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

M5 Family discussion here:

Page 431 - Discussion - Apple Silicon SoC thread

Page 431 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[Covfefe]

I’m pretty sure Notebookcheck, and ComputerBase run their power draw test via a meter. If I’m wrong someone correct me.

I think ComputerBase uses Power Metrics. Their chart says CPU Package Power in their M4 Pro Mac Mini review. It could vary from review to review though.



Looking at Notebookcheck's testing methodology, they measure at the outlet for their power consumption charts. However, I noticed that the power consumption in the CPU performance section of their Macbook Pro M4 review is lower so I think they're using Power Metrics for that section. "In the multi-core tests, the new M4 in the MBP 14 consumed a maximum of 24 watts" compared to 37.8 watts average and 43.7 watts max for Cinebench R23 in the power consumption charts.

 

[name99]

Anyone ever seen any figures or even hints what Apple TV's installed base is? I love mine but I've never once thought "man I would like to game on this" because it isn't designed for that and would need some sort of controller since the remote isn't exactly useful beyond a narrow category of games.

If the market isn't very large, and it would require the purchase of a 3rd party gaming controller which pushes you into an even smaller niche, I can see why developers ignore it. If Apple wanted gaming to be a thing they'd promote it for that, incentivize developers, offer a controller for it etc. It would still be a drop in ocean compared to the revenue they make from mobile gaming - and would suffer in comparisons with real game consoles.

Apple Game Controllers & TV Accessories

Discover the world of Apple TV remotes and controllers. Compare models, like game controllers, shop online and find more TV accessories.

www.apple.com

 

[poke01]

Looks accurate to me. Another issue is a lot of reviewers don’t measure power under a worst case scenario. Instead they will do things like simulate web browsing, watch a video, etc. These are areas where the M-series excels. However, when the chip is fully loaded, efficiency tends to fall.

This is because Apple laptops do not change to a different power profile automatically when unplugged, leading to worse battery life when fully loaded. If you want even better battery life, then you can change the macbook to low power mode.

We see this in countless reviews.

 

[johnsonwax]

Anyone ever seen any figures or even hints what Apple TV's installed base is? I love mine but I've never once thought "man I would like to game on this" because it isn't designed for that and would need some sort of controller since the remote isn't exactly useful beyond a narrow category of games.

If the market isn't very large, and it would require the purchase of a 3rd party gaming controller which pushes you into an even smaller niche, I can see why developers ignore it. If Apple wanted gaming to be a thing they'd promote it for that, incentivize developers, offer a controller for it etc. It would still be a drop in ocean compared to the revenue they make from mobile gaming - and would suffer in comparisons with real game consoles.

Games aren't relevant to the market. But Apple's AI plans are - if they want AI powered Siri on AppleTV, they're going to need the RAM/NPU of their more current SOCs.

 

[poke01]

some info on SRAM in A19 Pro

 

[mvprod123]

https://twitter.com/x/status/1999537832273096824

 

[poke01]

"Apple Mac True Power Consumption: power meter vs powermetrics"

The youtuber says the "telemetric" data that most people trust to show Apple Silicon power consumption, can be understating the actual power consumption significantly. Or at least in his results...

Powermetrics command never showed total system power consumption.

It’s shows CPU/GPU/NPU package power, that’s it. Anandtech mentioned this in 2021.

 

[Jan Olšan]

Powermetrics command never showed total system power consumption.

It’s shows CPU/GPU/NPU package power, that’s it. Anandtech mentioned this in 2021.

That's obvious and not what the video is about, the measurements if legit show a huge discrepance that can hardly be explained by powermetrics being accurate and rest of the system taking the excess power.

 

[Nothingness]

That's obvious and not what the video is about, the measurements if legit show a huge discrepance that can hardly be explained by powermetrics being accurate and rest of the system taking the excess power.

How do reviewers measure power consumption on x86 machines? I agree that the better way seems to use a wattmeter at the wall. Comparisons should only be made at that level on all systems under review.

 

[Covfefe]

How do reviewers measure power consumption on x86 machines? I agree that the better way seems to use a wattmeter at the wall. Comparisons should only be made at that level on all systems under review.

It varies. Plenty of reviewers use software readings. Anandtech used software readings when they reviewed the M1 (archive link).

