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.
M1 Air with Thermal Pad Mod
- Thread starter downshiftdre
- Start date
- Sort by reaction score
Yes, there is a postage stamp sized thermal " sponge " fitted to one end of the heat sink. The thermal sponge can compress against the back cover of the Air and dissipate heat.
There's a line of thought that says that it was omitted from the base 7 core model so as to differentiate the benchmark scores. Fitting a thermal pad to the base Air gives benchmark scores similar to the M1 MacBook Pro, but it throttles as there is no fan fitted.
It is a very small one covering the deepest corner of the heatsink. Check back to some of the earliest posts.
Clearly Apple needed to boost the performance of the top MBA above the base model and a thermal pad was the way they chose.
Clearly Apple needed to boost the performance of the top MBA above the base model and a thermal pad was the way they chose.
My modded base model M1 MBA exceeds the performance of the top M1 MBP when on a stand, under sustained load, as the previous graphs and benchmarks show.
Thank you for your prompt reply.
So there wouldn't be much help from some additional mods for the MBA (8gpu), since there is already one.
Is it, right?
No, the deepest corner of the heatsink is furthest away from the M1 chip, and the thermal 'sponge' is tiny. A $2 2mm thermal pad covering the shallow majority of the heatsink will make all the difference. The exchange is performance for a hot bottom case. Add a stand and the sustained performance rips through the MBP's benchmarks but could superheat your lap without one.
Apple is adding a Low Power Mode to the Battery System Preferences in macOS Monterey. I wonder if that along with the thermal pad mod would allow use on the lap without overheating. Maybe get the best of both.
Edit: I just tested my M1 MBA on Monterey with low power mode for the Power Adapter. The mode seems to limit total package power to around 4 W. I tested both CPU and GPU and got the same 4 W limiter. So I don't think the bottom case of the MBA will warm up at all in this mode. WildLife Extreme I got about 2800 (16 fps) when power limited and about 4900 (29 fps) when full speed. At full speed the GPU was pulling over 10 W. For Cinebench 23 I got 4360 when limited and my normal score is around 7020.
Last edited:
Very interesting. So, the overriding question therefore is: does the mod make any difference in low power mode?
I love the idea of running this MBA, that trashes the performance of my old 8-core Mac Pro tower that used to heat the house, on 4W of juice, e.g. a small USB solar panel. Apple should add some PV panels into the MBA for ultimate greenness.
Just ran another test using the new Low power mode in Monterey. Using the command line powermetrics tool, I can see that with both the CPU and GPU maxed out (I ran Cinebench 23 and 3DMark Wild Life Extreme at the same time), the total package power stayed just about 4 Watts. This time I measured CPU temperatures and never got above 60° C.
If someone who has done the thermal pad mod and is running Monterey could check out what happens with Low power mode, that would be great information.
Looking for input from others that have done this mod.
I installed thermal pads on my 8-core GPU MBA 1GB/1TB back in early February and while I did notice a performance difference in Geekbench, RAW photo editing, and in FCP, all without a difference in usability or noticeable temps, I have recently noticed that my battery health is deteriorating faster than other Appel laptops I've owned. According to Coconut Battery after owning the laptop since November 2020 I am down to 90% of my original capacity (although the Battery Health reading within System Preferences says 97%), which is much worse after this much time than I would have expected. Obviously I knew when installing the thermal pad that more heat might be transferred throughout the lower case and therefore also the battery, but I wanted to see if others have had similar results so far.
Thanks in advance!
I installed thermal pads on my 8-core GPU MBA 1GB/1TB back in early February and while I did notice a performance difference in Geekbench, RAW photo editing, and in FCP, all without a difference in usability or noticeable temps, I have recently noticed that my battery health is deteriorating faster than other Appel laptops I've owned. According to Coconut Battery after owning the laptop since November 2020 I am down to 90% of my original capacity (although the Battery Health reading within System Preferences says 97%), which is much worse after this much time than I would have expected. Obviously I knew when installing the thermal pad that more heat might be transferred throughout the lower case and therefore also the battery, but I wanted to see if others have had similar results so far.
