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

 

[Nothingness]

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 ?

Without knowing the methodology used to measure power, I'm not sure what to do with their figures. Is it documented anywhere?

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 ?

People have been putting phone in freezers to run benchmarks for years, nothing new here

 

[Hitman928]

Without knowing the methodology used to measure power, I'm not sure what to do with their figures. Is it documented anywhere?


People have been putting phone in freezers to run benchmarks for years, nothing new here

It seems to be loaded - idle power measured at the socket, but not completely sure.

 

[Nothingness]

It seems to be loaded - idle power measured at the socket, but not completely sure.

Yes, that's how much we know. And I'm afraid that's not enough.

As a reminder, here is what Andrei documented

Last year when we reviewed the M1 inside the Mac mini, we did some rough power measurements based on the wall-power of the machine. Since then, we learned how to read out Apple’s individual CPU, GPU, NPU and memory controller power figures, as well as total advertised package power.

 

[Gideon]

The most interesting part for me is the 4.4Ghz boost in fridge /under liquid nitrogen (just check the .gb6 info)

If that's accurate and achieved just by lowering temps, macbook pros should be capable of reaching close to that on active cooling alone.

This should make M4pro very competitive with Zen5/Arrow lake or beat them outright in ST (at least in laptop form)

 

[Henry swagger]

The most interesting part for me is the 4.4Ghz boost in fridge /under liquid nitrogen (just check the .gb6 info)

If that's accurate and achieved just by lowering temps, macbook pros should be capable of reaching close to that on active cooling alone.

This should make M4pro very competitive with Zen5/Arrow lake or beat them outright in ST (at least in laptop form)

Lets wait for benchmarks from zen5 arrow lake first.. but apple do have the best efficiency

 

[Hitman928]

Yes, that's how much we know. And I'm afraid that's not enough.

As a reminder, here is what Andrei documented

Agreed, but I don't think the app they used for monitoring SoC power is available on iOS (someone let me know if this isn't correct), so this may be the best we get for now.

 

[Mopetar]

I'm waiting for the 13" model that I ordered to ship, but after reading some of the reviews I'm definitely excited for it to arrive.

I'm somewhat curious to see if improvements in the media engines means that it could actually encode any of my Final Cut projects faster than my M1 MBP. I remember when I got an older iPad where Apple had put an encoder on the SoC and it was rendering videos faster than an Intel-based Mac Pro that cost almost 10x as much as the iPad.

 

[Ajay]

I'm waiting to see info on the M4 Pro and Max. Curious if the Pro will be a bigger step up than with the M3 vs M3 Pro.

 

[Doug S]

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.

I'll make you a bet now that the M4 in the Macbooks will be the exact same one, with the same clock rate, just like they have done with previous Apple Silicon iPad Pros & Macs. They aren't going to do any "fancy cooling" because Apple does not want loud fans in Macs, especially not in laptops. It is hilarious you think Apple will be "desperate" to beat x86. They don't care that much, because they don't compete directly with them. Windows users aren't switching to a Mac if the M4 MBP beats x86, and Mac users aren't switching to a PC if x86 beats M4. That's not how those markets work, it would take an extreme performance differential (at least 50%) before enough people to matter considered such a thing.

If people want to know what performance they'll be getting out of a Macbook with an M4 they already know based on the iPad Pro results. We'll have to see about M4 Pro/Max, they will have a little more cache and I wouldn't be surprised if they did LPDDR5-8533 instead of the -7500 the iPad Pro has. I imagine they'll have the same clock rate, too, unless another few months of tweaking allows them to squeeze out an extra 100 MHz or so.

 

[Doug S]

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.

Supercooling it doesn't make it operate at a higher frequency, it just operates at the standard 4.4 GHz for longer before throttling.

 

[Eug]

I'll make you a bet now that the M4 in the Macbooks will be the exact same one, with the same clock rate, just like they have done with previous Apple Silicon iPad Pros & Macs.

---

If people want to know what performance they'll be getting out of a Macbook with an M4 they already know based on the iPad Pro results. We'll have to see about M4 Pro/Max, they will have a little more cache and I wouldn't be surprised if they did LPDDR5-8533 instead of the -7500 the iPad Pro has. I imagine they'll have the same clock rate, too, unless another few months of tweaking allows them to squeeze out an extra 100 MHz or so.

Yeah, and IIRC, for M1, M2, and M3, the Pro/Max/Ultra have also had the same clock speed for each series: 3.2 GHz, 3.5 GHz, 4.05 GHz respectively (and now 4.4 GHz). I remember for some Geekbench tests the recorded clock speed has been higher with the Max/Ultra, but it was only for some individual results, so it may just be an anomaly with the data captured.

I might expect though that the M4 non-Pro chips between the 13" MacBook Air and the 14" MacBook Pro could diverge, with the Air getting the 9-core and the Pro getting 10-core.

 

[poke01]

I'll make you a bet now that the M4 in the Macbooks will be the exact same one, with the same clock rate, just like they have done with previous Apple Silicon iPad Pros & Macs. They aren't going to do any "fancy cooling" because Apple does not want loud fans in Macs, especially not in laptops. It is hilarious you think Apple will be "desperate" to beat x86. They don't care that much, because they don't compete directly with them. Windows users aren't switching to a Mac if the M4 MBP beats x86, and Mac users aren't switching to a PC if x86 beats M4. That's not how those markets work, it would take an extreme performance differential (at least 50%) before enough people to matter considered such a thing.

If people want to know what performance they'll be getting out of a Macbook with an M4 they already know based on the iPad Pro results. We'll have to see about M4 Pro/Max, they will have a little more cache and I wouldn't be surprised if they did LPDDR5-8533 instead of the -7500 the iPad Pro has. I imagine they'll have the same clock rate, too, unless another few months of tweaking allows them to squeeze out an extra 100 MHz or so.

