2020 Air Heatsink Modification Thread

[Loog]

@Robotronic The consequence of limiting the airflow over the batteries didn't cross my mind, but below is a screenshot of the temps from Macs Fan Control just before I stopped the wind channel stress test. Do you see anything out of the ordinary?

View attachment 910352
@Loog The rest of my Amazon care package should be coming this afternoon, so hopefully I can take some measurements tonight. The weird thing is that I don't own a proper ruler 🤪 just a paper one that came with my Moleskine notebook haha.

Thanks that would be really helpful. Looking on Intel website the dimensions are quoted as 22mm x 16.5mm for the Icelake and 20.5mm x 16.5mm for the Coffee lake. With the additional space in the heatsink of the Coffee lake I suspect that the same size shim could be used just with a different thickness for ease of 22mm x 16.5mm as there is space to house this. IIRC mine was 18.5 mm x 22mm which just fitted nicely but overlapped the CPU

What I did notice on my device is there is a shroud around the CPU which the heatsink makes contact and the shim fills the void between the CPU and the heatsink along with compound so it also needs to be considered.

 

[edubfromktown]

MacBook Air: i7/16GB/1TB

TL;DR: Sealing up the wind channel appears to only offer limited performance improvements.

I only got my iFixit toolset yesterday night, so I couldn't do the heat sink mod. I decided to open my MacBook Air anyway to see what I'm dealing with. The first thing I noticed was that the heat sink on my i7 looks quite different from the i3. It looks like corrugated cardboard but in metal form. I think the centre foam pad that is stuck to the inside of the bottom case is there to press against the heat sink and force all the air to go through the heat sink.

View attachment 910232 View attachment 910233

View attachment 910235 View attachment 910236


Wind Channel Mod
Based on @Jordi Padreny's experiments with the wind channel, I sealed the area between the fan and heat sink to see the effect on performance. The idea is to restrict the fan to draw all its air from behind the heat sink, thus maximising airflow through the heat sink for hopefully higher cooling effect. I cut strips of corrugated cardboard from my Amazon box and lined them against the metal separators next to the batteries (indicated by red boxes below). I found that performance was worse when the cardboard strips were taller than the metal separators, compared to them being the same height.

View attachment 910240 View attachment 910241

View attachment 910242


Results
I used Intel Power Gadget (v3.7.0) to monitor the temperature and CPU/GPU frequencies, and used its built-in tests to load the CPU (All Thread Frequency) and GPU (Maximum Frequency). I let each test run until the frequency stabilised, and then maxed out the fans to test the full wind channel effect. I then let the frequency stabilise again (usually goes up). My battery was charged to 100% and plugged in, and I set my screen to maximum brightness and always on.

I ran the tests before and after the mod, and back-to-back to minimise differences in ambient temperature. What I saw was that the wind channel mod had a very slight effect on CPU (+70 MHz at best) and GPU (+30 MHz) performance (Forgot to take a screenshot of the original GPU performance). Would be interesting to see if the benefit improves with the heat sink/shim mod.

View attachment 910258 View attachment 910259 View attachment 910261

Wow- great info. Thank you for posting the photos and all of the details. I am one of the Core i7 Air 2020 users who has not observed any excessive heat issues whatsoever (so I haven't cracked mine open or considered modding it). I'm on day 9 of thrashing it with tests, macOS reinstalls (just completed the 4th iteration), photo processing performance and general use.

It might be a stretch to assume that the heat sink on the i3 and i5 are different than the i7.

Might a few others who have popped the covers on i3 and i5 models be able to confirm this theory?

 

[edubfromktown]

I agree, this unit is built for a specific market segment and engineered to be 'fit for purpose' with thermal limitations to hold back long sustained workloads. By virtue the device does seem to run a little hotter than the previous model, use office, browse the web with safari and do light CPU intensive work and boom all is well. There are some poorly tuned apps that make the overall experience poor but these are broadly outside of the Apple eco systems so someone else problem to resolve...