Thanks in advance!
I bought mine 8th December and did the mod late January. Apple System Report also reports mine as at 97% capacity. Is there any reason to doubt the Apple System Report?
Hello,
I am gonna try the thermal pad mod today but I have a few questions firstly and after reading most of the posts around here I got a bit worried.
My MBA is an 8 core GPU, 8 core CPU, 16GB RAM, 512 GB storage.
At this moment, right before the thermal pad mod, if I do a Cinebench R23 10 minutes benchmark I only managed to get a max score of around 6200 for the multicore test. I can see people around here getting scores of 7000+ doing the same benchmark as I did, unmodded and on a flat desk (raw MBA with no additional help). My room temps are not that high, around 23C and I made sure I closed all running apps and tools. Is there anything I should worry about?
Also, going to the system report I can see my battery life is at 98% as of today. My MBA being only 5 months old. Anything I should be concerned about?
Thanks in advance for your suggestions. I will come back with the results after the mod. As thermal pads i will be using Thermal Grizzly 2mm and 3mm pads.
I am gonna try the thermal pad mod today but I have a few questions firstly and after reading most of the posts around here I got a bit worried.
My MBA is an 8 core GPU, 8 core CPU, 16GB RAM, 512 GB storage.
At this moment, right before the thermal pad mod, if I do a Cinebench R23 10 minutes benchmark I only managed to get a max score of around 6200 for the multicore test. I can see people around here getting scores of 7000+ doing the same benchmark as I did, unmodded and on a flat desk (raw MBA with no additional help). My room temps are not that high, around 23C and I made sure I closed all running apps and tools. Is there anything I should worry about?
Also, going to the system report I can see my battery life is at 98% as of today. My MBA being only 5 months old. Anything I should be concerned about?
Thanks in advance for your suggestions. I will come back with the results after the mod. As thermal pads i will be using Thermal Grizzly 2mm and 3mm pads.
I think those un-modded scores posted above are about as reliable as self-measured penis sizes.
If you read the whole thread, and that would take days, you will see that Cinebench only produces an average score over multiple runs. To know the split performance as the MBA throttles back you need to do a longer test, e.g. an hour, record the numbers each pass then do the 'de-averaging' calculations as I explained in one of my early posts. This reveals that the MBA, modded or not, reaches a steady state of performance and that the modded is far better (I don't have the number to hand). The raw CB figures just show a steady trend towards the steady state but never actually get there.
As far as battery life is concerned, Apple guarantees at least 80% capacity after 1000 charge/discharge cycles, as measured by System Info. I am on 17 cycles and 97% after 7 months - but mostly used plugged in. Capacity falls like a half-life graph, fast at first then forever more slowly.
I don't think you have anything to worry about.
If you read the whole thread, and that would take days, you will see that Cinebench only produces an average score over multiple runs. To know the split performance as the MBA throttles back you need to do a longer test, e.g. an hour, record the numbers each pass then do the 'de-averaging' calculations as I explained in one of my early posts. This reveals that the MBA, modded or not, reaches a steady state of performance and that the modded is far better (I don't have the number to hand). The raw CB figures just show a steady trend towards the steady state but never actually get there.
As far as battery life is concerned, Apple guarantees at least 80% capacity after 1000 charge/discharge cycles, as measured by System Info. I am on 17 cycles and 97% after 7 months - but mostly used plugged in. Capacity falls like a half-life graph, fast at first then forever more slowly.
I don't think you have anything to worry about.
Thank you so much for the prompt reply
Got a rock off from my heart I'll come back with results/impressions after the mod.
I did the thermal pad mod to my 2020 i5/8/512 Air and my battery is also down to about 89.7% after 49 cycles. I don't know if its related or not. I get much better temps and GB5 Scores are around 3400.
I finally did the thermal pad mod and I got a 10% improvement, indeed. Great news!