 

[Covfefe]

I decided to dig a little deeper into notebookcheck's power consumption numbers. I used the max power draw mentioned in the CPU part of the review and compared it to the max from the power consumption chart. The Macbooks' outlet measurements are 70-82% higher than their software readings. On the x86 side, the outlet measurements are only 32-40% higher.

This is all very messy. The power consumption wasn't measured at the same time, and the benchmarks used might vary from review to review. Nevertheless there's a clear pattern of the Apple's software readings understating the actual power draw.

​	software reading​	outlet reading​	discrepancy​
Macbook Pro M4​	24 watts​	43.7 watts​	82.1%​
Macbook Pro M4 Pro​	46 watts​	84.4 watts​	83.5%​
Macbook Pro M5​	30.5 watts​	52 watts​	70.5%​
Lenovo Yoga 7 14 258V​	37 watts​	51.8 watts​	40.0%​
Framework hx 370​	46 watts​	61 watts​	32.6%​

 

[poke01]

I decided to dig a little deeper into notebookcheck's power consumption numbers. I used the max power draw mentioned in the CPU part of the review and compared it to the max from the power consumption chart. The Macbooks' outlet measurements are 70-82% higher than their software readings. On the x86 side, the outlet measurements are only 32-40% higher.

This is all very messy. The power consumption wasn't measured at the same time, and the benchmarks used might vary from review to review. Nevertheless there's a clear pattern of the Apple's software readings understating the actual power draw.

​	software reading​	outlet reading​	discrepancy​
Macbook Pro M4​	24 watts​	43.7 watts​	82.1%​
Macbook Pro M4 Pro​	46 watts​	84.4 watts​	83.5%​
Macbook Pro M5​	30.5 watts​	52 watts​	70.5%​
Lenovo Yoga 7 14 258V​	37 watts​	51.8 watts​	40.0%​
Framework hx 370​	46 watts​	61 watts​	32.6%​

That software reading is only for the total CPU core package (it doesn't include the cache, uncore etc) for the Macbooks.

There are Mac sensor apps like Mactop and Stats that show near total system power consumption unlike powermetrics.

GitHub - context-labs/mactop: mactop - Apple Silicon Monitor Top

mactop - Apple Silicon Monitor Top. Contribute to context-labs/mactop development by creating an account on GitHub.

github.com

GitHub - exelban/stats: macOS system monitor in your menu bar

macOS system monitor in your menu bar. Contribute to exelban/stats development by creating an account on GitHub.

github.com

 

[poke01]

I did some testing. Watch it till end. Pretty much confirms that the standard powermetrics command is only reporting total CPU cores power consumption.

I used a Tapo smart plug and Stats for software sensor software. Its pretty accurate.

https://imgur.com/a/52513eC

 

[jdubs03]

Seems to line up with what notebookcheck was outputting for multicore power draw at the outlet with the M4.

@poke01 did you try a single-core run?

 

[Nothingness]

I decided to dig a little deeper into notebookcheck's power consumption numbers. I used the max power draw mentioned in the CPU part of the review and compared it to the max from the power consumption chart. The Macbooks' outlet measurements are 70-82% higher than their software readings. On the x86 side, the outlet measurements are only 32-40% higher.

This is all very messy. The power consumption wasn't measured at the same time, and the benchmarks used might vary from review to review. Nevertheless there's a clear pattern of the Apple's software readings understating the actual power draw.

​	software reading​	outlet reading​	discrepancy​
Macbook Pro M4​	24 watts​	43.7 watts​	82.1%​
Macbook Pro M4 Pro​	46 watts​	84.4 watts​	83.5%​
Macbook Pro M5​	30.5 watts​	52 watts​	70.5%​
Lenovo Yoga 7 14 258V​	37 watts​	51.8 watts​	40.0%​
Framework hx 370​	46 watts​	61 watts​	32.6%​

IMHO comparing laptops leads to more differences due to other components (screen and its settings in particular). It'd be more interesting to measure this on desktops. That's why I found the measurements on the Apple Studio in the video relevant.

 

[MS_AT]

It'd be more interesting to measure this on desktops.