Yes indeed... there are always tradeoffs between cost, design and performance. I'm over the hill and getting a bit lazy so a 10-15% theoretical (in some regard at least) benchmark performance hit doesn't bother me at all. This was part of the reason I decided to go for the i7 Air 2020 and see what happened. The "gamble" paid off in that I've had no excessive heat issues after more than a week of thrashing it with a battery of tests, 4 macOS reinstalls and general daily usage.

Some applications are definitely more resource hungry. Chrome has always been among the worst offenders in my experience so I've outright avoided it and use Opera as primary and SRWare Iron as secondary web browsers. Poorly tuned and/or developed applications appear to be growing in number. Many with "new and improved" cloud-based components and dreaded monthly $ub$cription fees are among them.

I am grateful that Catalina still plays nice with a few "perpetual" licensed products that I use on a frequent basis (Office 2016 and Lightroom 6 primarily). Adobe development and patch management (or lack thereof!) has stymied me for a long time.

For many years one could easily crush VDI infrastructure performance globally by holding down a mouse button for an extended period of time in Adobe Reader from a minimal access user VDI session! In the case of Lightroom 6, they shipped a 64-bit app with a 32-bit installer. So there is no longer an easy way to install/reinstall LR6 in Catalina. Adobe updates are an excellent way to experience a denial of service attack firsthand against hundreds if not more users at the office (unless you lock down how they obtain them).

 

[srkirt]

The batteries do not undergo any heating at all ... I have already looked at leaving sensors and transistors in the air uncovered.
Here is a photo of Cinebench before and after.

 

[srkirt]

I would take a 2020 MBA with my invention to get the power of a Pro ... at least very close ...

[automerge]1588187421[/automerge]
Increasing the power of an i7 with the wind tunnel. From now on, I will only publish the results of my 2019 MBA.

 

[Loog]

[automerge]1588187421[/automerge]
Increasing the power of an i7 with the wind tunnel. From now on, I will only publish the results of my 2019 MBA.

Good research Jordie ... I think we're coming to the max of the MBA Retina 2018/19 unit unless you can unleash some hidden power from within !

These are great machines, so glad I kept mine.

 

[EValentino]

Agreed, I think the CPU’s are locked to a maximum clockspeed.

 

[RiaKoobcam]

Good research Jordie ... I think we're coming to the max of the MBA Retina 2018/19 unit unless you can unleash some hidden power from within !

These are great machines, so glad I kept mine.

Loog, question from someone who owns the 2019 MBA - what kind of performance difference are you seeing after the copper mod? My cooling is already fine and it's certainly fit for what I use it for, but are you able to quantify it in some way?

Cheers

 

[kinchee87]

MacBook Air 2020: i7/16GB/1TB

Just finished my heat sink (+ wind channel) mod and preliminary results look good! It's already 1:30 am here, so I will post pictures later.

I did the same stress test using Intel Power Gadget as before at 100% battery and plugged in, and ran it for about 15 min. The first thing I noticed was that the CPU temperature was much less sensitive to changes in CPU load, which might mean that the fans switch between on/off less often. In the screenshot below, the temperature rises gradually, whereas before it would shoot up much steeper.

I'm getting about 2.70 GHz for the CPU (~400 MHz improvement over no mod) and no noticeable thermal throttling for the GPU (it basically hits the requested frequency). The temperatures around the case looks okay (?)

CPU result:

GPU results:

 

[kreasonos]

Wow- great info. Thank you for posting the photos and all of the details. I am one of the Core i7 Air 2020 users who has not observed any excessive heat issues whatsoever (so I haven't cracked mine open or considered modding it). I'm on day 9 of thrashing it with tests, macOS reinstalls (just completed the 4th iteration), photo processing performance and general use.

It might be a stretch to assume that the heat sink on the i3 and i5 are different than the i7.

Might a few others who have popped the covers on i3 and i5 models be able to confirm this theory?

Why have you already re-installed Mac OS 4x? Just curious.

