Modern SoCs throttle in a fanless design...

[leman]

The lady doth protest too much, methinks…

A lot of people had success with this mod on the M1 gen machines for almost 2 years now, and I can’t find a soul on the interwebz who has had any fallout from it. All this weeping and wailing and gnashing of teeth over the findings and the thermal mod is literally worse than the maligned YouTubers‘ findings and worse than any perceived damage that might occur.

No matter how you slice it, the chassis isn’t going to get hotter than my old 2018 MacBook Pro which got unbearably hot when running heavy loads. A known thermal problem that Apple shipped for YEARS with no compunction. It was at LEAST 60° C during 20 minute Visual Studio builds. And to answer another poster, the heat is being drawn away from the battery in the front toward the back. Also, internal temps are still being monitored exactly the same as before and the SOC will throttle once it gets too high.

Anyway. Peace! I’m out, these quixotic apology threads are a waste of time.

Are you sure you are replying to the right person? I wasn’t mentioning or discussing any mods, whatever that means.

 

[exoticSpice]

You're misunderstanding the issue. Yes, we know that this is how a modern chip is designed to work. If it gets too hot, it throttles by slowing down the clock speed. We know this. We've known this for many years.

The point is that Apple combined the new M2 chip that can use a LOT more power than the M1 chip with a very thin chassis that DOES NOT have an actual proper heatsink, according to iFixit's teardown.

The previous M1 MacBook Air actually had a decent heatsink, which allowed the temps to stay lower.

This time around, Apple not only got rid of the proper heatsink and replaced it with a thin metal shield, but they also put in a chip that uses more power.

This leads to throttling that happens sooner than it did before.

So yes, the throttling is working as intended, but it could've been avoided if put the M2 chip over to the left side and connected a proper heatsink that fills in that large open area on the left side that isn't filled up with the logic board. (see image)

That's what people aren't happy about. We're not mad at throttling. It works as intended like it always has. We're not happy with Apple's passive cooling design which is vastly worse than the previous M1 MacBook Air's passive cooling design.

View attachment 2032243

I respect what you do Vadim. I have seen your sustained Cinebench test and in that test, the M2 Air was faster even with a weaker cooling design. It's clear to me that in this M2 generation Apple is pushing people who want sustained workloads to the Pro models.

There is a big difference between the 13" Pro and M2 Air in terms of sustained performance over 10 minutes.

How do you think Apple will handle M2 on the iPad Pro? The current M1 iPad Pros have NO heatsinks or any tape.

 

[Wizec]

Are you sure you are replying to the right person? I wasn’t mentioning or discussing any mods, whatever that means.

My reply was to multiple people, but yes primarily you. No you didn’t mention the thermal mod specifically, but it’s part of the general topic. Multiple people have been bashing Max Tech’s suggested thermal pad mod - in this thread.

You did write “Did you do thermal conductivity tests or did you just hold the new heatsink in your hand and decided that it's not adequate?”

It’s not a heat sink; there isn’t a heat sink! There’s just an EM shield. By that very fact alone, he is perfectly within reason to conclude that the new thermal solution is inferior to the previous generation.

 

[darngooddesign]

How do you think Apple will handle M2 on the iPad Pro? The current M1 iPad Pros have NO heatsinks or any tape.

That's probably less of a concern because even Lightroom on the iPad isn't as demanding as LR on the Mac. Can you imagine if apple allowed a MacOS mode on the iPad like some people want; just think about all the "I ran a 30 minute Cinebench benchmark and my iPad overheated" videos. #plannedobsolescence

 

[dgdosen]

On the bright side... If Apple puts an M2 or M3 into a 2lb 12" laptop, the current lack of passive cooling helps us know the thermals will be as good or better

 

[leman]

My reply was to multiple people, but yes primarily you. No you didn’t mention the thermal mod specifically, but it’s part of the general topic. Multiple people have been bashing Max Tech’s suggested thermal pad mod - in this thread. You did write “Did you do thermal conductivity tests or did you just hold the new heatsink in your hand and decided that it's not adequate?” It’s not a heat sink; there isn’t a heat sink! There’s just an EM shield. By that very fact alone, he is perfectly within reason to conclude that the new thermal solution is inferior to the previous generation.

I really don't understand the argument. Previous M1 Air reaches thermal equilibrium at around 10 watts of sustained power consumption. The new M2 Air reaches thermal equilibrium at around 10 watts of sustained power consumption. Experimental evidence thus shows that the thermal dissipation capabilities of both chassis are identical. And yet you keep claiming that the new chassis is worse.

