Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

excelsior.ink

macrumors regular
Apr 15, 2020
134
78
My Geekbench Multicore CPU results were:
3169 (shim mod)
3443 (shim + heatpad mod)
3595 (shim + heatpipe mod)
You might want to redo the shim ( I did it 5-6 times till I got it properly and I might do it again). The shim is the most important mod. I also have the i5 and I get consistently 3400-3600. This one I did it after reading your post, with Safari and Edge and Slack and Mail running and no cooling down before the test:
Screenshot 2020-06-08 at 20.41.05.png

3169 is too low, I think you can get better results with just the shim.
 

demize

macrumors newbie
Jun 17, 2012
15
27
Tampa, FL
Ultimately for me the biggest benefit of the mod (i5/8/512, .3mm shims, Thermal Grizzly Kryonaut, Minus 8 Pad 30x20x2) is the the silent operation. For the last couple of years most of my 'general home computing' has been done on an iPad Pro; as such I've gotten quite used to a fan-less design. This has been increasingly important over the last couple months throughout this 'work from home' period as most of the day I'm dealing with the fans on my work laptop or desktop PC.

Pre-mod my MacBook Air fans would spin up only occasionally with general web browsing, email, Spotify etc. This changed when video conferencing. FaceTime, Zoom and Skype all caused the fans to come on very quickly and very loudly. Connecting my MBA to an external display caused the fans to run continuously no matter what the task I was performing.

Post-mod this thing is silent, even while running benchmarks the temps never get higher than 92 degrees. I have not tried a Zoom call while connected to an external monitor (I suspect this will cause the fans to kick on) but during an hour long FaceTime call fans never came on.

As for performance, I'll be honest, I didn't buy the MBA for intensive tasks. I have a desktop PC for gaming and, seemingly unlike the rest of the world, I have no desire to edit photos or videos. For everyday computing needs the 2020 MBA is extremely fast, and now also silent. Yes, the improvements in the benchmarks are nice... so much so that I even ordered a 100x100 Minus 8 pad to cut to the entire size of the heat sink to 'maximize the performance'... but I stopped short of installing it since I'm so pleased by how the machine performs currently.

As I mentioned before, if it wasn't for the 'worse than my 2015 MBA/2012 MBP' webcam, this would be the absolute perfect computer (for me).
 
Last edited:

Bones13

macrumors regular
Oct 7, 2008
144
62
@srkirt - I’m not nearly as brave as you are senor!

@demize - thanks for comments, those are exactly the parts, and mods I’m planning on the same MBA build. My MBA is in shipping pipeline now, it left Taiwan this morning. Your description of MBA function mirrors mine, quiet and powerful (enough). Some fan is good though, cooler CPU will live longer.

I do wonder if a third party will create a heat sink - heat pipe - fan replacement unit.
 
  • Like
Reactions: srkirt and Spectrum

kazune_karin

macrumors member
May 9, 2020
38
50
You might want to redo the shim ( I did it 5-6 times till I got it properly and I might do it again). The shim is the most important mod. I also have the i5 and I get consistently 3400-3600. This one I did it after reading your post, with Safari and Edge and Slack and Mail running and no cooling down before the test:
View attachment 922585
3169 is too low, I think you can get better results with just the shim.

Thanks. What paste did you used? I only have Arctic MX-4 on my hands.
I repasted again and with shim + heat pipe mod (I am too lazy to remove the heat pipe),
my score rises to 3713. I did it 3 times, and its all above 3700. Wow.
スクリーンショット 2020-06-09 7.47.22.png


And more importantly, the frequency profile is perfect for me. The pink (requested frequency) line completely overlapped with the blue (average frequency) line THROUGH the test. That means the CPU provided what was needed for the test without thermal restricted. And the temperature never reached 100 degree.
New-GB5-CPU.png
 

RiaKoobcam

macrumors regular
Apr 17, 2020
225
289
Thanks. What paste did you used? I only have Arctic MX-4 on my hands.
I repasted again and with shim + heat pipe mod (I am too lazy to remove the heat pipe),
my score rises to 3713. I did it 3 times, and its all above 3700. Wow.
View attachment 922650

