Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Modern SoCs throttle in a fanless design...
- Thread starter exoticSpice
- Start date
- Sort by reaction score
47c is above the temp required to burn skin and well into the range that will significantly degrade battery longevity.
Congrats! you just reduced the life of your MacBook battery, voided your warranty and got a lesser improvement to sustained throughput that you could have had with a MacBook Pro for similar money - with a warranty.
- It takes at least 5 minutes, at 48C to get a 3rd degree skin burn.
- The mod makes a small portion of the rear chassis - away from the battery which is in the front - a couple of degrees warmer. The effect on the battery aging is going to be negligible
- "However, cell temperature should not exceed 55 • C to prevent accelerated aging or safety-critical state"
- The thermal pad mod actually makes the Air faster in some cases than the Pro:
- The mod is reversible if done right, so I wouldn't really be worried about warranty.
Last edited:
Thanks for the links! Always great to see some data, though I wish they actually labeled the X-axis or at least put what the X-axis meant in a figure caption.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
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.
I'm confused as to the differences between the combo Cinebench + 3DMark Extreme stress test and the power consumption where they show the power consumption for various loads. The Cinebench+3Dmark stress test is measured with an OS utility while the power consumption is measured by a multimeter.
The power consumption tests show power consumption of 23 W at the end of the data for a Furmark+Prime95 run and 22 W for a Witcher Ultra run. These numbers look like they include the display (which is about 7.4 W at max brightness) Then the other power consumption chart shows 23 W for a Cinebench R23 multi run that hasn't flattened and 19 W for the Witcher run. The latter don't appear to include the display power, if I understand what "external monitor" means. The Witcher runs look like they've flattened out and hit equilibrium. I don't think there are any major power users other than the CPU, GPU, and displays. Ones that would make up for a 10 W difference, at least.
So, these power numbers are higher than the Cinebench + 3DMark Extreme test. So, why the difference? Either the power tests don't show equilibrium for throttling and need to run longer, or the Cinebench+3DMark test is throttling package power for other reasons, or perhaps the reported package power doesn't include a high power user in the package.
Yeah I agree that methodology documentation is severely lacking. I was also confused by that CB23 test. Maybe they took breaks just long enough for the thermals to recover?
Regardless, as lacking as these tests are they seem to show that the new M2 Air chassis has a larger thermal buffer compared to the old one.
You DO realize that if you require each generation of a passively cooled laptop to have better heat dissipation than the last one, you'll be quickly caught up by, well, physics?
In order to increase the amount of heat the machine can dissipate per a unit of time, you either have to
- add fans (can't do)
- increase its surface temperature (uncomfortable above a certain point, also regulated)
- increase the total surface (heavily limited by design and the target size of the device)
Basically, there's only so much you can do with a slab of aluminum that can't get hotter than 45 °C and larger than 13" diagonally + pretty thin.
It doesn't matter how many heatsinks or pads or put INSIDE the machine, because the boundary at which you're thermally limited is OUTSIDE the machine.
Tangentially - can anyone comment on:
- Did the 12" MacBook have any cooling tech?
- Does any ipad pro have any cooling tech?
- Does any iOS device have any cooling tech?
Has anyone done any testing on the M1 air to see if removing the heat sink had any impact in it's performance?
The M1 has made me realize I'll be all in on whatever class of computer that can be keyboard-width and deliver a thermal-protection-free punch as good or better than the M1.
- Did the 12" MacBook have any cooling tech?
- Does any ipad pro have any cooling tech?
- Does any iOS device have any cooling tech?
Has anyone done any testing on the M1 air to see if removing the heat sink had any impact in it's performance?
The M1 has made me realize I'll be all in on whatever class of computer that can be keyboard-width and deliver a thermal-protection-free punch as good or better than the M1.
Last edited:
This is correct over a long time period, but the thermal mass directly contacting the SoC matters quite a bit in the short term (seconds and minutes). Most computing is bursty, and you want to draw that initial burst of heat away from the SoC to buy some time before hitting the max temp. During the next dip in processing the heat has time to flow outward allowing the thermal mass to cool enough that it has room for the next burst. The thermal mass evens out the temperature during the burst/dip cycles and reduces the amount of time the SoC operates at thermal max.
This can help stabilize frame rates during gaming sessions, for example.
Leman posted the M2 MBA review from notebookcheck. They compare it directly to the M1 MBA. The results are what you would expect. The M2 MBA has better performance than the M1 MBA, about what Apple is advertising. The M2 MBA throttles a little slower than the M1 MBA, which is indicative that the M2 MBA has a better cooling system than the M1 MBA. The M2 MBA has comparable surface temperatures to the M1 MBA. The M2 uses a little more power than the M1, especially in single core, but it is hard to say how it impacted results. That the throttle rate, ie, how fast the M2 MBA throttles, appears slower than the M1 MBA and is doing it with a little hotter SoC implies the M2 MBA removes heat a little faster.
