Request: can some with M1 Max run this Powermetrics code to test idle + load power draw?

[helix21]

Background here is The Verge reporting that 16" M1 Max had only 10 hours of battery life (16 hours on the M1 Pro). This has lead to a fair few people wanting to cancel their M1 Max orders and go for the Pro instead. I certainly want do.

There's an excellent reddit thread which goes into detail on power draw and CPU/GPU state with a terminal command that you can copy paste to look at the power draw your machine is having.

Can someone with a M1 Max 14 inch please paste the following command in terminal and look at the live power draw on idle and when doing something intensive? Full disclosure: I am not a developer and I don't 100% know what it does, but have just run it on my M1 Pro and didn't do anything bad but give me live readout of power usage.

sudo powermetrics -i 1000 -a --samplers cpu_power | grep -e "Package Power" -e "CPU Power" -e "GPU Power"

I borrowed a friend's M1 Pro (10 core CPU, 16 core GPU, 16GB RAM) for this test.

Idle load (not doing anything): ~600mW total; CPU 240mW power ; GPU 10 mW.
Running an intense app (proprietary app): ~10,000 mW package power ; 5000mW CPU power; 400mW GPU power.
Safari watching YouTube HD in theatre mode: ~1400 mW package power ; 400mW CPU power; 10mW GPU power.

A big concern with M1 Max is that either the additional GPU cores or the extra silicon with the memory controllers/architecture cause a higher power drain. I think most are worried of a higher idle draw, as it would reduce battery life doing simple things (email/word processing).

The optimal finding here would be if the M1 Max also has very low power draw so that idle battery runtime would be preserved. Then you power to spare/use if needed and it doesn't "cost" you runtime when you don't need it.

 

[bill-p]

Anandtech already ran that command and did their own investigation:

Apple's M1 Pro, M1 Max SoCs Investigated: New Performance and Efficiency Heights

www.anandtech.com

They already stated their suspicions as well:

In multi-threaded scenarios, the package and wall power vary from 34-43W on package, and wall active power from 40 to 62W. 503.bwaves stands out as having a larger difference between wall power and reported package power – although Apple’s powermetrics showcases a “DRAM” power figure, I think this is just the memory controllers, and that the actual DRAM is not accounted for in the package power figure – the extra wattage that we’re measuring here, because it’s a massive DRAM workload, would be the memory of the M1 Max package.

The command won't give us much useful information since power draw still differs depending on the type of load. And the prime suspect right now are the memory modules. Everything else seems like it's already accounted for.

And it makes sense. You can't get 400GB/s memory bandwidth without paying some penalty in power consumption over the M1 Pro and M1.

 

[helix21]

Thank you. So you think that the overall package power draw does not include the memory controller power draw?

It's so frustrating that this is so little information on this. I am really tempted to get rid of the M1 Max.

 

[Sanpete]

Taking the casually tossed off and poorly supported Verge claim seriously is the root problem. I don't see what makes sense about it. Nothing in the Anandtech link supports it, quite the opposite in regard to idle usage.

 

[adamk77]

Taking the casually tossed off and poorly supported Verge claim seriously is the root problem. I don't see what makes sense about it. Nothing in the Anandtech link supports it, quite the opposite in regard to idle usage.

Yeah. Anandtech link is saying 0.2W idle... doesn't seem right. Or maybe I missed some details in the post? Or my math is off. With a 100 W/Hr battery that comes out to 500Hours or almost 21 days, doesn't it?

 

[bill-p]

Yeah. Anandtech link is saying 0.2W idle... doesn't seem right. Or maybe I missed some details in the post? Or my math is off. With a 100 W/Hr battery that comes out to 500Hours or almost 21 days, doesn't it?

You missed this:

Starting off with device idle, the chip reports a package power of around 200mW when doing nothing but idling on a static screen. This is extremely low compared to competitor designs, and is likely a reason Apple is able achieve such fantastic battery life. The AC wall power under idle was 7.2W, this was on Apple’s included 140W charger, and while the laptop was on minimum display brightness – it’s likely the actual DC battery power under this scenario is much lower, but lacking the ability to measure this, it’s the second-best thing we have. One should probably assume a 90% efficiency figure in the AC-to-DC conversion chain from 230V wall to 28V USB-C MagSafe to whatever the internal PMIC usage voltage of the device is.