And more importantly, the frequency profile is perfect for me. The pink (requested frequency) line completely overlapped with the blue (average frequency) line THROUGH the test. That means the CPU provided what was needed for the test without thermal restricted. And the temperature never reached 100 degree.
View attachment 922651

Any insights into what you changed with your repaste/what method you used?
 

excelsior.ink

macrumors regular
Apr 15, 2020
134
78
Thanks. What paste did you used? I only have Arctic MX-4 on my hands.
I repasted again and with shim + heat pipe mod (I am too lazy to remove the heat pipe),
my score rises to 3713. I did it 3 times, and its all above 3700. Wow.
I used Thermal Grizzly Kyronaut and Noctua NT-H2. The Kryonaut seems to provide better results but no major difference. I actually think the method, how to and how much to apply, is more important.
About your results... Wow, indeed! I wonder if anyone can score 4000 with the i5-1030NG7
 
  • Like
Reactions: kazune_karin

IngerMan

macrumors 68020
Feb 21, 2011
2,016
905
Michigan
My best single core is 1304 and multi core 4007. With thermal pad only. Searching GB the best i7 score is 4200’s And very few of them break 4000, seen just a few. Mine (KTF) and Daystar are in there. @daystar has a high of 4120 with both mods.

no way will the i5 break 4000, read and weep boys. The i7 is the only 4000 club. 😂
 

ctjack

macrumors 68000
Mar 8, 2020
1,529
1,547
Impromptu test, full office suit open and working, safari, mail... and timemachine running. Lets add the intel power gadget to max and see what happens to my MBA 2018...
I have noticed you are running also MacsFanControl widget on the top, especially with the stats of fan rpm and C degrees. That widget alone is making my machine crazy while idling, like it was doing something heavy (+15C to temps and +20 to CPU use).
 

kazune_karin

macrumors member
May 9, 2020
38
50
Any insights into what you changed with your repaste/what method you used?

I think it is the amount of paste. I tried to keep it minimal but maybe it is too little. When I opended the heatsink to repaste, the shim did not stick to the heatsink. Maybe that's why. I increased a bit to ensure even coverage.

I used Thermal Grizzly Kyronaut and Noctua NT-H2. The Kryonaut seems to provide better results but no major difference. I actually think the method, how to and how much to apply, is more important.
About your results... Wow, indeed! I wonder if anyone can score 4000 with the i5-1030NG7

I think 4000 will be difficult. The barrier for my i5 seems to be the frequency 1.8Ghz cap. As your see in my Geekbench graph, the latter part stuck at 1.8Ghz eventhough temperature is below 100 degree. That's even more true for Cinebench R20. Air i7 can be sustained over 2Ghz (Nice if I can find a screenshot for the profile). Unless I can 'overvolt' to the settings like i7-1060NG7, I don't think my score will improve dramatically anymore.
[automerge]1591661166[/automerge]
My best single core is 1304 and multi core 4007. With thermal pad only. Searching GB the best i7 score is 4200’s And very few of them break 4000, seen just a few. Mine (KTF) and Daystar are in there. @daystar has a high of 4120 with both mods.

no way will the i5 break 4000, read and weep boys. The i7 is the only 4000 club. 😂

Do you have any frequency/thermal profile for Geekbench to share?
 
  • Like
Reactions: RiaKoobcam

ctjack

macrumors 68000
Mar 8, 2020
1,529
1,547
I think 4000 will be difficult.
I think as if you guys modding the CPU then you should compare Cinebench R20 results. Because your graph is not always maximizing CPU and also it heavily relies on the RAM size, so someone with bigger RAM will get bigger score no matter what.
I really liked your full Intel screenshot, could you take the same one with Cinebench R20?
I do have one opinion, that Apple pegged performance to the CPU temperature. Like Max Tech was having sometimes worse results with maximum Ghz achieved/sustained when water cooled.
It seems like Air is getting more power when temperatures are rising. So when he water cooled it to 55C degrees, his chart was looking like a throttling.
 

