M1 Air with Thermal Pad Mod

[downshiftdre]

Ok 30 minute cinebench results are in:

In summary, no throttling.

30 min Cinebench R23 score: 7734!!! (in line with the 7713 10 minute score).

Before temps
CPU cores: 22C
Battery: 22C
Battery Proximity: 22C

During temps (halfway 15min mark)
CPU cores: 82-90C
Battery: 35C
Battery Proximity: 28C

End temps (1min remaining)
CPU cores: 82-90C
Battery 35C
Battery Proximity: 28C

Battery drained from 100% to 80% within the 30 minutes.

My conclusion is that the battery is not impacted by the heat in the backshell as shown by the “battery proximity” temperature only rising 6C delta from 22C to 28C. The actual battery temp rose from 22C to 35C likely due to the massive and rapid battery drain occuring during the long and high powered cinebench test.

 

[Tenkaykev]

Ok 30 minute cinebench results are in:

In summary, no throttling.

30 min Cinebench R23 score: 7734!!! (in line with the 7713 10 minute score).

Before temps
CPU cores: 22C
Battery: 22C
Battery Proximity: 22C

During temps (halfway 15min mark)
CPU cores: 82-90C
Battery: 35C
Battery Proximity: 28C

End temps (1min remaining)
CPU cores: 82-90C
Battery 35C
Battery Proximity: 28C

Battery drained from 100% to 80% within the 30 minutes.

My conclusion is that the battery is not impacted by the heat in the backshell as shown by the battery proximity temperature only rising 6C delta from 22C to 28C. The actual battery temp rose from 22C to 35C likely due to the massive and rapid battery drain occuring during the long and high powered cinebench test.

Many thanks for doing this. Very useful information.

 

[kepler20b]

As @adrianlondon mentioned, I don't think I'd bother with the thermal paste UNLESS doing what you're looking to do doesn't seem to produce the expected thermal spread to the cover. (eg If the cover gets hotter, the thermal pads are working and hence so is the heatsink, and therefore the current thermal paste.) That would save you the hassle of peeling the heatsink.
I'm interested in your results, vs @downshiftdre. If you are able to accomplish similar improvements with a slower heat spread and/or (hopefully) a lower overall cover temp (by avoiding the heatsink directly above the CPU from piping heat directly into the cover, using air insulation as a damper), it would be even better IMHO (better "solution" for laps, especially!).

I know full well that reapplying thermal paste will have a negligible impact. this is well documented in the MBP intel 13". While it did decrease ambient temperatures by a few degrees, it has no real impact on actual load temperatures because of the flimsy cooling systems that apple uses. its too easy to saturate the heatsink because the heatsink is simply too small.


When my MBA comes I will open it and reapply the thermal paste because I feel like it. There are risks involved, of course, like losing a screw. It's a computer. anyone with even a modicum of electronics or computer building experience should be able to disassemble and reassemble it with 0 issues. there are unlimited YouTube videos showing people step by step how to remove a motherboard out of a MBP/MBA for the unexperienced.

 

[downshiftdre]

I know full well that reapplying thermal paste will have a negligible impact. this is well documented in the MBP intel 13". While it did decrease ambient temperatures by a few degrees, it has no real impact on actual load temperatures because of the flimsy cooling systems that apple uses. its too easy to saturate the heatsink because the heatsink is simply too small.


When my MBA comes I will open it and reapply the thermal paste because I feel like it. There are risks involved, of course, like losing a screw. It's a computer. anyone with even a modicum of electronics or computer building experience should be able to disassemble and reassemble it with 0 issues. there are unlimited YouTube videos showing people step by step how to remove a motherboard out of a MBP/MBA for the unexperienced.

Please provide details on your thermal paste reapplication procedure. In particular what you will do with the unified memory chips. Cant tell if that is thermal paste or black insulating goop on those chips.

You might be able to eek out some additional performance which would give the heatsink and thermal pad a head start in getting heat out.

 

[kepler20b]

Please provide details on your thermal paste reapplication procedure. In particular what you will do with the unified memory chips. Cant tell if that is thermal paste or black insulating goop on those chips.

You might be able to eek out some additional performance which would give the heatsink and thermal pad a head start in getting heat out.

my MBA is on backorder, got it for 899 on the blackfriday deal. I honestly have no idea when its going to come.


