2020 Air Heatsink Modification Thread

[Tenkaykev]

My 20mm shim has arrived, I have the correct screwdrivers and a syringe of Thermalite TF8 heat sink compound.

I've been running the GB5 benchmark on my i5/8/256 stock Air.
Battery was charged to 100% and the charger then disconnected. All programs closed except Intel Power Gadget and GB5.
Ambient temperature was 20 degrees, MacBook Air on a solid wood table. Ten runs of GB5 with the temperature allowed to drop to 35 before starting the next run.

Average single core score: 1021, highest 1172, lowest 1039
Average multi core score: 3168, highest 3371, lowest 2860

I then ran five back to back GB5 tests. the temperature was 35 degrees at the start of the first test and had typically dropped to 75 degrees at the start of the subsequent tests. Max fan speed of 6.5k at the end of test 5.

Average single core score: 1080, highest 1114, lowest 1035
Average multi core score: 2917, highest 3120, lowest 2787

I also have a " Gelid GP Extreme " 1.5mm 12w /mk heat pad. Having seen some posters expressing concerns about the bottom case temperature I'll hold fire on fitting that.

Looking at the figures on my i5/8/256, the stock figures look quite respectable.
[automerge]1593530357[/automerge]

@Tenkaykev sry you misunderstood me. Those longer screws are for the bottom cover. Till now i didnt found a replacement for the heatsink.

Ah, OK ( I had a similar issue when changing the battery on my 2012 Air, I naively thought all the screws were the same length... )

 

[doctor_evil]

Have you tried a T4 and some epoxy or other glue? I think with epoxy you can mold a T3 in. I didn't found a source yet. They are 1.2mm width and have a flat head. BTW. I had the same Problem.

No, I decided to give up on the shim mod for now. If I can find a replacement heatsink screw, I will try to remove the stripped one and do the mod at that time, but I don't want to make it any worse without a replacement. Sounds like it's going to be a while until there's any replacements as this doesn't seem to be a screw used for anything besides this laptop.

That being said, with just the thermal pad, I got up to 1180 single core / 3780 multi core (i5), so not too shabby, assuming the thing doesn't crap out in under a year like a lot of the 2018 MBA's did!

 

[IngerMan]

A lot of us used the T4 because that’s what the iFixit starter kit came with. I highly recommend using a magnifying glass when trying to fit the wrench in the screw head, regardless if your using T3 or T4. Doing this you can also see if your angle is correct to the screw. Being on an angle only half seated In the head is a big reason for stripping. This thing is so small you need every benefit advantage you can get. If you don’t have a magnifying glass then try your camera on your iPhone. Knowing you bought a MBA you have to have a iPhone ?

 

[fcracer]

Having the right tools for the job is always important when working with computers or other delicate items. For the 2020 i7, the heatsink requires T3 and for the outer case screws, it requires P5. What's confusing in this thread is that there are several different computers being modified including a 2019 MBA and two versions of the 2020 MBA.

One hint I picked up on a youtube video (sorry, forgot who it is) is that when re-inserting the screw, to ensure the thread has been engaged correctly, turn the screw counter-clockwise first until you hear a click; then carefully go clockwise to tighten. That counterintuitive move makes it much safer to ensure you don't strip the threads (much more challenging to fix than a stripped screw head!).

 

[nill1234]

Still i think that the threads are ****** and that one of my threads was faulty from the beginning. The bottom screws are pretty short. Replacing two of the short ones ( right + left in the middle) with the longer ones from the mbp is possible, they are deep enough.

 

[excelsior.ink]

The Apple screws are too soft, I remember I found replacement screws for some older MBP models in the past from ebay. Much better, made of steel. But we have to wait for the Chinese to start manufacturing these.

 

[PhightinPhils26]

Just realized that my iPad Air 3 out performed my MBA 2020 i5/8GB with no modifications.

Couple takeaways:
- Thankfully the copper shim mod has corrected this.
- It's going to be interesting to see what the ARM based Mac's are going to do with Apple Silicone.

 

[nill1234]

I have just read that apple enables VP9 in Big Sur and Safari. So its possible to watch youtube in 4k without hear and noise...finally

 

[excelsior.ink]

With the upcoming transition of Apple from Intel to Apple Silicone I am now convinced that they did the recessed heatsink design on purpose. They will likely create Macbook Airs with the new Apple Silicon and they know reviewers will use GeekBench to compare the Intel and Apple Silicone. They wanted to be sure Apple Silicone will outperform by quite some margin Intel. Think about it, they put the effort to mill the heatsink. It's that space we are filling with a copper shim.
[automerge]1593671349[/automerge]

I have just read that apple enables VP9 in Big Sur and Safari. So its possible to watch youtube in 4k without hear and noise...finally

Is is already enabled or is it an upcoming feature in a future beta release?