IngerMan

macrumors 68020
Feb 21, 2011
2,016
905
Michigan
Thanks. The result is great. Is it shim+heatpad mod? Is fan speed profile system default?
My heat dissipation may still need improvement.

I never hear the fans on GB or Cinebench R20, I score over 1200 on CB. I did check the fans on CB and they where running 3300 but I could not hear them. I don't think they run on Geek Bench.

Edit: The fans do run on GB test, I just checked, about 3,500 but not audible.

It is a 1.5mm Artic pad only, not the top of the line but seems to do the job. I purchased the 140x140 on Amazon.

No other mods. I cut a piece 35mm x 45mm and applied it below.



MBA Pad outline.png
 
Last edited:

kazune_karin

macrumors member
May 9, 2020
38
50
I never hear the fans on GB or Cinebench R20, I score over 1200 on CB. I did check the fans on CB and they where running 3300 but I could not hear them.

It is a 1.5mm Artic pad only, not the top of the line but seems to do the job. I purchased the 140x140 on Amazon.

No other mods. I cut a piece 35mm x 45mm and applied it below.



View attachment 922676

Thank you. Heatpad is tempting. Um, I will try to improve my mod further.
 

RiaKoobcam

macrumors regular
Apr 17, 2020
225
289
Just finished the copper shim/repaste on my 2019 - good fun!

Good news - went well!
Bad news - the video I took was completely out of focus so it's unusable!

I ended up just using one 20x20 0.3mm copper shim, as that fits over the 2018/2019 CPU die and GPU nicely. Didn't have to use the half-shim I'd made as it'd be pretty marginal in improving performance. I would recommend going for a 0.5mm thickness shim if you've got a 2018/2019 MBA though, I had to use a little more MX-4 than I'd normally use to get good contact.

As far as whether it's T3/T4 for the heatsink - I think it is T3, but T4 fits. I'd recommend having both on hand to see what feels right, they're pretty cheap to buy as a set.

Even though I only did this to improve internal component lifespan, there are performance improvements in benchmarks (which don't really mean anything to me, but here you go):

Pre-mod:
Geekbench 5: Single Core 700, Multi Core 1516
Cinebench R20: 618

Post-mod:
Geekbench 5: Single Core 868, Multi Core 1754
Cinebench R20: 701

Thermals (audible emissions, internal temperatures and external temperatures) were not an issue before the mod, and certainly aren't an issue after. Web browsing/video temps are still 40-50c.

@ilikewhey - maybe something to add to page 1:
________________________________________________________
Getting the factory applied Thick Black Goo off was way different than any thermal compound removal I've done before.

After trying a bunch of stuff (isopropyl solution, alcohol wipes), the only thing that really worked was using gloved fingers to peel back as much as possible. After that was done, I used a toothpick to clean in/around the the CPU/GPU, and then finished it off with by polishing with some isopropyl solution, using a coffee filter so as not to leave fibres/residue.

The heatsink was similarly hard to clean - physically peeling it off with gloved fingers and then using something abrasive (fine grit sandpaper is best, you can use a kitchen scrub bud in a pinch) is the only way to get rid of the factory whatever-the-hell-they-used-black-stuff, and ready it for your new thermal paste.

It took much longer to actually get rid of this stuff fully on the CPU/GPU and the heatsink than I was expecting (and I was expecting it'd be kinda tricky), so the important thing is to just take it slow in your fight with the evil goo.
________________________________________________________

Whilst for 2018/2019 owners this mod definitely isn't necessary, it's super easy to do. I'm 30 now and when I got my first Macbook Air at the tender age of 23 (with my hard earned student job money), I had to do a battery replace. It was like defusing a bomb, the thing was so complicated. Sweaty palms and shaky hands, having to delicately prise up tiny ribbon cables and fight my way through adhesive.

As a contrast, this is by far the easiest, most accessible MBA I've cracked open. If you're even slightly curious about whether you can get better thermals or performance with any MBA since 2018, it'll cost you about $30 (shims, paste, P5 and T3/T4 screwdrivers) and take about half an hour of your time.
 