 

[russell_314]

Why have you already re-installed Mac OS 4x? Just curious.

By him saying he was thrashing it with tests I'd assume just to see if it can do it. Some people buy cars to drive and others buy them to take to the shop to see the dyno HP numbers LOL

 

[Robotronic]

MacBook Air 2020: i7/16GB/1TB

Just finished my heat sink (+ wind channel) mod and preliminary results look good! It's already 1:30 am here, so I will post pictures later.

I did the same stress test using Intel Power Gadget as before at 100% battery and plugged in, and ran it for about 15 min. The first thing I noticed was that the CPU temperature was much less sensitive to changes in CPU load, which might mean that the fans switch between on/off less often. In the screenshot below, the temperature rises gradually, whereas before it would shoot up much steeper.

View attachment 910624

I'm getting about 2.70 GHz for the CPU (~400 MHz improvement over no mod) and no noticeable thermal throttling for the GPU (it basically hits the requested frequency). The temperatures around the case looks okay (?)

CPU result:
View attachment 910625 View attachment 910628

GPU results:

View attachment 910626 View attachment 910627

Nice work! Did you happen to run Geekbench with the stock heatsink, and then again after the mod?

 

[Loog]

Loog, question from someone who owns the 2019 MBA - what kind of performance difference are you seeing after the copper mod? My cooling is already fine and it's certainly fit for what I use it for, but are you able to quantify it in some way?

Cheers

To be honest I don't really notice any performance difference day to day, the Cinebench shows around a 15% improvement for sustained and heady workloads. I'm an office type user with the odd VM in use so I don't really push this device, its my go to when on the sofa or sidecar for day to day work and BYOD for business. What I have noticed is that the machine is running a little cooler in general, idle this is circa 40 deg when little is happening, I would say it runs 5-10 deg cooler, when on video calls (for which I'm doing quite a few WFH) CPU temp is around 60-70 deg with a 30% ish load. MS teams used to kill the device with 90+ temp and the fan quite loud, now this has gone. Zoom was never an issue.

The time for the fans to spin up where you can start to hear them is much slower when using VMWare Fusion, in general this just really improves the heat transfer to from the CPU and flattens the spike curve when you apply short and heavy loading, opening multiple apps etc. Recovery time also seems to be slightly quicker as well. See the above post from Kinchee87 on his 2020, the same concept applies here.

I wasn't disappointing with the MBA 2018 thermals for general use until I spent more time on video calls so I pondered over doing anything. Is the mod worth it ? In my opinion yes it does make some subtle changes to the device when you do need to push it hard. The question is are people happy stripping down their circa £1500 device with a possibility of bricking it ... the thought did cross my mind. Taking it slow the process is quite simple for anyone who has built PC in the past. Would the MBA still perform well enough without it, sure it would just be a little slower and warmer.

This modification doesn't turn a MBA into a pro device for sure but what it does do is squeeze a little more out of it to make it a great machine better. Looking at a lot of content of Macbook devices over the years, there has been numerous articles over better CPU paste etc as this does seem to be an area which could be improved. This just adds to this concept by removing the engineering tolerances between the heat sink and CPU which the thermal compound would fill and providing direct contact.

Hope that helps?

 

[RiaKoobcam]

To be honest I don't really notice any performance difference day to day, the Cinebench shows around a 15% improvement for sustained and heady workloads. I'm an office type user with the odd VM in use so I don't really push this device, its my go to when on the sofa or sidecar for day to day work and BYOD for business. What I have noticed is that the machine is running a little cooler in general, idle this is circa 40 deg when little is happening, I would say it runs 5-10 deg cooler, when on video calls (for which I'm doing quite a few WFH) CPU temp is around 60-70 deg with a 30% ish load. MS teams used to kill the device with 90+ temp and the fan quite loud, now this has gone. Zoom was never an issue.