As long as the thermal paste you're using is NON CONDUCTIVE, it'll be fine to put a little dot on each dimm. if the area is too far recessed, its normal to place a thermal pad on top of the dimm instead (that's what ill probably end up doing). that's quite common in overlocking memory sticks or GPU memory chips that sit under heatsinks

 

[jay-m]

Okay, I got my M1 Air today and did some modifications and benchmarking. 16GB / 1TB / 8core GPU on macOS 11.0.1

I only added thermal pads (2mm thick, 6W/mK) to higher part of the radiator - there is a factory installed pad touching lower part (is is attached to bottom cover) and I did not want to do any permanent modifications yet. For the same reason I did not remove black plastic glued to bottom cover too. For temperature measurements I attached simple thermocouple right over where CPU is. Ambient temperature was 18°C, laptop was placed on a wooden table. Did over 30 minutes of Cinebench R23 before modification and over 30 min after. I noted down temperature and benchmark score after every run (about every 2 minutes), here are the results:

It started to throttle a bit after 10 minutes, but very slightly, score dropped from initial 7784 to 7530 after 30 minutes. For comparison without thermal pads score dropped from 7751 to 6879.

But temperature-wise it was bad. Without mod, bottom reached comfortable 37°C and it felt like my Intel MBP while watching full HD Youtube video. After mod final temp was 46.2°C and it was really burning, I would definitely not want to have that on my lap. But I run such heavy and long workloads only while I'm at my desk, so it will be no problem for me. It also cooled very quickly on its own and was under 30°C in 60 seconds.

 

[downshiftdre]

Okay, I got my M1 Air today and did some modifications and benchmarking. 16GB / 1TB / 8core GPU on macOS 11.0.1

I only added thermal pads (2mm thick, 6W/mK) to higher part of the radiator - there is a factory installed pad touching lower part (is is attached to bottom cover) and I did not want to do any permanent modifications yet. For the same reason I did not remove black plastic glued to bottom cover too. For temperature measurements I attached simple thermocouple right over where CPU is. Ambient temperature was 18°C, laptop was placed on a wooden table. Did over 30 minutes of Cinebench R23 before modification and over 30 min after. I noted down temperature and benchmark score after every run (about every 2 minutes), here are the results:

View attachment 1688837
View attachment 1688840

It started to throttle a bit after 10 minutes, but very slightly, score dropped from initial 7784 to 7530 after 30 minutes. For comparison without thermal pads score dropped from 7751 to 6879.

But temperature-wise it was bad. Without mod, bottom reached comfortable 37°C and it felt like my Intel MBP while watching full HD Youtube video. After mod final temp was 46.2°C and it was really burning, I would definitely not want to have that on my lap. But I run such heavy and long workloads only while I'm at my desk, so it will be no problem for me. It also cooled very quickly on its own and was under 30°C in 60 seconds.

Very nice results in performance. Thank you for sharing. I am with you that when it is on my laptop I will not be doing CPU intensive tasks anyways. Those are usually when its on the desk. So the heat wont bother me.

Can you elaborate more on the thermal pad that is already there on the lower part of radiator? Pics?

 

[jay-m]

Can you elaborate more on the thermal pad that is already there on the lower part of radiator? Pics?

I mean this, it is thermally conductive foam glued to bottom of the case that touches lower part of radiator, it is 3mm thick. Was your computer missing it?

 

[downshiftdre]

I mean this, it is thermally conductive foam glued to bottom of the case that touches lower part of radiator, it is 3mm thick. Was your computer missing it.

Interesting. No did not have that. Does it seem dense enough to transfer heat

 

[Nitrogen]

What about battery temperature during sustained workload?

 

[jay-m]

Does it seem dense enough to transfer heat

Yes, it is silicone thermal sponge, I've seen similar but smaller ones in many smartphones and USB chargers. It does not transfer heat as well as dense pad but it is super soft and will not put any pressure on cooled parts. I wonder if Apple puts it only into selected models or started adding it at a later date (my MBA was manufactured on November 23rd).

What about battery temperature during sustained workload?

