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

2020 Air Heatsink Modification Thread

prism said:
Bummer, I ordered the 1mm pad as suggested in the first post of this thread!
Click to expand...
I did a test of which thermal pad was better 1mm or 1.5mm and the 1.5 was the same single-core score on geekbench but was 50 points higher on multicore I have a i5 MacBook air. So either work in my eyes if you really want the performance then switch to 1.5 or just wait 1-2 years and then swape out the thermal pad when it starts to degrade.
 
fcracer said:
I've got a few DM's from people asking about the modifications. To make it easier for people contemplating the paste/shim/pad modifications, I've put a how-to guide online with pictures and test results. Hope you find it useful.
Click to expand...
Thank you for the writeup. Long time lurker second or third time poster (my previous posts were about the USB-C disconnection issue, which I fixed by throwing money *cough* CalDigit *cough* at the problem)...

Anyway here's a video I cooked up applying the things in your post (random copper shim, 1mm pad)

Has affiliate links, but also links to this thread and your guide. (I have a low-subscriber cycling channel so wanted to make a new one for this)


Sadly I lost my old Cinebench R23 result (edit: according to the "result file" it's 2667pts), new one is at 2967pts (11.3% improvement). Geekbench5 MC 2833 to 3806 (32% improvement)
 
Last edited:
  • Like
Reactions: Loog and Tenkaykev
Ok, I did the shim and paste mod because that's what was delivered from Amazon first, lol.

If I do the test from cold start, I get
single: 1165
Multi : 3558

vs. before mod scores of
single: 1082
Multi : 3036

I ran the modded test several times, the singles are always above 1100 and the multis vary between 3100 and the highest score (3558).

Before the mod, the lowest multi I had was around 2300.
What do you think of all this?

I will report back once I get and install the thermal pad.
 
disq said:
Thank you for the writeup. Long time lurker second or third time poster (my previous posts were about the USB-C disconnection issue, which I fixed by throwing money *cough* CalDigit *cough* at the problem)...

Anyway here's a video I cooked up applying the things in your post (random copper shim, 1mm pad)

Has affiliate links, but also links to this thread and your guide. (I have a low-subscriber cycling channel so wanted to make a new one for this)


Sadly I lost my old Cinebench R23 result (edit: according to the "result file" it's 2667pts), new one is at 2967pts (11.3% improvement). Geekbench5 MC 2833 to 3806 (32% improvement)
Click to expand...
Your final scores are impressive, what I'm trying to figure it is if this is the highest you achieved, ie from cold wakeup or if you systematically hit those ranges irrespective of warm or cold conditions for running bench.
 
prism said:
Your final scores are impressive, what I'm trying to figure it is if this is the highest you achieved, ie from cold wakeup or if you systematically hit those ranges irrespective of warm or cold conditions for running bench.
Click to expand...
These were just ran once from cold, just after the mod in fact. Ambient temp was 22ish, no external monitor/CalDigit, just connected to power. Geekbench was done first, then launched Cinebench and waited to cool down back to low 40s, and ran the test. Bluetooth, Dropbox, iStat were all disabled, no browser open, but WiFi was connected. Intel Power Gadget was running but no Fanny. Max temp on the CPU was 92 deg C (during Cinebench), to my surprise it didn’t even hit 100.
 
disq said:
These were just ran once from cold, just after the mod in fact. Ambient temp was 22ish, no external monitor/CalDigit, just connected to power. Geekbench was done first, then launched Cinebench and waited to cool down back to low 40s, and ran the test. Bluetooth, Dropbox, iStat were all disabled, no browser open, but WiFi was connected. Intel Power Gadget was running but no Fanny. Max temp on the CPU was 92 deg C (during Cinebench), to my surprise it didn’t even hit 100.
Click to expand...
Thanks, did you also test consecutive tests from warm? I would be curious as to how much of a hit the scores take.
 
prism said:
Thanks, did you also test consecutive tests from warm? I would be curious as to how much of a hit the scores take.
Click to expand...
Didn’t have time had to get back to work (and/or cutting the video) I’ll try to do it later/tomorrow. But check out the linked guide (almost the same) or other modders benchmarks for an idea.
 
disq said:
Thank you for the writeup. Long time lurker second or third time poster (my previous posts were about the USB-C disconnection issue, which I fixed by throwing money *cough* CalDigit *cough* at the problem)...

Anyway here's a video I cooked up applying the things in your post (random copper shim, 1mm pad)

Has affiliate links, but also links to this thread and your guide. (I have a low-subscriber cycling channel so wanted to make a new one for this)
Click to expand...
Nice work on the video and no worries about affiliate links. You put work in to make the video, you should get benefits from it. When I have time, I’ll link to your video from the guide. The easier it is for people to make the mods, the better.
 
  • Like
Reactions: disq
Ok, so I added the thermal paste and lo and behold my new scores are
single: 1168
multi : 3800

and the best part is that I constantly get similar results after running the test a couple of times consecutively !
 
  • Like
Reactions: disq and IngerMan
Wojciech Lampa said:
Any teardown/disassembly or possibility of upgrading M1 MBA with some thermal pads?

I saw one pulling a fan underneath that improved Heaven results in loop to almost non-throttled, like MBP 13, so there is no software power limit but only thermal limitation and there might be a potential there.
Click to expand...
Hey where did you see this fan experiment?
 
Nitrogen said:
In some topic where subOP took Noctua 140mm fan under the MBA and said the machine pulled 25W again instead of throttled 14W in 100% combined CPU+GPU scenario.
Click to expand...

thats why I posted early wondering if someone plans to make a fan retrofit. I bought the air because I dislike the touchpad but wished it had a fan. Someone could easily create a kit that utilizes fan parts from the intel version and then create a chip that is programmed to adjust the fan based on CPU speeds. The only requirement I can see would be requiring a few wires to be soldered (or a manufactured piece that clips between another wire set).
 
I'm reporting back to say how impressed I am with the mod, it is like as if I have a new machine, I tried to push it testing FireWatch on it and I still couldn't hear the fan. It now actually doesn't go above the low 80's when I push it to its limits!
If I knew I would get such amazing results, I would have done this mod a long time ago.
 
  • Like
Reactions: agaskew and IngerMan
Been thinking of getting a thermal pad for intel mba 2020 i5. Does anyone know if the thermal pad stays in place? Travel a lot and I don't plan to open up my macbook often...
 
Interesting to see someone using thermal pads on the new M1 Air and getting sustained performance similar to the M1 Pro 👍
 
  • Like
Reactions: disq
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back