The time for the fans to spin up where you can start to hear them is much slower when using VMWare Fusion, in general this just really improves the heat transfer to from the CPU and flattens the spike curve when you apply short and heavy loading, opening multiple apps etc. Recovery time also seems to be slightly quicker as well. See the above post from Kinchee87 on his 2020, the same concept applies here.

I wasn't disappointing with the MBA 2018 thermals for general use until I spent more time on video calls so I pondered over doing anything. Is the mod worth it ? In my opinion yes it does make some subtle changes to the device when you do need to push it hard. The question is are people happy stripping down their circa £1500 device with a possibility of bricking it ... the thought did cross my mind. Taking it slow the process is quite simple for anyone who has built PC in the past. Would the MBA still perform well enough without it, sure it would just be a little slower and warmer.

This modification doesn't turn a MBA into a pro device for sure but what it does do is squeeze a little more out of it to make it a great machine better. Looking at a lot of content of Macbook devices over the years, there has been numerous articles over better CPU paste etc as this does seem to be an area which could be improved. This just adds to this concept by removing the engineering tolerances between the heat sink and CPU which the thermal compound would fill and providing direct contact.

Hope that helps?

Yeah definitely, might give it a go. Where'd you acquire your copper sheet and what kind of scissors did you use to cut it?

The only other times I've done this I've bought copper shims from eBay long ago that were the right size for the laptop I was doing the mod for, never done a DIY job before.

 

[Loog]

Yeah definitely, might give it a go. Where'd you acquire your copper sheet and what kind of scissors did you use to cut it?

The only other times I've done this I've bought copper shims from eBay long ago that were the right size for the laptop I was doing the mod for, never done a DIY job before.

I got some off cuts from an engineering firm and used tin snips to cut, 0.5mm is a little thick for normal kitchen scissors. I have sent over the dimensions to a good friend of mine who has a precision engineering workshop to make production samples of these in 0.3mm and 0.5mm suitable for the MBA 2018/19 and new MBA 2020. Looking at the dimensions one size should both but in different thicknesses. If you're in no rush I'll be asking for some community feedback on these and will be offering some samples in return for reviews.

 

[edubfromktown]

By him saying he was thrashing it with tests I'd assume just to see if it can do it. Some people buy cars to drive and others buy them to take to the shop to see the dyno HP numbers LOL

LOL

I was wrestling with some external drives not booting as a startup disk (Mike Bombich, author of Carbon Copy Cloner has amassed quite a list), getting Lightroom 6 perpetual to behave (executed fine 1st time and then failed to launch thereafter due to sandbox security issues; finally got it playing nice), evaluating other photo apps and testing 10.15.3 vs 10.15.4 changes.

 

[srkirt]

My tests with the wind tunnel are giving results, from now after getting more than 100pts in Cinebenck and being above an i7 it is not bad. Now I have just tried removing the tunnel and placing shark fin type diffusers on the heatsink, I have managed to increase a little but I am already reaching the limit, later I will try gull wing type diffusers always separating the cover from the heatsink.

photo 1 : reposo
photo 2 : test 1 8.000rpm baterie.
photo 3 : test 2 " " record !!

 

[Dannuts]

Nothing to add other than that this has been a great read!

 

[srkirt]

I can edit 4K video with DaVinci.

 

[edubfromktown]

I can edit 4K video with DaVinci.

Nothing melted - including your fingers? 🙃

 

[srkirt]

[QUOTE = "edubfromktown, publicación: 28420149, miembro: 494942"]
¿Nada derretido, incluidos tus dedos? 🙃:RE
[/ CITAR]

Max93º.

 

[srkirt]

Nothing melted - including your fingers? 🙃

 

[kinchee87]

MacBook Air 2020: i7/16GB/1TB with wind channel and heat sink mod

Here are more benchmark results to supplement my previous ones. I always ran the benchmarks with the MacBook Air fully charged and plugged in, and always waited for the CPU to cool down to around 42 ℃ before running the next benchmark. Each benchmark was executed at least three times and I averaged the results.