43.3°C (110°F) after 35 minutes of 100% CPU workload according to coconutBattery. I did not check battery temperature before adding pad, but from your post on previous page it sounds like unmodded MBA's reach the same temperature. This is not terrible, I've seen such battery temperatures under heavy load on my Late 2013 MBP and it still has 75% battery capacity left after 7 years of heavy, daily use (I did only 480 battery cycles on it though).

 

[melgross]

All of this is really interesting but silly. Of course, you’ve voided your warrantee. Yes, Apple will be able to tell.

if you really do work where throttling is something you don’t want, then you should have spent the extra $300 and gotten the Macbook Pro as I did. Then this wouldn’t be something you would have to do. Not that you had to do it anyway.

 

[rumorguy]

sounds like a good way to void your warranty.

 

[landonh12]

All of this is really interesting but silly. Of course, you’ve voided your warrantee. Yes, Apple will be able to tell.

if you really do work where throttling is something you don’t want, then you should have spent the extra $300 and gotten the Macbook Pro as I did. Then this wouldn’t be something you would have to do. Not that you had to do it anyway.

IMO, this mod makes the M1 MBA miles better than the M1 MBP. No fan noise (ever), no Touch Bar, and better ergonomics with the wedge shape.

Not to mention you're saving $300. For some people, $300 is a lot. And how could Apple tell that you did this? If it leaves no residue, then I really don't see a way. I have gotten in-warranty repairs several times with my Macs, and I've opened every single one of them before getting repairs. Just take the pads out before getting a repair.

 

[melgross]

IMO, this mod makes the M1 MBA miles better than the M1 MBP. No fan noise (ever), no Touch Bar, and better ergonomics with the wedge shape.

Not to mention you're saving $300. For some people, $300 is a lot. And how could Apple tell that you did this? If it leaves no residue, then I really don't see a way. I have gotten in-warranty repairs several times with my Macs, and I've opened every single one of them before getting repairs. Just take the pads out before getting a repair.

You clearly know little about this. Even with the fan on you can only hear the Macbook Pro if you listen for it in a quiet room. I’ve never found the Touch Bar to be the negative some do. It has advantages. The wedge shape is small. Is it better, maybe a little, but not so much that people would but the computer because of it.

‘’if you’re doing WORK, you can put this in as a capital expense. You will get a good amount off your taxes because of that, and you’re also allowed to claim depreciation. Keep it for three years and sell it, and you’ll be even.

just opening a device is one thing. Reversing this so that Apple can’t tell, particularly if it results in a heat related repair, is almost. Impossible. I’ve seen times when people had their warrantee revoked when they brought their machines in because of some unauthorized, supposedly reversed mods.

 

[downshiftdre]

if you really do work where throttling is something you don’t want, then you should have spent the extra $300 and gotten the Macbook Pro as I did. Then this wouldn’t be something you would have to do. Not that you had to do it anyway.

Thank you for your opinion. For my use case, this was a no brainer especially how simple and low risk it is. I love the thin and light form factor of the MBA and the fanless and ventless build ensures my internals wont ever have foreign material in it. I now have the performance potential of the M1 MBP and I saved 300 and bought an Apple Watch with it. Plus my engineering brain gained some satisfaction in the whole process. All wins to me.

Not silly one bit.

 

[SesameWasher]

Ok 30 minute cinebench results are in:

In summary, no throttling.

30 min Cinebench R23 score: 7734!!! (in line with the 7713 10 minute score).

Before temps
CPU cores: 22C
Battery: 22C
Battery Proximity: 22C

During temps (halfway 15min mark)
CPU cores: 82-90C
Battery: 35C
Battery Proximity: 28C

End temps (1min remaining)
CPU cores: 82-90C
Battery 35C
Battery Proximity: 28C

Battery drained from 100% to 80% within the 30 minutes.

My conclusion is that the battery is not impacted by the heat in the backshell as shown by the “battery proximity” temperature only rising 6C delta from 22C to 28C. The actual battery temp rose from 22C to 35C likely due to the massive and rapid battery drain occuring during the long and high powered cinebench test.

According to anandtech, running all CPU cores at full speed results in 20~24W while some reviews I've read states it only runs at ~15W in Cinebench R23. I guess that means 90C at 15w, maybe near 20w @ 100C.