Well, here too it could be a problem I frequently see reviewers show 10+ W higher idle power draw measured at the wall for a hw similiar to mine. The biggest difference they are using oversized power supplies, and usually the efficiency <5% is poor.

 

[poke01]

Seems to line up with what notebookcheck was outputting for multicore power draw at the outlet with the M4.

@poke01 did you try a single-core run?

Will try later this week

 

[name99]

I decided to dig a little deeper into notebookcheck's power consumption numbers. I used the max power draw mentioned in the CPU part of the review and compared it to the max from the power consumption chart. The Macbooks' outlet measurements are 70-82% higher than their software readings. On the x86 side, the outlet measurements are only 32-40% higher.

This is all very messy. The power consumption wasn't measured at the same time, and the benchmarks used might vary from review to review. Nevertheless there's a clear pattern of the Apple's software readings understating the actual power draw.

​	software reading​	outlet reading​	discrepancy​
Macbook Pro M4​	24 watts​	43.7 watts​	82.1%​
Macbook Pro M4 Pro​	46 watts​	84.4 watts​	83.5%​
Macbook Pro M5​	30.5 watts​	52 watts​	70.5%​
Lenovo Yoga 7 14 258V​	37 watts​	51.8 watts​	40.0%​
Framework hx 370​	46 watts​	61 watts​	32.6%​

Do the people creating these sorts of charts have a clue what they are ACTUALLY measuring?
For example let's suppose (I have no idea if this is true) that
- Apple's wall power draw varies extremely rapidly (goes up and down multiple times per second) because, eg, it operates on a principle of filling up some capacitors as rapidly as possible, then draining them while throttling the AC draw
- the power meter is designed to capture the *maximum* draw over some longish period (say a second or two)

Then an Apple internal value reporting essentially a "genuine" average will look very different from a HW meter reporting essentially a "maximum over one second" value.

This could be resolved very easily by using a scheme that measures ENERGY not power. Either a meter that's designed (and designed WELL...) to measure energy, or by driving the mac off a battery with a trustworthy "capacity" measurement.

But I've concluded that the amateur internet world is simply uninterested in actually thinking about this issue [energy vs power] like an adult; they've being making idiotic statements on the issue since the M2, so I'll bow out now and leave the rest of you to fight about this in blissful ignorance.

 

[poke01]

AMD used to also report ONLY the GPU cores power consumption pre-RDNA3.
Meaning the TBP was off by 100watts in some GPUs like RX 6800 XT

This isn’t unique to Apple silicon. Any software measurement is inaccurate, period.

 

[adroc_thurston]

AMD used to also report ONLY the GPU cores power consumption pre-RDNA3

pre-idk but definitely not RDNA3.

Meaning the TBP was off by 100watts in some GPUs like RX 6800 XT

nope. 300W is 300W.

 

[poke01]

nope. 300W is 300W

It doesn’t include memory or VRM on cards prior rdna3. The software doesn’t read the full TBP.

 

[poke01]

Total Board Power (TBP) vs. real power consumption and the inadequacy of software tools vs. real measurements | Page 5 | igor´sLAB

Since many users and especially YouTubers as well as twitchers still take the values of tools like the MSI Afterburner or AMD’s Radeon software for the power consumption of their graphics card at face…

www.igorslab.de

Software is inaccurate.

 

[adroc_thurston]

It doesn’t include memory or VRM on cards prior rdna3. The software doesn’t read the full TBP.

yeah it does.

Total Board Power (TBP) vs. real power consumption and the inadequacy of software tools vs. real measurements | Page 5 | igor´sLAB

Since many users and especially YouTubers as well as twitchers still take the values of tools like the MSI Afterburner or AMD’s Radeon software for the power consumption of their graphics card at face…

www.igorslab.de

Software is inaccurate.

that's a factory OC mememodel in furmark crasher bios mode.

 

[poke01]

that's a factory OC mememodel in furmark crasher bios mode

so he wrote an article based on that?
Looks like he just wants to show AMD in a bad light then.

yeah it does

Yeah looks like you are right looking at some HWInfo pics for the RX 6800XT

 

[poke01]

Yeah looks like you are right looking at some HWInfo pics for the RX 6800XT

hmm looking at my RX9070XT, it does offer more measurements


yeah RDNA2 is limited. No TBP, no GPU power maximum ie spikes