Geekbench 5
The single-core and graphics performance are only slightly better. I think it's because thermal throttling didn't happen much for those tests with the original heat sink. Multi-core performance, however, is about 20% higher. Click on a Mod result to view the best that I got.

Test Type	Original (Points)	Mod (Points)	Improvement
CPU (Single-core)	1,214​	1,271​	4.8%​
CPU (Multi-core)	2,992​	3,666​	22.5%​
Compute (OpenCL)	3,825​	3,944​	3.1%​
Compute (Metal)	4,285​	4,491​	4.8%​

Cinebench R20
Single-core and multi-core performance both improved by about the same amount, unlike the large performance improvement gap seen with Geekbench.

Test Type	Original (Points)	Mod (Points)	Improvement
CPU (Multi-core)	979​	1,110​	13.5%​
CPU (Single-core)	353​	390​	10.5%​
MP Ratio	2.77​	2.84​	2.7%​

Intel Power Gadget 3.7.0
Lastly, I did some plain-old stress tests, which show about 20% improvement in sustained CPU and GPU performance.

Test Type	Original (GHz)	Mod (GHz)	Improvement
CPU (All Thread Frequency, Scalar)	2.25​	2.70​	20.0%​
CPU (All Thread Frequency, AVX-256)​	2.12​	2.59​	22.2%​
CPU (All Thread Frequency, AVX-512)	2.00​	2.40​	20.0%​
Intel Graphics Test (Max Frequency)	0.90​	1.10​	22.2%​

I don't remember if my MacBook Air was originally able to sustain max Turbo Boost for single-core.

 

[Loog]

MacBook Air 2020: i7/16GB/1TB with wind channel and heat sink mod

Here are more benchmark results to supplement my previous ones. I always ran the benchmarks with the MacBook Air fully charged and plugged in, and always waited for the CPU to cool down to around 42 ℃ before running the next benchmark. Each benchmark was executed at least three times and averaged the results.


Geekbench 5
The single-core and graphics performance are only slightly better. I think it's because thermal throttling didn't happen much with the original heat sink. Multi-core performance, however, is about 20% higher. Click on a Mod result to view the best that I got.

Test Type	Original (Points)	Mod (Points)	Improvement
CPU (Single-core)	1,214​	1,271​	4.8%​
CPU (Multi-core)	2,992​	3,666​	22.5%​
Compute (OpenCL)	3,825​	3,944​	3.1%​
Compute (Metal)	4,285​	4,491​	4.8%​

Cinebench R20
Single-core performance improved more than multi-core performance. However, the multi-core improvement is less pronounced compared to the Geekbench results.

Test Type	Original (Points)	Mod (Points)	Improvement
CPU (Single-core)	979​	1,110​	13.5%​
CPU (Multi-core)	353​	390​	10.5%​
MP Ratio	2.77​	2.84​	2.7%​

View attachment 910875


Intel Power Gadget 3.7.0
Lastly, I did some plain-old stress tests, which show about 20% improvement in sustained CPU and GPU performance.

Test Type	Original (GHz)	Mod (GHz)	Improvement
CPU (All Thread Frequency, Scalar)	2.25​	2.70​	20.0%​
CPU (All Thread Frequency, AVX-256)​	2.12​	2.59​	22.2%​
CPU (All Thread Frequency, AVX-512)	2.00​	2.40​	20.0%​
Intel Graphics Test (Max Frequency)	0.90​	1.10​	22.2%​

View attachment 910904 View attachment 910905 View attachment 910906 View attachment 910907

I don't remember if my MacBook Air was originally able to sustain max Turbo Boost for single-core.
View attachment 910927

Great feedback and some great performance improvements !

 

[kinchee87]

Great feedback and some great performance improvements !

This has been an awesome engineering journey that wouldn't have happened if @Jordi Padreny didn't take the heat sink off his MacBook Air. Those who subsequently did the heat sink mod also gave me some confidence to try it out too

I'm going to edit my photos of the heat sink now, so hopefully you'll get some rough measurements soon!