 

[silvia_18]

Safari can now stream 4K HDR and Dolby Vision content on Netflix with macOS Big Sur

With the upcoming transition of Apple from Intel to Apple Silicone I am now convinced that they did the recessed heatsink design on purpose. They will likely create Macbook Airs with the new Apple Silicon and they know reviewers will use GeekBench to compare the Intel and Apple Silicone. They wanted to be sure Apple Silicone will outperform by quite some margin Intel. Think about it, they put the effort to mill the heatsink. It's that space we are filling with a copper shim.
[automerge]1593671349[/automerge]

Is is already enabled or is it an upcoming feature in a future beta release?

Kinda? Not yet available in Safari?

Safari can now stream 4K HDR and Dolby Vision content on Netflix with macOS Big Sur - 9to5Mac

We recently mentioned that macOS Big Sur includes a new version of Safari with several enhancements, including support for HDR...

9to5mac.com

 

[srkirt]

After seeing the performance of Huawei, I doubt that I will return to Apple at least on laptops.

 

[IngerMan]

After seeing the performance of Huawei, I doubt that I will return to Apple at least on laptops.

its probably a good thing, I see a spark in your future about the 6th time you take apart your Huawei and your next replacement won't be as much as Apple. FYI I see lots of places in your picture where you can get better air flow, get cracking

 

[srkirt]

[QUOTE = "IngerMan, publicación: 28634844, miembro: 543032"]
Probablemente sea algo bueno, veo una chispa en su futuro sobre la sexta vez que desmonta su Huawei ags y su próximo reemplazo no será tanto como Apple. Para su información, veo muchos lugares en su imagen donde puede obtener un mejor flujo de aire, agrietarse:RE
[/CITAR]
?

its probably a good thing, I see a spark in your future about the 6th time you take apart your Huawei and your next replacement won't be as much as Apple. FYI I see lots of places in your picture where you can get better air flow, get cracking

Here the problem is that there is no sotfware to control the fans ... and the case is overheated ... do not think I'm afraid because I have the MBA as decoration as a book ... this is more economic € 624 with Amd Ryzen 5 3,500 and with a score of 1445 in Cinebench. Screws panic me ... when the strands catch few threads and you can run out of threads. If I could speed up the fans this would be a missile !!!

 

[Tenkaykev]

Finally got round to fitting shim to my i5/8/256
I had 2 off 15x15 shims. I cut 5mm off each to give me an effective 20x15.
Fairly thick black paste on existing heatsink. Cleaned the copper part of the heatsink and used a bit of 6000grit wet and dry. The Thermalrite TF8 heat sink compound didn't seem to want to adhere and it was a bit of a job to make sure there was at least some on the relevant parts.
After reassembly I ran Geekbench and was concerned as I thought I might have damaged the fan connection as the fan didn't start until the temp reached 76C
quite pleased with a score of 1155 / 3986, shim only and no pad, that's tantalisingly close to the magic 4000 for an i5.
2781945

 

[nill1234]

Wow thats a good score dont know why your score is so different to the others? I didnt polish it just sanded it a bit down...maybe thats the problem?
Can you rerun the test?

Found this test, which measures the temperatures with different sand papers.

Overall the sand paper has only little effect, thermal paste will take care of the scratches.

 

[fcracer]

Wow thats a good score dont know why your score is so different to the others? I didnt polish it just sanded it a bit down...maybe thats the problem?
Can you rerun the test?

There are three ways people in this thread are measuring the performance of the modifications:

1. Turn on from cold and immediately run the benchmark. On an i7, with shim and paste mod, this will yield ~4050 GB MC score.
2. Allow to reach idle temperature of <40C and run benchmark. On i7, this will yield ~3850 GB MC score.
3. Run the benchmark at least once and then run the benchmark a few times in succession without letting it cool. For the i7, this will yield ~3450 GB MC score.

Therefore you need to follow the same process to get comparable numbers. If you add in the pad modification, the gap between the #1 and #3 methods closes and you’ll yield ~4050-4150 using any of the three methods.

 

[Tozovac]

If I were these engineers I would Ask my money back from whatever phd program they came out of.

