artificial and technical lines are blurring.

Where does an iPhone start and an iPad end when they use the same SoC and the vast majority of the code base?

Likewise the same with iPad and Mac Laptop, and then Mac Desktop and Workstation. Apple is now targeting* 5w / 15 / 50 / 70 / 125 / 200** / 300** watt TDPs. And if Apple is going to “keep” something like the 2019 MacPro we will soon see up to a 1000 watt systems being powered by a 1400w power supply.

*not literally these **specific** numbers.
**speculation they may go as high as this for there are some Xeons that Apple use that are in this range.

 

[Eug]

Good Graph but I just want to point out that in some ways the artificial and technical lines are blurring.

Where does an iPhone start and an iPad end when they use the same SoC and the vast majority of the code base?

Likewise the same with iPad and Mac Laptop, and then Mac Desktop and Workstation. Apple is now targeting* 5w / 15 / 50 / 70 / 125 / 200** / 300** watt TDPs. And if Apple is going to “keep” something like the 2019 MacPro we will soon see up to a 1000 watt systems being powered by a 1400w power supply.

*not literally these **specific** numbers.
**speculation they may go as high as this for there are some Xeons that Apple use that are in this range.

Thinking in old terms, not in Apple's new paradigm, I had postulated in this thread that Apple might create a laptop SKU just using something like a rebadged A14 for entry level fanless MacBook Airs, just like they would for their non-Pro iPads. Why? Cuz it's more than fast enough.

In this house, we have in active use:

2014 A8X iPad Air 2
2019 A10 iPad*
2017 A10X iPad Proº
2008 Core 2 Duo 2.0 GHz MacBook
2009 Core 2 Duo 2.26 GHz MacBook Pro
2017 Core m3-7Y32 MacBook*
2007 2 x Xeon X5365 Mac Pro*
2010 Core i7-870 iMacº
2017 Core i5-7600 iMacº

* - All I consider decently fast for regular mainstream use even in 2020.
º - All I considered fast enough for a new entry level fanless Mac in 2020.

None of these machines are faster than A15. However, Apple chose to push forward and release M1 as the chip for their entry level fanless Mac, and it blows all of these machines out of the water performance-wise.

 

[DrMrLordX]

I genuinely do not know why you people keep trying to compare.

Mostly it's due to constant speculation that someday Apple may attempt to break out of their walled garden and snap up extra market share, invade other markets, etc. Apple's CPU design team has been on the radar as top talent for awhile. Whether or not you really believe that they can and should perform such feats is another matter. If you really want to know why? That is why.

 

[JasonLD]

I actually wish AMD and Nvidia would start down clocking large die's again for top end laptop GPU's. The trend they both fell into using smaller die's and clocking them high has resulted in pretty piss poor perf/watt. AMD really are in the best position to have done that, as NAVI 21 memory bus is only 256Bit which is practical for a laptop. Instead they've clocked NAvi 22 to the moon for the 6800M, which is very disapointing as it would have been interesting to see a very low power high perf option (at a high price) for premium thin gaming laptops.

As much as I would love to see Nvidia and AMD doing that, I think Apple is the only company that can pull that off since they are not selling the chips to OEMs but the complete systems for themselves. It just wouldn't be financially viable for Nvidia and AMD.

 

[LightningZ71]

It's also a cost and supply issue. They can get a lot more GPU chips out of a die if they go smaller and faster instead of larger but slower. There are already supply chain issues,they need to produce as much as they can with what they have.

 

[bigggggggg]

Some laptops with 11900H definitely draw more power tho, running in 107 W Short Burst, 90 W sustained. ~100W in bursts is especially common. If I should guess, given GB5's short bursts, I would say it was running at around 100W.

We need AT analysis and SPEC benchmark to confirm the performance.

Don't know why nobody (or almost) but AT makes a serious analysis of performance with proper benchmarks.