The results in Cinabench R23
Before the mod: 6100 ish
After the mod: 6800 ish
Even though I did this mod I'm still not even close to the scores posted by other people. Would it be because i don't use a stand of a cooler stand under my laptop? Can a stand improve my scores even more?
The results in Cinabench R23
Before the mod: 6100 ish
After the mod: 6800 ish
Even though I did this mod I'm still not even close to the scores posted by other people. Would it be because i don't use a stand of a cooler stand under my laptop? Can a stand improve my scores even more?
Did you de-average the results and run for 60 mins or so? Also make sure nothing else is running and you start from 'cold' to compare.
A stand for me made as much difference as the mod.
From post #96 "Yes, the approximate 2-minute score updates are the average per pass to that point. They do not tell you the score of that particular pass, it has to be calculated:
(latest score x total duration so far - previous score x total duration to its scoring) / duration of last pass."
It is an hour of writing times and scores, then entering them into a spreadsheet and doing the calculations and finally the beautiful chart
A stand for me made as much difference as the mod.
From post #96 "Yes, the approximate 2-minute score updates are the average per pass to that point. They do not tell you the score of that particular pass, it has to be calculated:
(latest score x total duration so far - previous score x total duration to its scoring) / duration of last pass."
It is an hour of writing times and scores, then entering them into a spreadsheet and doing the calculations and finally the beautiful chart
Last edited:
Your ambient temp is 10C greater than that of the data presented in the charts, hence why the mod is far less efficient in your case with subsequent lower results. If in hotter climates your better served by a system with active cooling if sustained performance is required.
FWIW M1 13" MBP 8GB base model, macOS 11.5.1, Cinebench R23; 30 minute run, ambient temp 26C, max temp 92C, max fan 4.5K RPM, max battery temp 32C, scored 7814 pts. Monitored by Macs Fan Control, fan set to Apple default auto setting.
The M1 MBP active cooling maintains full CPU speed with negligible impact from the ambient temperature. I've tried several passively cooled notebooks and they all have the same positives and negatives with the latter being they slowdown when impacted by elevated temperature.
The modification can and does increase the M1 Air's thermal capability to maintain higher CPU/GPU frequency longer with the trade off of a significantly higher base plate temperature and increased battery temperature. Nor will the Air surpass the M1 MBP, in the right conditions it will be close for a period of time, that's it...
Q-6
Last edited:
Since I'm a tinkerer, I've had a little project where I...
- Took a 3mm thick piece of pure copper bar and cut it so that it would fit perfectly over the portion of the original heatsink that covers the package area.
- Sanded the end that would go towards the back of the chassis slightly rounded to ensure there would be no conflict with the shape of the bottom plate.
- Tested multiple times if everything would come together and the bottom plate fit properly. Used a little moisturizing cream to see contact points.
- Then some proper Noctua thermal paste on the top and bottom of the copper piece (the plastic insulating stuff from the chassis bottom plate removed).
- Last assembly, then removed the bottom plate to see if good contact was being made. It was.
Result? Well, it is obviously more thermal mass so things heat up a lot slower compared to stock or with thermal pads, but the ability of the bottom plate to transfer heat away is somewhat limited.
In Cinebench when watching temperatures and running powermetrics in terminal to see cpu frequency, it takes a good 7-8 or so minutes before any tendency of throttling (dropping down from constant 3200mhz) while watt draw stays in the 15w range all the time. Cinebench score improved with about 200 points compared to thermal pads (up to around 7800pts), but that could be down to room temperature and loads of different things so I'll take that with a grain of salt.
What annoys me is that it is still clearly throttling a little bit in the 10minute Cinebench run, but then again its a passive design and has limitations regardless of how efficiently you transfer heat from the package to the bottom plate.
What I like is that it heats up less during moderate load and short bursts, such as creating previews after importing RAW images in Capture One or doing a quick short video encode. Sustain load champs these M1 airs will never be. Battery temps (when run from the wall) stays around 30-31c as well.