The rMB12 had passive cooling. Ie, no fan. They used 5W TDP Intel processors, or 7W with TDP-up. 5W is about what you see for a phone processor, or perhaps a tablet processor. From the teardowns, it looks like a heat spreader wasn't used. Intel likely built the processor at Apple's request. Intel took a Kaby Lake and limited it to 5 W, or 7W in the TDP-up config. They could turbo to 3.4, 3.6 GHz at around 20 W levels, but would drop down to 1 to 1.2 GHz for sustained levels at 5 to 7 W.
For the iPad Pro, it looks like they started using metallized (copper?) tape with a graphene coating (?) starting with A12Z models and continues with the M1 models. The tape acts as the heat spreader. Performance is about the same as the M1 MBA? Maybe? Hard to say. I think the sustained Watts is on order the same as M1 MBA, perhaps 1 Watt less. Something like that. Starting with iPadOS 16, people will be able to really push the M1 models and we will see, so we might so more data after iPadOS 16 is available.
For iPhones, I don't think they use the combo EMI+heat spreader metallized tape yet. Maybe with the A16? If the iPhone Pro 14 has a 48 MP camera, it will be interesting how much heat processing 48 MP images will generate. Feels inevitable that iPhones will start using this metallized tape as heat spreaders though. Features that sell phones are getting harder to come by. The M2 MBA uses a thin piece of metal plus the metallized tape as the heat spreader.
So far the published tests suggest that the thermal performance of the new Air has improved compared to its predecessor. On the graphs published by notebookcheck it is able to stay in higher power state for longer at least.
You can also constantly redline the engine in your car but not without a cost.
M2 is built on TSMC like AMD which has a Tjmax of 95C. Even Intel 7 node built for high frequency has a Tjmax of 100C while M2 junction temp is measured at ~108C. In the end, if it burns out then you were using it wrong and Apple benefits from another sale.
M2 is built on TSMC like AMD which has a Tjmax of 95C. Even Intel 7 node built for high frequency has a Tjmax of 100C while M2 junction temp is measured at ~108C. In the end, if it burns out then you were using it wrong and Apple benefits from another sale.
In the Six Colors review, both throttle, but the M2 remains significantly faster.
M2 MacBook Air Review: A new era
The definitive Mac of the last decade was the 13-inch MacBook Air, introduced in 2010 and featuring a wedge-shaped design. It influenced all the Apple laptops (and let’s face it, a lot of PC …
sixcolors.com
This is not a situation where the M2 MBA is always overheating and always running slow. For almost every task you throw at it, the M2 MBA will remain cool and fast. You have to throw something at it that uses all the CPU cores and all the GPU cores at once to get it to heat up to its prescribed limits, at which point it starts slowing down to cool off. Even the it only slows down about 10-20% and remains faster than the M1 even when the M1 is not throttling.
Also, that iFixit video that declared that the M1 had a heat spreader but the M2 doesn’t flat out was wrong. What they were deriding as “too thin to be a heat spreader“ is exactly that. Yes, the heat spreader in the M1 was thicker but it had less surface area. The M2 heat spreader is larger and with more surface area can dissipate heat faster. It also has a graphene film to help transfer the heat away. IFixit was largely ignoring that. This was not the iFixit A-team. They even tore the MBA apart and then put it back together before they did their performance tests eliminating the default state for their tests!
BTW here is information about the Apple patent on that graphene heat spreader.
Fresh Apple Patent Shows New Graphene Heat Dissipator, That Could Get Rid of Potential Overheating | GSMDome.com
Fresh Apple Patent Shows New Graphene Heat Dissipator, That Could Get Rid of Potential Overheating - In reltion with: anti overheating Apple patent, anti overheating iPhone, Apple patent, Apple patent graphene heat dissipator, graphene heat dissipator, graphite sheets, overheating prevention
It is ridiculous for iFixit to say that the M2 does not have a heat spreader. That is exactly what that is. It is in the same position as the heat spreader in the M1 and is attached the SOC by thermal paste. Yes, it is thinner metal but the surface area is larger than the one in the M1. In heat dissipation, surface area matters more than mass. This is not a heat sink that aborbs heat. It is a spreader that dissipates heat into the case. It is even covered in the new graphene thermal transfer film. I am shocked that iFixit missed that.
Fresh Apple Patent Shows New Graphene Heat Dissipator, That Could Get Rid of Potential Overheating | GSMDome.com
Fresh Apple Patent Shows New Graphene Heat Dissipator, That Could Get Rid of Potential Overheating - In reltion with: anti overheating Apple patent, anti overheating iPhone, Apple patent, Apple patent graphene heat dissipator, graphene heat dissipator, graphite sheets, overheating prevention
When you say “drastically better performance” how much are you talking about? In the tests that I saw, the thermal throttling only had at most a 10-20% reduction in speed. How much of that would a thermal pad offset? At most it could be no more than 10-20% which is not very “drastic”. Especially since very few uses cases will ever get into the throttling temperatures. This seems like a lot of risky modification for very little benefit.