It's like "the old car can do 150kmh, the new car can do 150kmh, but the new car must be slower since I don't like the look of that bodywork"

 

[NT1440]

I really don't understand the argument. Previous M1 Air reaches thermal equilibrium at around 10 watts of sustained power consumption. The new M2 Air reaches thermal equilibrium at around 10 watts of sustained power consumption. Experimental evidence thus shows that the thermal dissipation capabilities of both chassis are identical. And yet you keep claiming that the new chassis is worse.

It's like "the old car can do 150kmh, the new car can do 150kmh, but the new car must be slower since I don't like the look of that bodywork"

But click driven media that only makes money by stoking irrational anger in people said something different!

 

[Xiao_Xi]

Previous M1 Air reaches thermal equilibrium at around 10 watts of sustained power consumption. The new M2 Air reaches thermal equilibrium at around 10 watts of sustained power consumption. Experimental evidence thus shows that the thermal dissipation capabilities of both chassis are identical. And yet you keep claiming that the new chassis is worse.

What p-value did you get in the hypothesis test?

Until there is solid evidence, his claim is as far-fetched as yours.

 

[leman]

What p-value did you get in the hypothesis test?

What is this, the eighties? The academic world is working hard to banish p-values and that's what you are asking for?

Until there is solid evidence, his claim is as far-fetched as yours.

Fair enough. But if you want to be this strict then this entire thread is utterly meaningless. Circumstantial evidence is still evidence. The data that we have might be crap, but there is not a single data point showing that the new M2 chassis is settling at a significantly lower power draw level than the old one. Quite on contrary, reviewers and users who report the observed power appear to agree that the equilibrium is around 10 watts. I don't think this information is useless. Quite on contrary.

 

[Xiao_Xi]

What is this, the eighties? The academic world is working hard to banish p-values and that's what you are asking for?

Out of curiosity, what statistic method would you use to demonstrate that the heat dissipation capability of both chassis is identical?

But if you want to be this strict then this entire thread is utterly meaningless. Circumstantial evidence is still evidence. The data that we have might be crap, but there is not a single data point showing that the new M2 chassis is settling at a significantly lower power draw level than the old one. Quite on contrary, reviewers and users who report the observed power appear to agree that the equilibrium is around 10 watts. I don't think this information is useless. Quite on contrary.

Although these two or three threads on thermal throttling have only weak anecdotal evidence, they are very useful (at least to me) because people can read them and decide what to believe. It's about as useful as a thread on why you should buy product A instead of product B. You read the cons and pros, and decide for yourself.

 

[NT1440]

Out of curiosity, what statistic method would you use to demonstrate that the heat dissipation capability of both chassis is identical?


Although these two or three threads on thermal throttling have only weak anecdotal evidence, they are very useful (at least to me) because people can read them and decide what to believe. It's about as useful as a thread on why you should buy product A instead of product B. You read the cons and pros, and decide for yourself.

It’s fine to accept your own evidence, but answer this please: is the fact that the CPU throttles a bad thing?

 

[Apple_Robert]

The M1 throttles when needed. The M1 has a heat sink. The M2 throttles when needed. The M2 uses a more advanced method to dissipate heat than the M1. I am not seeing the problem with the M2. It seems to me a lot of people on here are still trying to create a problem with the M2 where there isn't one.

 

[bobcomer]

It’s fine to accept your own evidence, but answer this please: is the fact that the CPU throttles a bad thing?

Definitely not in a hardware sense. If it didn't throttle when getting too hot, it would fail, and nobody wants a broken computer.

However, throttling can make a computer not fit the job you want it to do, and that's on you, you didn't buy the right computer for the job.

 

[Xiao_Xi]

is the fact that the CPU throttles a bad thing?

Of course, not. Thermal throttling protects the hardware.

But, we should keep in mind that Apple's obsession with thickness (especially in the Ive era) has brought some problems to their hardware. Some customers want thicker phones for bigger batteries and thicker laptops for less thermal throttling. Indeed, the M1 MBP got praised because it got thicker and better thermal envelope.

None of this would have happened if the 13" MBP had the same design as the 13" MBA. You would customize the laptop by "adding a fan" as you do now by adding more RAM or more SSD space.

 

[Apple_Robert]

And while I am thinking about it, I think Vadim is way out of line in encouraging people to make a modification to their brand new Mac. Doing the single temp test you did after adding pads doesn't really mean too much, especially with long term use. In my opinion, you are misleading people and possibly putting their Macs at risk. I don't know many people who are willing to void the warranty in a brand new Mac like you did but, your video encourages them to make changes that affect the hardware. and voids the warranty which, I don't recall you mentioning in the video. You should take the video down.