Should you do what I did? Nah, its not worth the time and it might get very tricky to get the right tickness of this massive "shim".
What will I do next? Well, I have more copper of various dimensions and an idea to replace the stock heatsink completely, foregoing the hovering wing part of it which is useless once you connect it to the chassis bottom and just make a single massive piece of copper around 5mm thick that connects the cpu directly to the bottom plate. The stock heatsink mounting springs can be used with nuts, and countersunk bolts on the heatsink side. But, that is a bit of work and I lack machining tools except a drill press, files and sand paper, so probably not
- Took a 3mm thick piece of pure copper bar and cut it so that it would fit perfectly over the portion of the original heatsink that covers the package area.
- Sanded the end that would go towards the back of the chassis slightly rounded to ensure there would be no conflict with the shape of the bottom plate.
- Tested multiple times if everything would come together and the bottom plate fit properly. Used a little moisturizing cream to see contact points.
- Then some proper Noctua thermal paste on the top and bottom of the copper piece (the plastic insulating stuff from the chassis bottom plate removed).
- Last assembly, then removed the bottom plate to see if good contact was being made. It was.
Result? Well, it is obviously more thermal mass so things heat up a lot slower compared to stock or with thermal pads, but the ability of the bottom plate to transfer heat away is somewhat limited.
In Cinebench when watching temperatures and running powermetrics in terminal to see cpu frequency, it takes a good 7-8 or so minutes before any tendency of throttling (dropping down from constant 3200mhz) while watt draw stays in the 15w range all the time. Cinebench score improved with about 200 points compared to thermal pads (up to around 7800pts), but that could be down to room temperature and loads of different things so I'll take that with a grain of salt.
What annoys me is that it is still clearly throttling a little bit in the 10minute Cinebench run, but then again its a passive design and has limitations regardless of how efficiently you transfer heat from the package to the bottom plate.
What I like is that it heats up less during moderate load and short bursts, such as creating previews after importing RAW images in Capture One or doing a quick short video encode. Sustain load champs these M1 airs will never be. Battery temps (when run from the wall) stays around 30-31c as well.
Should you do what I did? Nah, its not worth the time and it might get very tricky to get the right tickness of this massive "shim".
What will I do next? Well, I have more copper of various dimensions and an idea to replace the stock heatsink completely, foregoing the hovering wing part of it which is useless once you connect it to the chassis bottom and just make a single massive piece of copper around 5mm thick that connects the cpu directly to the bottom plate. The stock heatsink mounting springs can be used with nuts, and countersunk bolts on the heatsink side. But, that is a bit of work and I lack machining tools except a drill press, files and sand paper, so probably not
Interesting, not seen much past 7800 on Cinebench R23 for the M1 which I think is the SOC's upper limit. So your on the right path with decent battery temps. You'll need to run multi hour tests to see how the battery fairs. Seen some where the battery is approaching 45C which isn't desirable.Since I'm a tinkerer, I've had a little project where I...
- Took a 3mm thick piece of pure copper bar and cut it so that it would fit perfectly over the portion of the original heatsink that covers the package area.
- Sanded the end that would go towards the back of the chassis slightly rounded to ensure there would be no conflict with the shape of the bottom plate.
- Tested multiple times if everything would come together and the bottom plate fit properly. Used a little moisturizing cream to see contact points.
- Then some proper Noctua thermal paste on the top and bottom of the copper piece (the plastic insulating stuff from the chassis bottom plate removed).
- Last assembly, then removed the bottom plate to see if good contact was being made. It was.
Result? Well, it is obviously more thermal mass so things heat up a lot slower compared to stock or with thermal pads, but the ability of the bottom plate to transfer heat away is somewhat limited.
In Cinebench when watching temperatures and running powermetrics in terminal to see cpu frequency, it takes a good 7-8 or so minutes before any tendency of throttling (dropping down from constant 3200mhz) while watt draw stays in the 15w range all the time. Cinebench score improved with about 200 points compared to thermal pads (up to around 7800pts), but that could be down to room temperature and loads of different things so I'll take that with a grain of salt.