I hope that 11.1 will be more optimized for the new chipset.

 

[Jack Neill]

I'm not an expert in these things, but logic tells me that if the thermal pad is enough to make the bottom of the casing hot, then the thermal paste must already be doing a good job. So I'm not sure it's worth removing the heatsink to check and replace the paste.

I have a 2020 i5 Air, and I read the main thread on this topic but so far have done nothing. It's winter here at the moment, so the only time I notice the fans come on is when upgrading Big Sur betas. However, a few months ago in the main heat of summer, watching videos would also do it, so I'll probably install the thermal pad before it gets hot here again. I can't be bothered to do the shim/paste. Lazy.

I did the pad on my i5 and it was af and really made a difference. I dropped 10-20 degrees. I am getting 35C while typing this post. I used to idle at 50-60. Big Sur made it jump up again, but after I clean installed 20B50 I am getting normal temps.

 

[melgross]

Thank you for your opinion. For my use case, this was a no brainer especially how simple and low risk it is. I love the thin and light form factor of the MBA and the fanless and ventless build ensures my internals wont ever have foreign material in it. I now have the performance potential of the M1 MBP and I saved 300 and bought an Apple Watch with it. Plus my engineering brain gained some satisfaction in the whole process. All wins to me.

Not silly one bit.

The problem is that you don’t know if it’s low risk. It could be very high risk, and that’s a oroblem. You want to assume it’s low risk because you want to do it. Not everything we want to be true, is.

 

[melgross]

According to anandtech, running all CPU cores at full speed results in 20~24W while some reviews I've read states it only runs at ~15W in Cinebench R23. I guess that means 90C at 15w, maybe near 20w @ 100C.

I hope that 11.1 will be more optimized for the new chipset.

Temperatures in most testing ive seem were lower, often by 20 degrees.

 

[downshiftdre]

So

The problem is that you don’t know if it’s low risk. It could be very high risk, and that’s a oroblem. You want to assume it’s low risk because you want to do it. Not everything we want to be true, is.

You are absolutely right. But to me it is just a thermal pad designed for this purpose. I will report back every now and then whether positive or negative in the name of transparency to the forums.

 

[SesameWasher]

Temperatures in most testing ive seem were lower, often by 20 degrees.

That’s a good sign. Will OP @downshiftdre also do a test after the update?

 

[scatopie]

All of this is really interesting but silly. Of course, you’ve voided your warrantee. Yes, Apple will be able to tell.

if you really do work where throttling is something you don’t want, then you should have spent the extra $300 and gotten the Macbook Pro as I did. Then this wouldn’t be something you would have to do. Not that you had to do it anyway.

The problem is that you don’t know if it’s low risk. It could be very high risk, and that’s a oroblem. You want to assume it’s low risk because you want to do it. Not everything we want to be true, is.

True, but you also don't know that it's high risk! Although we can't assume that it is low risk, we have reason to hypothesize that it is low risk, because we are not physically altering any electrically connected parts. We are simply creating better heat connectivity to dissipating heat. I could see a fear to say that the M1 was not designed to be "pushed this far," but I believe there are similar temperature outputs from the M1 Pro (also around 46 degrees) when pushed to the limit. So since they are exactly the same processors, there really shouldn't be reason to assume we have increased risk in any way. With the risk argument aside, what we DO know is that we are getting immediate benefits. I saw that you put forth argument for the Pro, but points like Touch Bar is moot, so it depends on the preference of the user.

One thing I'll concede is that the Pro has a slightly better battery. But with the Air, for the same price as the $1299 Pro ($50 cheaper), you can get double the SSD on the 8 core. Again, these are very tangible benefits, where risk does seem to be limited and reduced (not heightened) compared to other kinds of modifications.

But yes, there is SOME risk to doing this. And the people who own the M1 machines are the perfect ones to take it, because after all, this is the first generation M1, new tech, limited software compatibility. If we can get an extra 20% performance under these circumstances, I think it's a great opportunity to take.

 

[mwidjaya]

have reason to hypothesize that it is low risk, because we are not physically altering any electrically connected parts.

Makes sense, after all temp sensor(s) still function and will kick in throttling same as before, just later.

 

[SesameWasher]

Still, anyone tested their score after 11.1 update?