With a 100WHr battery, that translates to about 13.8 hours of just sitting there doing nothing. If even one single background task is running, it'll result in much less than that figure. When Anandtech measured the M1 Mac Mini, the same scenario came out to be about 4W, so idle power consumption is twice as bad in M1 Max vs M1 from what I can see. That makes sense because again, you can't turn off RAM. It's most likely the extra memory modules in the M1 Max causing this increase in power consumption.

 

[adamk77]

You missed this:



With a 100WHr battery, that translates to about 13.8 hours of just sitting there doing nothing. If even one single background task is running, it'll result in much less than that figure. When Anandtech measured the M1 Mac Mini, the same scenario came out to be about 4W, so idle power consumption is twice as bad in M1 Max vs M1 from what I can see. That makes sense because again, you can't turn off RAM. It's most likely the extra memory modules in the M1 Max causing this increase in power consumption.

Ah, thanks!

That 13.8 hours make sense. Yes, I agree with you that the memory modules are the culprit. Now, decisions, decisions.

 

[dk831]

14" MBP, M1 Max 32C GPU, 64G RAM

At 99.16% Idle right after a reboot and things have settled:
DRAM Power: 227 mW
CPU Power: 25 mW
GPU Power: 3 mW
Package Power: 470 mW

Watching a YouTube video:
DRAM Power: 746 mW
CPU Power: 66 mW
GPU Power: 35 mW
Package Power: 1492 mW

Anyone with a M1 Pro wanna do the same?

EDIT: These seems to be in the same ball park as the M1 Pro #'s that OP provided.

Some random thoughts:
- What is the difference in consumption for 4 vs 2 DRAM chips. I wonder if on an M1 Max with 32GB vs 64GB there would be much difference since both have 4 DRAM chips.
- Additional cache blocks on M1 Max.
- The GPU seems to power down to almost nothing when not being used.

 

[Sanpete]

With a 100WHr battery, that translates to about 13.8 hours of just sitting there doing nothing. If even one single background task is running, it'll result in much less than that figure. When Anandtech measured the M1 Mac Mini, the same scenario came out to be about 4W, so idle power consumption is twice as bad in M1 Max vs M1 from what I can see. That makes sense because again, you can't turn off RAM. It's most likely the extra memory modules in the M1 Max causing this increase in power consumption.

Idle is measured here with the screen on, so you can't compare to the Mini that way and just assume the difference is RAM. According to Apple the idle Mini draws 6.8W, btw, so something is askew somewhere.

The Tom's Guide review got over 15 hours web browsing on battery with their Max, so clearly something is askew.

MacBook Pro 2021 (16-inch) review

Apple's MacBook Pro 2021 (16-inch) is an across-the-board upgrade that delivers best-in-class performance

www.tomsguide.com

Idle RAM does use a small amount of power, but we really can't tell how much from the information cited here.

Yes, I agree with you that the memory modules are the culprit.

That doesn't follow.

 

[bill-p]

Idle is measured here with the screen on, so you can't compare to the Mini that way and just assume the difference is RAM. According to Apple the idle Mini draws 6.8W, btw, so something is askew somewhere.

The Tom's Guide review got over 15 hours web browsing on battery with their Max, so clearly something is askew.

MacBook Pro 2021 (16-inch) review

Apple's MacBook Pro 2021 (16-inch) is an across-the-board upgrade that delivers best-in-class performance

www.tomsguide.com

Idle RAM does use a small amount of power, but we really can't tell how much from the information cited here.


That doesn't follow.

Anandtech measured actual number, so their number is real.

Tom's Guide does not give as much detail as Anandtech so I'd ignore their number as well.

In fact, I no longer trust Apple's benchmarks because it is very clear they nitpicked only the ones that show the biggest improvements. I did have the 16" with M1 Max in for a week to test, and just exchanged it with the M1 Pro 14" on Friday. I concur with The Verge's numbers as far as battery life is concerned.

In fact, the 14" M1 Pro doesn't seem to get much better battery life compared to the 16" with M1 Max either. Likely because the battery is smaller. But the 14" is much more portable and runs cooler overall, so I can forgive that. Plus I'm not losing out on CPU performance, which is more important for my workflow.

Going back to your "that doesn't follow" remark, the truth is: it does. The M1 Max gets twice the bandwidth of the M1 Pro. That means it needs to have memory modules arranged in 2x wider bus and also these memory modules need to be "active" almost all the time, so I'd expect memory system of M1 Max to consume 2x more power compared to M1 Pro. You really cannot get faster RAM without having to pay any penalty for that. Apple cannot defy physics.