Last edited:

hmhmmhmmm?

macrumors newbie
Apr 23, 2020
28
42
I found hmhmmhmmm? shim only mod on i5 Air with full fan blast and 3852.
I may still have some room for improvement?

The other day i ran a geekbench after i got home from a sunny day ~25 outside(not sure what the room temperature was, but it’s def. cooler).
Started the i5/16/256(shim) let it idle for a bit (maybe 5 min) and then ran geekbench. Best yet!

Waiting for pad to see if it will improve anything.

8EDABC40-8876-42DC-8692-1E1E6E92FE49.jpeg
 

Loog

macrumors regular
Apr 14, 2020
164
167
I have noticed you are running also MacsFanControl widget on the top, especially with the stats of fan rpm and C degrees. That widget alone is making my machine crazy while idling, like it was doing something heavy (+15C to temps and +20 to CPU use).

@ctjack MacsFanControl plays nice with my system, usually taking around 0.3 which was similar to Fanney.

I would recommend going for a 0.5mm thickness shim if you've got a 2018/2019 MBA though, I had to use a little more MX-4 than I'd normally use to get good contact.

Just finished the copper shim/repaste on my 2019 - good fun!

Pre-mod:
Geekbench 5: Single Core 700, Multi Core 1516
Cinebench R20: 618

Post-mod:
Geekbench 5: Single Core 868, Multi Core 1754
Cinebench R20: 701
@RiaKoobcam, thanks for the review, seems that the scores you got were in the error or margin as mine with the 0.5mm shim, I didn't run GB only Cinebench with a score of 704 so there is no fundamental difference between the shim thickness other than the 0.5mm you only need to use a very small amount of thermal compound. The tolerances on the Gen 1 retina MacBooks seems to be little more than the Gen 2.
 
  • Like
Reactions: RiaKoobcam

Loog

macrumors regular
Apr 14, 2020
164
167
Freebee Tuesday.

Firstly, this thread is awesome, there has been so much community input and support by everyone sharing their experiences and homing different mods.

For those who have been following this thread throughout will know I had a sample run of custom shims made through an aerospace engineering firm that were water cut to exact size that drop straight into all generations of MBA retina with nothing more than a quick rub down on emery cloth. These will never go into full production by me due to cost, but the project was fun and proved an assumption and kept me busy during lockdown. This was a subjective move by me, warmly accepted by most but challenged by others as profiteering. Whatever side of the fence you were on, it was always my intention for this to be community driven.

I’m now left with 6 x 0.5mm shims 22mm x 16.5mm which are of no use to me I only have 1 device (note to you who buy pack of 10 and have spares). I’m so grateful for all those that have purchased these, supported this idea and provided feedback to this thread. Financially I’m sitting around £50 short of breaking even of this project. Reflecting on this... was the modification worth this cost to me, yeah for sure.

So … I’m offering to post world-wide at my expense my last remaining 6 custom shims (1 per person) to anyone who is willing to undertake the mod, in return they post a detailed review with pictures or even better a video to this thread with benchmarks etc.

If you want a free shim delivered to your door reply openly on the forum so we know who you are and to expect a review in the next week or so, tag me in the post so I get a notification.

I would be grateful if you can review the numbers who have already requested and if <6 requests, then state how many are shims are left your comment to help others, then drop me a PM with your address and I’ll get these out in the post before the weekend. Open forum posts will top the speed of PM request.

Please only request one of these if you are willing to participate and provide quality feedback for others to follow.

Happy modding people!
 
Last edited:

srkirt

Suspended
Apr 12, 2020
257
179
Barcelona
Well this from the heatsink to the fan is not a panacea ... I have got even more power and less temperature but at the same time I have occupied the wind tunnel ...
I have already broken two threads of two screws ... when I press the cover it presses the motherboard and the keyboard keys sound ... that is if I close the laptop and I press the case it turns on by itself ... try to disassemble the motherboard to see what is behind if a screw had been inserted, but it was late and I gave up. In the end I think that the best thing is a good copper sink direct to the cpu as close as possible to the cover because if it bends where the processor goes, the errors begin ...