 

[thenewperson]

Indeed, the M1 MBP got praised because it got thicker and better thermal envelope.

Did it? Because as far as I could tell people bashed it for not adding much over the M1 Air even with the fan. Same with the M2 MBP.

Also why would they design a fanned and fanless version of the same computer? The compromise of one of those modes would definitely affect the other, so that seems like a silly way to go about it. The people who want M2 at its full capacity should just get an M2 MBP and suck it up re: the design. I too would love a more-modern display on it (I prefer the body to the Air's personally) but oh well.

 

[NT1440]

Of course, not. Thermal throttling protects the hardware.

But, we should keep in mind that Apple's obsession with thickness (especially in the Ive era) has brought some problems to their hardware. Some customers want thicker phones for bigger batteries and thicker laptops for less thermal throttling. Indeed, the M1 MBP got praised because it got thicker and better thermal envelope.

None of this would have happened if the 13" MBP had the same design as the 13" MBA. You would customize the laptop by "adding a fan" as you do now by adding more RAM or more SSD space.

Okay good, there are definitely some less savvy users here whom seem think the act of throttling at all means there is a defect. There’s a major lack of technical understanding beyond what done of these YouTube videos say…which is laymen at best and sometimes just flat out wrong.

 

[NT1440]

Did it? Because as far as I could tell people bashed it for not adding much over the M1 Air even with the fan. Same with the M2 MBP.

Also why would they design a fanned and fanless version of the same computer? The compromise of one of those modes would definitely affect the other, so that seems like a silly way to go about it. The people who want M2 at its full capacity should just get an M2 MBP and suck it up re: the design. I too would love a more-modern display on it (I prefer the body to the Air's personally) but oh well.

Apple’s performance tiers are now designed around a desired thermal envelope. That’s the thing people aren’t getting about Apple silicon.

MX Fanless<MX with fan<MX Pro<MX Max<MX Ultra<Whatever will be in the Pro

That’s the line, the Straight MX series is differentiated by fan or not, then you continue to climb in performance from there.

 

[Xiao_Xi]

Okay good, there are definitely some less savvy users here whom seem think the act of throttling at all means there is a defect.

The question is not whether thermal throttling is good or bad, it should be whether the thermal envelope of the new MBA is better or at least as good as the previous MBA. And so far, I've only read unfounded claims that it is sufficient, but not any hard evidence for it.

 

[leman]

Out of curiosity, what statistic method would you use to demonstrate that the heat dissipation capability of both chassis is identical?

In theory? I’d do multiple side by side sustained load runs and measure the package power and performance and then compare the observed distributions. If you insist of it being hypothesis-driven one can use your favorite modeling framework to compare the the equal mean vs. unequal mean models.

But why bother? For all practical purposes the variance should be fairly low, so if the measured sustained TDP is within 0.5 watts from each other it’s sufficient to claim no difference.

 

[dgdosen]

This thread debates this topic in a thoughtful way (IMO). There's truth on both sides:
- This is the best, fastest, fanless laptop Apple has ever made
- Apple could have done much better

WRT being better - sorry to point to yet another youtuber - but these are good points:
- Few attempts to save weight
- Generously spaced square batteries
- Sparsely populated logic board
- Uninteresting cooling solution

 

[Xiao_Xi]

For all practical purposes the variance should be fairly low, so if the measured sustained TDP is within 0.5 watts from each other it’s sufficient to claim no difference.

If you think the variance could be very low, "about 10" could mean a little more than 10 (i.e. 10.3) for the M1 MBA and a little less than 10 (i.e. 9.7) for the M2 MBA, and that small difference could be statistically significant. Therefore, stating that "the thermal dissipation capacity of both chassis is identical" is a bit adventurous without precise values.

 

[leman]

If you think the variance could be very low, "about 10" could mean a little more than 10 (i.e. 10.3) for the M1 MBA and a little less than 10 (i.e. 9.7) for the M2 MBA, and that small difference could be statistically significant. Therefore, stating that "the thermal dissipation capacity of both chassis is identical" is a bit adventurous without precise values.

“Statistical significance” boils down to “what margin of error am I willing to accept and still hold that X is true”. I believe the term is often overused. If the two measures only differ by 0.5 watts I would be ok with saying that there is no practical difference - that’s 5% accepted relative error given the target value of 10W. I mean, if you want to be precise you can always do repeated measurements and study the distributions so that you can compute the observed variance and adjust your prior accordingly.