 

[Sanpete]

Anandtech measured actual number, so their number is real.

Tom's Guide does not give as much detail as Anandtech so I'd ignore their number as well.

In fact, I no longer trust Apple's benchmarks because it is very clear they nitpicked only the ones that show the biggest improvements. I did have the 16" with M1 Max in for a week to test, and just exchanged it with the M1 Pro 14" on Friday. I concur with The Verge's numbers as far as battery life is concerned.

In fact, the 14" M1 Pro doesn't seem to get much better battery life compared to the 16" with M1 Max either. Likely because the battery is smaller. But the 14" is much more portable and runs cooler overall, so I can forgive that. Plus I'm not losing out on CPU performance, which is more important for my workflow.

Going back to your "that doesn't follow" remark, the truth is: it does. The M1 Max gets twice the bandwidth of the M1 Pro. That means it needs to have memory modules arranged in 2x wider bus and also these memory modules need to be "active" almost all the time, so I'd expect memory system of M1 Max to consume 2x more power compared to M1 Pro. You really cannot get faster RAM without having to pay any penalty for that. Apple cannot defy physics.

I suspect Apple also measured.

What details of the Tom's Guide test (one of many that don't fit) could possibly affect the implication that your analysis based on Anandtech's numbers can't be correct?

RAM doesn't need to be "active" when not in use, but twice the idle RAM should consume twice the power, no doubt. We just don't know how much that is.

(It's a different topic, but tests under heavy load show the 16" to run significantly cooler and quieter than the 14". It has a larger cooling system, and the fans don't have to turn as fast to move as much air.)

 

[AFK]

by ram needing to be active, do you mean it needs constant power for it maintain data? but wouldn't that be different from xfering the data, which is what i would assume consumes more juice.

also wouldnt 64gb be more power hungry than 32gb? all these reviews seem to be 32gb machines

 

[helix21]

14" MBP, M1 Max 32C GPU, 64G RAM

At 99.16% Idle right after a reboot and things have settled:
DRAM Power: 227 mW
CPU Power: 25 mW
GPU Power: 3 mW
Package Power: 470 mW

Watching a YouTube video:
DRAM Power: 746 mW
CPU Power: 66 mW
GPU Power: 35 mW
Package Power: 1492 mW

Anyone with a M1 Pro wanna do the same?

EDIT: These seems to be in the same ball park as the M1 Pro #'s that OP provided.

Some random thoughts:
- What is the difference in consumption for 4 vs 2 DRAM chips. I wonder if on an M1 Max with 32GB vs 64GB there would be much difference since both have 4 DRAM chips.
- Additional cache blocks on M1 Max.
- The GPU seems to power down to almost nothing when not being used.

Thanks for doing this! Seems quite similar numbers to mine.

What's battery life looking like for you? For me its about comparable to my M1 Air, maybe slightly less, so losing about 10% every hour doing wireless web stuff (email etc).

Could you also please specific if you use Chrome or Safari, and if you have any Intel apps?

 

[dk831]

Thanks for doing this! Seems quite similar numbers to mine.

What's battery life looking like for you? For me its about comparable to my M1 Air, maybe slightly less, so losing about 10% every hour doing wireless web stuff (email etc).

Could you also please specific if you use Chrome or Safari, and if you have any Intel apps?

You're welcome.

I was using Safari for that test.

I've also been experiencing ABOUT 10% loss of battery per hour but I haven't been tracking it much. From a full battery and after doing some light work on the laptop, Activity Monitor reports 10-12 hours Time remaining.
Right now Activity Monitor is showing 75% battery remaining, and 9:12 Time remaining.

I do have some Intel apps in the background still, mainly whatever comes with Adobe CC.

TBH, I think the battery life is good but I'm was trying to make a final decision between M1 Pro and M1 Max before the return window is finished. Just wanted to make sure the difference wasn't huge.

I was initially worried about the GPU potentially not being as efficient. Apple presented a graph showing the M1 Max having a higher lower power consumption range. I'm not worried about that anymore after seeing the GPU only take 3mW when idle. The only other major difference was the DRAM consumption but we seem to be about the same there between the two different models.

 

[helix21]

You're welcome.

I was using Safari for that test.

I've also been experiencing ABOUT 10% loss of battery per hour but I haven't been tracking it much. From a full battery and after doing some light work on the laptop, Activity Monitor reports 10-11 hours Time remaining.

I do have some Intel apps in the background still, mainly whatever comes with Adobe CC.