So completely impractical for using the device as intended then.
Also, by dumping that heat into the case you will be causing battery temps to be much hotter thus degrading your battery.
Seriously, if you want an M2 to run sustained workloads faster the solution is clear: Buy an M2 13" Macbook Pro.
I really don't understand why some people seem so determined to buy a product that is not intended for their use case, for pretty much the same money and almost identical portability, to hack away at, void warranty, degrade battery life and end up with worse performance anyway than just buying the right machine or the job.
Yes, the machine throttles (but even throttled is faster than virtually any comparable form factor machine on the market). Yes there is a reason Apple actually insulated the chassis from the CPU and doesn't dump heat into it. No, this is not an oversight. It's a conscious design choice for this market segment.
I would wager that most MacBook Air customers are intending to use this machine away from a desk (e.g., at least some of the time on their lap).
Last edited:
I find it hard to be critical of iFixit. They're not hyping about benchmarks like others. Nor do they post multiple videos rehashing the same subject. They seem to tell it like it is.It is ridiculous for iFixit to say that the M2 does not have a heat spreader. That is exactly what that is. It is in the same position as the heat spreader in the M1 and is attached the SOC by thermal paste. Yes, it is thinner metal but the surface area is larger than the one in the M1. In heat dissipation, surface area matters more than mass. This is not a heat sink that aborbs heat. It is a spreader that dissipates heat into the case. It is even covered in the new graphene thermal transfer film. I am shocked that iFixit missed that.
Fresh Apple Patent Shows New Graphene Heat Dissipator, That Could Get Rid of Potential Overheating | GSMDome.com
Fresh Apple Patent Shows New Graphene Heat Dissipator, That Could Get Rid of Potential Overheating - In reltion with: anti overheating Apple patent, anti overheating iPhone, Apple patent, Apple patent graphene heat dissipator, graphene heat dissipator, graphite sheets, overheating preventionwww.gsmdome.com
I am skeptical of your touting the M2 Air heat dissipation. If you look at any of those youtube videos that includes a thermal camera shot, including the iFixit video, the images show quite a temperature bullseye on the exact spot of the SOC. If there was some great graphene solution (which would surprise me as Apple Marketing has not hyped it) I'd think those would look more thermal-camera-friendly.
I find the conclusion that "Apple was fine with the perf of the M2 in the Air, and didn't try overly hard to passively cool it" to be a reasonable conclusion. Because as pointed out above, there's not much passive cooling in similar Apple devices. It's the best they've got for passively cooled laptops, and it arguably runs circles around the competition.
Why would you expect the thermal hotspot to be somewhere else than on the only heat-producing element?
You might try a simple google search. First hit that I got.
Thermal Management: Heat Spreader - The Global Graphene Group
Thermal Heat Spreader Overview and Value Proposition G3’s graphene thermal heat spreader provides superior thermal management in the form of high thermal conductivity films / sheets and enhanced heat dissipation in a lightweight and non-combustible film The thermal heat spreader can be easily...
www.theglobalgraphenegroup.com
Obviously Apple isn't hyping it because it is a normal solution for a thin heat spreader. It seems very likely that Apple used something like this for their heat spreader on the M2 MBA.
First off, not a hardware engineer and don't claim to be. I agree that the hottest spot would be that heat source.
I fast-forwarded through the iFixit video, and didn't see a thermal camera shot. Not sure where I saw one, but here's a couple found on a search:
First - a scan of the M1 Air (left) vs M2 Air (right) - and you see the M1 heat center being around the size of the heat spreader, whereas on the right - the hottest spot looks to be much more zeroed in on the SOC.
I'd expect (as a layman) heat dissipation to show a spread a bit more evenly over any heat sync - as in this mac studio image:
I'm not sure where in that blob the SOC might be.
For reference:
first image from:
I'm not an expect either, but since we are already two layman's discussing this stuff, I am not sure one can interpret much from these pictures. The second one looks at the heat spreader directly, while in the first one there is the aluminium chassis in the way. I can imagine that complicating things already. And then, we don't know much about how the heatmaps are normalised etc. The only thing I can take from here is that the M2 is hotter in this specific example, but nothing about the heat is distributed over the cooling system.
The M1 MBA could dissipate a little more than 10W. The notebookcheck review of M2 MBA shows that the M1 MBA consumes around 13W at R23 multi and it appears that it is not thermally throttled. Unfortunately, the M2 draws around 21W and is throttled, so it is unclear how much heat the M2 MBA could dissipate.
The M1 MBA is absolutely thermally throttled running multi on Cinebench. I tested months ago and saw a 10% or so drop from peak to equilibrium. How did you get the idea that the M1 MBA didn't throttle in Cinebench?
Unlike the M2 MBA, the M1 MBA is still more efficient than the M1 13" MBP in R23 multi.
With the 30 minute stress test, the M1 MacBook Pro doesn't throttle. The M1 MacBook Air does.