2º Test...
 

Attachments

  • Captura de pantalla 2020-06-09 a las 13.25.27.png
    Captura de pantalla 2020-06-09 a las 13.25.27.png
    316.1 KB · Views: 240
  • IMG_0147.jpg
    IMG_0147.jpg
    531.6 KB · Views: 181
  • IMG_0149.jpg
    IMG_0149.jpg
    484.3 KB · Views: 192
Last edited:

silvia_18

macrumors member
May 24, 2020
47
45
Hi everyone,

Been lurking on here for a while. I wanted to do the mod so I could justify getting the i7 😅

Prior to this I had a 17" MacBook Pro 2009 so I wanted to future proof as much as possible. I've also wanted an Air for the longest time.

I got some Gelid GP Extreme Thermal Pad 1.5mm Thick for my i7 MacBook Air. Pre mod score was 1278 / 3437. Didn't quite hit 4000 so maybe didn't do the best job. I think I might do the shim once my one year warranty runs out.

Thanks everyone for all the guides and help.
 

Attachments

  • Screenshot 2020-06-09 at 4.12.14 pm.png
    Screenshot 2020-06-09 at 4.12.14 pm.png
    229.6 KB · Views: 149
  • Like
Reactions: agaskew and Loog

_neon

macrumors newbie
May 22, 2020
1
4
Hi,

i have the i7/16GB 2020 MBA.

I did the following:

Shim Mod:
2x Copper plate 15x15x0.3mm (one cut)
Arctic MX4

Pad:
Arctic 50x50x1.5mm; cut about 10mm off, so that it fits the whole cooler.

Geekbench before Mod:

1300/3510
https://browser.geekbench.com/v5/cpu/2346030

Geekbench after Mod:

1305/4260
https://browser.geekbench.com/v5/cpu/2464277

Cinebench after Mod (no pre mod available):

1230

Ambient temperature: 23.5°C

I wouldn´t recommend the shim mod unless you know what you are doing.
Instead i recommend doing the Pad Mod. As IngerMan showed us with the i7
you can reach the 4000 Pts in GB by applying only the pad.
And to be honest its not worth risking the warranty on a new device by adding the shim.
I also tried the Thermal Grizzly Minus Pad 8 30x30x2mm but as mentioned
earlier it is way too thick.


Happy modding!
 

kinchee87

macrumors 6502
Jan 9, 2007
289
212
New Zealand
Hi,

i have the i7/16GB 2020 MBA.

I did the following:

Shim Mod:
2x Copper plate 15x15x0.3mm (one cut)
Arctic MX4

Pad:
Arctic 50x50x1.5mm; cut about 10mm off, so that it fits the whole cooler.

Geekbench before Mod:

1300/3510
https://browser.geekbench.com/v5/cpu/2346030

Geekbench after Mod:

1305/4260
https://browser.geekbench.com/v5/cpu/2464277

Cinebench after Mod (no pre mod available):

1230

Ambient temperature: 23.5°C

I wouldn´t recommend the shim mod unless you know what you are doing.
Instead i recommend doing the Pad Mod. As IngerMan showed us with the i7
you can reach the 4000 Pts in GB by applying only the pad.
And to be honest its not worth risking the warranty on a new device by adding the shim.
I also tried the Thermal Grizzly Minus Pad 8 30x30x2mm but as mentioned
earlier it is way too thick.


Happy modding!
Those are really impressive numbers!

What scares me is that some i5 modders are getting Geekbench 5 scores that are close to my i7 scores (~1280/~4000) 😂

Would be great if we could put our results in a shared spreadsheet to get an understanding of all the mod variations and improvements. It seems that the performance of a stock Air really depends on how well Apple has applied the black thermal paste.
 
  • Like
Reactions: excelsior.ink

nill1234

macrumors 6502
Dec 22, 2012
311
215
with shim mod my max GB score was 3914 with an I5 the heatpad just lowered the temperatures so no gains there. I think with an i5 3800-3900 is the maximum score in GB 5.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.