TBH, I think the battery life is good but I'm just trying to make a final decision between M1 Pro and M1 Max before the return window is finished. Just wanted to make sure the difference wasn't huge.

That’s really good! I think that’s about the same (or bit better than me) on M1 Pro/16GB RAM. I am however limited to Chrome due to work apps. No Intel apps.

So I suspect the difference between M1 Max and Pro might be up to an hour-two for the 14”. I suspect I have higher draw from Chrome.

I find the theory that RAM/memory controllers seems to make the difference quite interesting. You have 64GB, I have 16. I suspect its the number of memory modules that’s probably matters more than the amount of RAM.

Right now I am tempted to keep my M1 Max order.

 

[bill-p]

Right now I am tempted to keep my M1 Max order.

Perhaps you should. So I found out yesterday that switching the default display calibration from the 1600 nits XDR profile to the 500 nits XDR profile makes quite a difference. Battery life has been consistently around 12 - 13 hours for me.



This is on M1 Pro, but I suspect M1 Max may also benefit from it. There may be something off with the way the machine is processing contents on the screen with HDR?

 

[archi penko]

request seconded!

 

[archi penko]

.

 

[bluegt]

On battery at 5 ticks brightness, 14" base model.

Edited results below as I realised TG Pro was causing increased CPU usage.

When Idle my results are:
ANE Power: 0 mW
DRAM Power: 98 mW
CPU Power: 94 mW
GPU Power: 3 mW
Package Power: 272 mW

If I move the mouse cursor in a circle rapidly
ANE Power: 0 mW
DRAM Power: 301 mW
CPU Power: 178 mW
GPU Power: 6 mW
Package Power: 1517 mW

Idle - At 8 ticks:
ANE Power: 0 mW
DRAM Power: 170 mW
CPU Power: 105 mW
GPU Power: 6 mW
Package Power: 309 mW

Circle test - 8 ticks
ANE Power: 0 mW
DRAM Power: 291 mW
CPU Power: 163 mW
GPU Power: 6 mW
Package Power: 1487 mW

idle - maximum brightness:
ANE Power: 0 mW
DRAM Power: 178 mW
CPU Power: 102 mW
GPU Power: 7 mW
Package Power: 339 mW

I can only conclude that Powermetrics does not capture power consumption by the display.

 

[archi penko]

I can only conclude that Powermetrics does not capture power consumption by the display.

Wouldn’t display power consumption be the package power minus all the other values listed?

Reading your results, presumably when you did the “mouse cursor in a circle rapidly” “circle test 8 ticks” and move the mouse rapidly, the ProMotion display went up in Hz.

Which if we think about low power mode that Apple allows, the main thing is it reduces/disables ProMotion & HDR, that seems to make sense.

As to DRAM, it does increase as display activity increases… which is expected, right? But I wonder what that means in relation to Cores then? If I have 32 GB and 32 Cores does that mean…? Less happens?

 

[bluegt]

I tried closing some Safari tabs and Whatsapp to reduce RAM usage, and power consumption dropped further:

ANE Power: 0 mW
DRAM Power: 81 mW
CPU Power: 88 mW
GPU Power: 1 mW
Package Power: 203 mW

 

[bluegt]

Wouldn’t display power consumption be the package power minus all the other values listed?

I thought so, but see my last result, even at max brightness Total package power is 339mW. Less the rest would mean the balance is 52mW.

No way a 14" max brightness screen only uses 52mW of power.

 

[archi penko]

No way a 14" max brightness screen only uses 52mW of power.

ok

anyway, I think it might have something more to do with ProMotion instead, or whatever computer components engage for ProMotion to engage.

 

[archi penko]

Watching a YouTube video:
Anyone with a M1 Pro wanna do the same?

@bluegt

Would it be possible to run your command while watching a YouTube video? So we can compare with what @dk831 got too?

Maybe we can all agree & run the same yet video… like…

 

[bluegt]

@bluegt

Would it be possible to run your command while watching a YouTube video? So we can compare with what @dk831 got too?

Maybe we can all agree & run the same yet video… like…

Watching the YT video you suggested - MacRumors open in another tab
ANE Power: 0 mW
DRAM Power: 712 mW
CPU Power: 4052 mW
GPU Power: 28 mW
Package Power: 7042 mW

Watching the YT video you suggested - MacRumors tab closed
ANE Power: 0 mW
DRAM Power: 217 mW
CPU Power: 147 mW
GPU Power: 3 mW
Package Power: 595 mW