As it stand it can’t load 100% for more than couple minutes without throttling, never mind 24/7

No doubt Apple hires the best but it doesn’t mean they don’t blunder, the utter failure of butterfly keyboard is a pure example of that

They don’t need to add another fan, just add a heat pipe, the thermal cooling benefit it brings would actually reduce fan rpm, thus lowering battery usage

the 2018 mba internal design is clear they went with form over function

Assuming the MBA’s prioritization on thinness can be blamed for this hampered performance condition (like with the butterfly keyboard), it’s not the engineers’ fault, it’s Tim’s senior product planning management that made another questionable decision focusing on form over function.

 

[fcracer]

Assuming the MBA’s prioritization on thinness can be blamed for this hampered performance condition (like with the butterfly keyboard), it’s not the engineers’ fault, it’s Tim’s senior product planning management that made another questionable decision focusing on form over function.

I don't think this is a design issue, I think it's an execution issue. With proper thermal paste, the i7 produces 4050 Geekbench MC scores from cold and 3450 when heat soaked after multiple runs. I think the heat soak result fits perfectly under the low-end MBP, while giving incredible performance in bursty workloads.

 

[Tozovac]

I don't think this is a design issue, I think it's an execution issue. With proper thermal paste, the i7 produces 4050 Geekbench MC scores from cold and 3450 when heat soaked after multiple runs. I think the heat soak result fits perfectly under the low-end MBP, while giving incredible performance in bursty workloads.

Good point. My possibly uninformed opinion (from having not used an MBA but just reading about them) is that I personally can’t get past the putting a V8 engine in a Cooper Mini here — those short bursts sound to be too short to be of significant usefulness, resulting in more frustration from most users than appreciation, you know?.

 

[Tenkaykev]

There are three ways people in this thread are measuring the performance of the modifications:

1. Turn on from cold and immediately run the benchmark. On an i7, with shim and paste mod, this will yield ~4050 GB MC score.
2. Allow to reach idle temperature of <40C and run benchmark. On i7, this will yield ~3850 GB MC score.
3. Run the benchmark at least once and then run the benchmark a few times in succession without letting it cool. For the i7, this will yield ~3450 GB MC score.

Therefore you need to follow the same process to get comparable numbers. If you add in the pad modification, the gap between the #1 and #3 methods closes and you’ll yield ~4050-4150 using any of the three methods.

Yes, I agree completely re the testing. I've rerun a 5 back to back test sequence on GB5, starting temperature was 40C.
3802 was max, 3101 was minimum with a 5 test average of 3270. Pre mod this figure was 2917.
The fan max was ~ 4.3K and there was a much longer delay before the fan started to ramp up.

I then closed everything apart from GB5 and Intel Power Gadget, waited for the temperature to drop to 25C and ran a single test. This gave 1179 - 3964. Of interest is that every GB5 test sees a power spike of 25W as soon as "Text Compression" initialises.

I'm really tempted to whip the back off and stick a thermal pad on the heat sink and see what figure I'd get from a fresh start.

As a matter of interest. when I did the shim / heatsink compound mod yesterday I was extremely methodical, all the tools laid out ready, sheet of paper with a sketch of the back of the machine so I could place the removed screws in the corresponding position on the drawing. Kitchen roll, alcohol wipes, 6000grit wet and dry tooth pick for getting the old compound from the nooks and crannies on the die / heatsink.
It was only when I turned the computer back over and opened the lid to be presented with my screensaver that I realised I'd forgotten to turn the bloody thing off before I started

 

[agaskew]

Yes, I agree completely re the testing. I've rerun a 5 back to back test sequence on GB5, starting temperature was 40C.
3802 was max, 3101 was minimum with a 5 test average of 3270. Pre mod this figure was 2917.
The fan max was ~ 4.3K and there was a much longer delay before the fan started to ramp up.

I then closed everything apart from GB5 and Intel Power Gadget, waited for the temperature to drop to 25C and ran a single test. This gave 1179 - 3964. Of interest is that every GB5 test sees a power spike of 25W as soon as "Text Compression" initialises.

I'm really tempted to whip the back off and stick a thermal pad on the heat sink and see what figure I'd get from a fresh start.

As a matter of interest. when I did the shim / heatsink compound mod yesterday I was extremely methodical, all the tools laid out ready, sheet of paper with a sketch of the back of the machine so I could place the removed screws in the corresponding position on the drawing. Kitchen roll, alcohol wipes, 6000grit wet and dry tooth pick for getting the old compound from the nooks and crannies on the die / heatsink.
It was only when I turned the computer back over and opened the lid to be presented with my screensaver that I realised I'd forgotten to turn the bloody thing off before I started

Wowsers, you mean you did the entire mod without shutting down first? I guess that proves it really was asleep.