 

[Aenean144]

I really don't understand the argument. Previous M1 Air reaches thermal equilibrium at around 10 watts of sustained power consumption. The new M2 Air reaches thermal equilibrium at around 10 watts of sustained power consumption. Experimental evidence thus shows that the thermal dissipation capabilities of both chassis are identical. And yet you keep claiming that the new chassis is worse.

Wizec was incorrect in saying there isn't a heat sink in the M2 MBA. The heatsink is the EMI shield and what looks to be a metallized black tape of some kind on the EMI shield. The cooling design of the M2 MBA appears fine and does its job. The thermal pad modders are prioritizing SoC performance over surface temperatures. Apple would rather have a device that feels cooler longer than something that can get hotter quickly. Modders aren't in the business of selling computers, so I hope people understand the difference there.

I do agree with you that the thermal dissipation between the M1 MBA chassis and M2 MBA chassis is very similar. It's just going to be surface material and surface area. The M1 MBA chassis probably has a little bit more surface area, but it isn't going to be much, like something less than 2 to 3%. The weight of the M2 MBA only dropped by 0.1 lb, but it's battery got larger. With the mass only 0.1 lb different, it just is going to come out in a wash. I wouldn't want to use Midnight or Space Gray in sunlight though. (Sunlight will be around 60 W peak for something the size of a 13" laptop, and a dark color, hot!)

I'm really responding to you as "thermal equilibrium" is really a function of how hot the designer wants to let the internal components and surfaces get, and ambient conditions. I have not seen anyone report Watt numbers for the M2 SoCs, though I haven't been looking. It's strange as PowerMetrics is right there along with other utilities that will tell the user directly how much power is being used for the workloads being used. I would think it would have been linked in various forums by now, and you know, would definitively answer all our questions.

You can make a decent estimate based on M1 numbers. As I recall, the max sustained CPU complex power consumption in the M1 is about 20 W, from Anandtech, using an M1 Mac mini. The rest is math. Transistors went up by about 1.2x. CPU clock rates went up from 3.2 GHz to 3.5 GHz, TSMC N5P is about 0.9x Watt per transistor over TSMC N5, and performance is throttled by about 0.7x (conservative number):

20 x 1.2 x (3.5/3.2)^2 x 0.9 x 0.7 = 18 Watts

By definition, at thermal equilibrium from a sustained max CPU load, the chassis is dissipating 18 Watts. The only difference is how hot the internal components and the surfaces are. I would guess the max chassis temperature will be in the 120 °F range.

Your 10 Watt number seems like one of those "average of the experience and data I've seen" type of number, and those typically include idling time, workloads not actually using the max performance of the SoC, so on and so forth. At 10 W, it probably won't even feel "hot". Maybe feel "warm". An iPhone Pro SoC can probably sustain 8 W and you will definitely feel that with 120 °F surface temps. An iPad Pro 12.9 and MBA has about 5x the surface area. It's going to be higher than 10 W, for both the M1 and M2 MBA.

As another check, 10 Watts will go through the 50 WHr battery in 5 hours. I'm pretty sure someone can find a work load that will run through and M1/M2 MBA battery in 2 hrs. That's 25 Watts. The thermal controller probably won't let it, so, 15 to 18 Watts is probably the max.

Right now, I do think the M2's max power consumption is a little higher than the M1's max power consumption, something like 10%. It all depends on how much of that 10% efficiency gain they got out of N5P, but users get a 20% to 40% performance improvement, depending on load, for it. So net-net, it is a nice sequential model to model improvement. The low workloads all seem the same, based on Apple's WiFi web browsing and Apple TV video benchmark numbers.

 

[leman]

Until there is solid evidence, his claim is as far-fetched as yours.

Notebookcheck has published their review. Look under stress test for throttling behavior of M2 vs M1 Air. If anything, the new Air seems to have slightly better thermal performance

Test Apple MacBook Air M2 Entry - Das sehr gute, aber zu teure Alltags-MacBook

Notebookcheck testet das neue Apple MacBook Air M2 in der Farbe Midnight und der Einstiegskonfiguration mit 8 GB RAM und 256 GB SSD-Speicher.

www.notebookcheck.com

Your 10 Watt number seems like one of those "average of the experience and data I've seen" type of number, and those typically include idling time, workloads not actually using the max performance of the SoC, so on and so forth.

No, what I meant is that the chassis cannot dissipate more than 10W in sustained operation - when you continuously keep pushing the system without any periods of idle activity it will keep throttling before stabilizing at 10W TDP. That’s the equilibrium I was referring to. Check the stress test graphs in the link above.