They have increased clock speed compared to base M chip for the M2 family.

M2 Max had a higher clock rate than M2. Since N3E is a much better node we might see increased clocks for M4 Max.

 

[Eug]

They have increased clock speed compared to base M chip for the M2 family.

M2 Max had a higher clock rate than M2. Since N3E is a much better node we might see increased clocks for M4 Max.

Are we sure about that? That was what I was wondering about earlier.

M2 is 3.5 GHz. In Geekbench, the majority of M2 Max and Ultra are 3.5 GHz as well, but I see some at 3.7 GHz, so I wondered if as a data collection issue. Or perhaps it is real.

EDIT:

Hmm... The 3.7 GHz Max & Ultra do tend to score higher than the 3.5 GHz Max & Ultra, for single-core.

 

[poke01]

M2 is 3.5 GHz. In Geekbench, the majority of M2 Max and Ultra are 3.5 GHz as well, but I see some at 3.7 GHz, so I wondered if as a data collection issue. Or perhaps it is real.

its real. My guess is N5P allowed for higher clocks.

 

[amosliu137]

A user named qzyqzy achieved a score of 4009 on the M4. He might have used a secret method to ensure every device received an exceptionally high score. For reference, the A17 Pro scored 3044, and the M3 Max scored 3267.
https://browser.geekbench.com/user/qzyqzy

 

[Eug]

A user named qzyqzy achieved a score of 4009 on the M4. He might have used a secret method to ensure every device received an exceptionally high score. For reference, the A17 Pro scored 3044, and the M3 Max scored 3267.
https://browser.geekbench.com/user/qzyqzy

I want to see 15000 multi-core now. Just because.

 

[amosliu137]

One more thing the real frequency of m4 single boost is 4515mhz.

 

[mr_roboto]

They have increased clock speed compared to base M chip for the M2 family.

M2 Max had a higher clock rate than M2. Since N3E is a much better node we might see increased clocks for M4 Max.

Even M1 Pro/Max/Ultra had higher clocks than plain M1 - it just wasn't enough of a bump that most people noticed. 3204 MHz vs 3228 MHz. Over time they've been increasing the gap between base and pro/max M series chips.

 

[Eug]

TweakTown reports on Geekerwan's M4 adventures:

Apple's new iPad Pro wtih M4 chip breaks Geekbench 6 world record, cooled with liquid nitrogen

Apple's new M4-powered iPad Pro destroys Geekbench 6 world record, pushing past 4000 points thanks to reviewer Geekerwan and LN2 (liquid nitrogen) cooling.

www.tweaktown.com

However, for some reason they've described it as liquid nitrogen overclocking of M4 from 4.40 to 4.41 GHz. Wut?

 

[Doug S]

I think the closest to 15000 so far is 3977 / 14955.

Just think of all the wasted CPU cycles running GB 6.3 over and over again Apple could have saved if they had just been able to clock M4 50 MHz higher so it would have more easily passed the 4K/15K hurdles

 

[roger_k]

Supercooling it doesn't make it operate at a higher frequency, it just operates at the standard 4.4 GHz for longer before throttling.

What I am trying to say is that under normal circumstances it might not actually operate at 4.4 Ghz. It could be 4.3 or 4.34 Ghz most of the time. Apple likes to push the frequency curve in a fairly aggressive manner, it is very likely that 4.4 Ghz is where you start hitting the extreme diminishing returns.

 

[roger_k]

Regarding the discussion about IPC improvements... I am becoming increasingly convinced that most of the discourse is useless because of the deeply flawed methodology. In different Geekerwan videos the results reported for A17 Pro have a relative error of almost 5%. Combine this with the non-precise frequency estimation and you end up with huge relative error for the IPC estimates.

We need more clear methodology, results from multiple devices (to circumvent device bias), and most importantly, we need to start looking at the variance instead of point estimates. I tried to do this earlier for some of the GB6 data and I hope I could illustrate how much more useful this approach it.

Bottomline: the data is crap, methodology is crap, the relative error is crap, meaning that the results are crap.

 

[roger_k]

One more thing the real frequency of m4 single boost is 4515mhz.

Based on a single entry that could be a fluke? Come on. These APIs are not that precise.

 

[igor_kavinski]

Just think of all the wasted CPU cycles running GB 6.3 over and over again Apple could have saved if they had just been able to clock M4 50 MHz higher so it would have more easily passed the 4K/15K hurdles

When searching for GB results, I kept seeing wayyyy too many Intel 4th gen and 5th gen macbook results. I don't know why these people are so crazy about GB, especially on older hardware. Seems more like some shop(s) checking their shipment of old laptops to see if they will post a GB score within an expected range or something. Maybe Apple has some internal test in MacOS that can tell these people that the CPU is working as expected?

 

[Eug]

When searching for GB results, I kept seeing wayyyy too many Intel 4th gen and 5th gen macbook results. I don't know why these people are so crazy about GB, especially on older hardware. Seems more like some shop(s) checking their shipment of old laptops to see if they will post a GB score within an expected range or something. Maybe Apple has some internal test in MacOS that can tell these people that the CPU is working as expected?

People do it because it’s free, it’s available as a downable binary, a test result database is available and freely searchable, there are a bazillion results to compare against, and it is dead simple to run.

If those things were all true with spec, people would run that too, but they aren’t so they don’t. People like to complain about Geekbench, but there are pretty good reasons why it is so popular. Hell, from what I’ve seen of CPU reviews, even big popular review sites only occasionally run SPEC.

As for old machines, they run GB on them because they can. Like why not, esp. as a comparison when they get new hardware? It’s free and easy and takes just a couple of minutes.