 

[Tenkaykev]

Here is the original Processor / Heatsink / Thermal Paste arrangement.

 

[srkirt]

Here is the original Processor / Heatsink / Thermal Paste arrangement.

View attachment 930280 View attachment 930279 View attachment 930281

The first thing I did in this thread and we were all surprised ... then came the wedge with Red Bull tin aluminum ... then with aluminum foil ... until I got copper and the rest is history ...

 

[WhiskyDel]

Ok so I've been following this thread ever since I received my MacBook Air (2020 i5/8gb/256) and was disappointed by the fans hitting 8000rpm on first boot up. I downloaded Intel power gadget to check temps etc and was a bit shocked to see the cpu temp at 100c without anything running. After contacting apple support by web chat, a reset of the NVRAM cured that issue; all good I thought until I tried a zoom call where temps immediately hit 100c and the fan was 8000rpm again but ending it and temps dropped. Thats when I started following on this thread and found I wasn't alone in my experience so out of curiosity I tried Geekbench 5 just to see how bad things were.

out of box scores were SC 918 and MC 2075!

Taking things slowly I tried the thermal pad mod and immediately noticed an improvement!
Single core = 1049
Multicore 2686

I thought this was promising and let it 'bed in' for a while with the scores increasing over time til I reached a plateau;
Single core = 1100
Multicore = 3500

I was tempted to leave it there since it was now silent and performing well.... until I saw the Maxtech watercooling video; I wasn't interested in water-cooling but the results from using a cold gel pack looked good so I tried it;
Single core=1194
Multi core= 3863

Without it the scores reverted to around 1100/3500 so I knew there was definitely a bit more thermal headroom to make use of.

That gave me the push to go for it and try the copper shim+Thermalright TF8 paste + Arctic thermal pad.

Immediately after doing this I ran Geekbench again and was pleased to see I was right; for 2 consecutive runs with no cooling period;
single core=1184,1196
Multi core=3834,3831

Temperature was sitting at 80c with the fan at 2700 and cpu utilisation at 97% in the multi core section.

To be honest I think I've hit the ceiling on the i5 and I'm now happy that this is the MacBook Air I wanted to purchase the first place. I suspect it would even give the base MacBook Pro a run for it's money!

 

[Tenkaykev]

Ok so I've been following this thread ever since I received my MacBook Air (2020 i5/8gb/256) and was disappointed by the fans hitting 8000rpm on first boot up. I downloaded Intel power gadget to check temps etc and was a bit shocked to see the cpu temp at 100c without anything running. After contacting apple support by web chat, a reset of the NVRAM cured that issue; all good I thought until I tried a zoom call where temps immediately hit 100c and the fan was 8000rpm again but ending it and temps dropped. Thats when I started following on this thread and found I wasn't alone in my experience so out of curiosity I tried Geekbench 5 just to see how bad things were.

out of box scores were SC 918 and MC 2075!

Taking things slowly I tried the thermal pad mod and immediately noticed an improvement!
Single core = 1049
Multicore 2686

I thought this was promising and let it 'bed in' for a while with the scores increasing over time til I reached a plateau;
Single core = 1100
Multicore = 3500

I was tempted to leave it there since it was now silent and performing well.... until I saw the Maxtech watercooling video; I wasn't interested in water-cooling but the results from using a cold gel pack looked good so I tried it;
Single core=1194
Multi core= 3863

Without it the scores reverted to around 1100/3500 so I knew there was definitely a bit more thermal headroom to make use of.

That gave me the push to go for it and try the copper shim+Thermalright TF8 paste + Arctic thermal pad.

Immediately after doing this I ran Geekbench again and was pleased to see I was right; for 2 consecutive runs with no cooling period;
single core=1184,1196
Multi core=3834,3831

Temperature was sitting at 80c with the fan at 2700 and cpu utilisation at 97% in the multi core section.

To be honest I think I've hit the ceiling on the i5 and I'm now happy that this is the MacBook Air I wanted to purchase the first place. I suspect it would even give the base MacBook Pro a run for it's money!

I'm tempted to open my Air up again and try the thermal pad mod. I'm getting regular readings in the low to mid 3900's with a high of 3986 on my first run after the shim mod. I'm sure that I've hit the limit but it's a bit like Schrodingers cat, there's always the possibility that I might see a 4k GB5 if I don't look ;-)