What annoys me is that it is still clearly throttling a little bit in the 10minute Cinebench run, but then again its a passive design and has limitations regardless of how efficiently you transfer heat from the package to the bottom plate.
What I like is that it heats up less during moderate load and short bursts, such as creating previews after importing RAW images in Capture One or doing a quick short video encode. Sustain load champs these M1 airs will never be. Battery temps (when run from the wall) stays around 30-31c as well.
Should you do what I did? Nah, its not worth the time and it might get very tricky to get the right tickness of this massive "shim".
What will I do next? Well, I have more copper of various dimensions and an idea to replace the stock heatsink completely, foregoing the hovering wing part of it which is useless once you connect it to the chassis bottom and just make a single massive piece of copper around 5mm thick that connects the cpu directly to the bottom plate. The stock heatsink mounting springs can be used with nuts, and countersunk bolts on the heatsink side. But, that is a bit of work and I lack machining tools except a drill press, files and sand paper, so probably not
FWIW M1 13" MBP 8GB base model, macOS 11.5.1, Cinebench R23; 30 minute run (2nd consecutively), ambient temp 26C, max temp 92C, max fan 4.5K RPM, max battery temp 32C, scored 7814 pts. Monitored by Macs Fan Control, fan set to Apple default auto setting.
Be fun to see if you can get the M1 to hold with the M1 MBP for 30/60 minutes. Might be possible, but will be hell of a hot on the bottom LOL...
Q-6
Clearly just having some fan running, no matter how crappy or whiny it is, is a considerable competitive advantage
In my opinion the passive cooling could have been considerably improved by Apple if they wanted to. Likely at a different price point though. The chassis design is very limiting, there is not really any room to for instance use heathpipes to spread heat to different points of the chassis bottom. I guess that is the Apple design over functionality mantra in practise (biased pc user here).
I am absolutely certain an M1 Pro is going to come out very slightly on top for sustained load over a few hours in terms of pure numbers crunched compared to my mod, but I would never run my Air like that. I got faster rigs.
Simply because of the chassis design, both units batteries are at serious risk for less than advantageous temperatures. But the Air moreso, since even removing the heat directly through the bottom plate, is bound to heat up some of the air and components internally that does not in themselves produce a lot of heat. Think about it, the bottom plate releases heat to the air outside the chassis...but also inside.
In essence this is a flawed design, IF you run it outside its intended operating parameters. And those are not encoding 1 hour Youtube podcasts or gaming something with serious graphics from the past decade. It is a short-bursts-and-mostly-close-to-idle type machine and for that it is amazing.
For me, losing 10% of the peak performance on a fast CPU isn't much of a competitive shortfall when compared to listening to a fan while using the computer. I'll admit the M1 MacBook Pro would probably be fine for me too but I actually like the MacBook Air form factor much better than the squared off MBP (I'm neutral on the Touch Bar).Clearly just having some fan running, no matter how crappy or whiny it is, is a considerable competitive advantage
M1 MBP fan will only spool up at 80C plus, with the cooling solution designed to remove the excess heat fast then shutdown the fan. In use you are hardily aware the MBP has an active fan as between the mass of the chassis and background noise the fan when running is near inaudible. Performance wise your looking at 20% on sustained loads, for me the additional performance, longer battery life and brighter display sold the MBP. Touchbar I can take or leave as it both adds and takes away from usability, mostly being a solution to a question that was never asked or required.
I prefer the Air's design, however from experience I know that passively cooled systems and heavy workloads are yet to be resolved. Has to be said the M1 MBP's battery life is nothing less than impressive nor exaggerated. I'd really like to see Apple bring back the 12" MacBook with M1 or updated SOC as seems an ideal solution for a small formfactor notebook with far less of a trade off than the original 12" Retina MacBook.
Q-6
FWIW, 6 months with the thermal mod, system information indicates 98% battery health after 67 cycles.