Be gone. Noisy Macbook Air!

[Tensakun]

If you go back in to the Apple store in the first couple of weeks after buying your computer, and tell them, "This is loud. I don't want a new one, but please check the thermal paste," will they do it?

I'd also like to know how they'd respond. Seems like this is a trick they could teach the Geniuses to do easily enuf, but a normal computer consumer should not have to crack open his/her machine and carefully spread goo around

If I were to do this myself, I'd be worried forever that I'd screwed something up.

And you'd have to change your name to 'Dekimasen' But I'm with you there--playing silly putty with an expensive hunk of hardware that I depend on to earn my living? I'm not that daring any more...

 

[wordy]

I was skeptical about thermal paste making that much of a difference, but I did have a small tube of AS5 lying around, so I decided to give this a try.

My idle temp decreased from ~52 degrees C to ~40. I actually observed it at a low of 38, which is much lower than prior to thermal paste operation.

Tip: You don't need special cleaners. From PC use, a good tip is to use regular coffee filters to clean excess paste off the chip. This works because the filters don't introduce fibers to the surface, which is exactly what you want.

Observations:
1. WAY too much thermal paste. So much so that it was oozing out off the chip and onto the board around it.

2. Screws were easy to remove. The 2 connectors are also easy to remove and put back. Keep track of them though, they're very small!

3. YES. You do need to remove the 2 connectors as there is ONE heatsink screw under them.

Total time ~30 mins. Could definitely do faster, but spent a good chunk of time looking at how this machine is put together, and also cleaning any dust I could find since I was in there anyway!

Procedure isn't difficult, but I'd say beginners should seek help.

 

[NC MacGuy]

I was skeptical about thermal paste making that much of a difference, but I did have a small tube of AS5 lying around, so I decided to give this a try.

My idle temp decreased from ~52 degrees C to ~40. I actually observed it at a low of 38, which is much lower than prior to thermal paste operation.

Tip: You don't need special cleaners. From PC use, a good tip is to use regular coffee filters to clean excess paste off the chip. This works because the filters don't introduce fibers to the surface, which is exactly what you want.

Observations:
1. WAY too much thermal paste. So much so that it was oozing out off the chip and onto the board around it.

2. Screws were easy to remove. The 2 connectors are also easy to remove and put back. Keep track of them though, they're very small!

3. YES. You do need to remove the 2 connectors as there is ONE heatsink screw under them.

Total time ~30 mins. Could definitely do faster, but spent a good chunk of time looking at how this machine is put together, and also cleaning any dust I could find since I was in there anyway!

Procedure isn't difficult, but I'd say beginners should seek help.

Okay. I can't stand it anymore. I must operate tomorrow.

I'll run a few baseline tests before and after. Curiosity is killing me.

Worst case scenario is it makes absolutely no difference w. heat at all and I clean out some dust & pet hair from the fan & vents.

Best case scenario is it may actually bring down temp a bit and prolong Air and battery life. Fan does use up some serious juice.

 

[weckart]

All I know is that I've heard the arctic silver paste in long term isnt good. I hear that the heat eats through the paste.

supporting evidence?

jj, you really gotta think or fact check before you post this kind of stuff. there are people that come here for guidance and see nonsense like what you've just said and take it to have value.

I've never once seen anything to support what you're saying and I've used AS for years and years. Maybe some batches were bad or something but to generalize like you have here is not helpful for anyone.

Plenty of evidence, anecdotal and otherwise. Here: http://www.google.com helps.

All pastes deteriorate over time - of course they do. However, it is very unlikely to bother the average user. We are talking about years rather than weeks. Just keep an eye on the average temperatures and decide on whether or not you need a top-up. Chances are, unless you have underapplied your coat of AS or other compound, this should last you until your next machine upgrade.

 

[hellobenny]

I tried reapplying the thermal paste today and actually I think that temperatures got hotter. The paste I used was the dow and corning tc-1996, some paste that I think comes with intel processors (i heard it was pretty good though on some overclocking sites). Maybe I'll try using arctic silver (thats stuff is pretty hard to come by in Vietnam though).

I think that it got hotter because I applied too little paste though, so I will open it up tomorrow and try adding more this time.

 

[wordy]

Paste application tip:

Don't follow the advice about using a razor blade to apply the coat. Instead, just put a dab in the middle of the chip, and when you put the heatsink back on, squish it around a little to spread the paste over the chip. The point of the paste is to provide a continuous surface for heat to travel from the chip to heatsink. I think if you guys are spreading it with a razor, yes, it will be thin, but it doesn't accomplish goal #1, which is to have maximum contact with both the processor and heatsink.

It just takes a little experimentation to see how big of an initial dab of paste you need to put. Don't be afraid to put the heatsink down, and then pull it up again and see how it has spread. Really, ever square millimeter doesn't need to be covered - it's all about getting maximum contact with minimum paste.

 

[texasdoc]

Did the paste re-application today with Arctic Silver. I'm pretty sure I got the right amount on there--not too much, but enough for good contact. AND.... It made ZERO difference. Temperatures are within a degree or two of what they were previously.

 

[NC MacGuy]

Did mine today too. Seems to have gotten hotter using Arctic Silver 5. Still analyzing results but does seem hotter. Not all bad though.

 

[silverblack]

Don't follow the advice about using a razor blade to apply the coat. Instead, just put a dab in the middle of the chip, and when you put the heatsink back on, squish it around a little to spread the paste over the chip.

This is good advice, I absolutely agree. Unlike the heat sink of a desktop computer, the one in MBA is a piece of thin metal, which is not necessarily perfectly parallel with the cpu's. So, if you only have a razor thin layer of thermal paste on the cpu, it's UNLIKELY to bridge the contact with the heatsink. THERE IS A REASON WHY APPLE APPLY SO MUCH PASTE at factory - it's to make sure it makes contact.

Learning from mistake, if you cpu cores run up to 90C at high load when heatsink temp stays at 60C - then you have a contact problem. When there's good contact, you cpu's should max out at ~70C, with heat sink always a few degrees cooler than cpu.

 

[Cheffy Dave]

First of all, let me start out by saying that I love Macbook Air.

Apple really knows how to attach emotion to the devices they make. It's something that no other company can do with their products.

If I had to pick one major fault for Macbook Air (1 USB port and lack of CD drive don't really bother me) is that it is noisy.

For a product that's designed mainly for light weight tasks, fan reving at 6200rpm after watching youtube or running screen saver is not favorable.

From browsing around the forum, I started to think that the source of heat issue on Air came from the way Apple applied thermal paste.

Poor application of thermal paste is a common practice for many companies and I guess it was also an oversight that Apple has made.

So I pop opened the back. I've done it by unscrewing screws in the back. It is probably one of the easiest laptops to open its innards.

I was greeted immediate with the heatsink. I was surprised. Heat sink is nothing but a thin plate of aluminum. No fins. No nothing.

I unscrewed the heatsink from the PCB.

And viola...

As you can see..they overused themal paste. This isn't a sandwich. It's a CPU core.



After a clean up.




Much better



Applied Arctic Silver 5..


Then I reassembled everything back up.

Ran youtube..and waited for the temperature to go up.

Result
When watching youtube video at full screen for 10 min.

Before
Core temp- 71 c
Fan- 6200rpm


After
Core temp- 63c
Fan- 4200rpm

Needless to say. HUGE improvement.

If you got a noisy Macbook Air...and if you got the gut to open up and reapply thermal paste..I'd say go for it.

Thank you so much, for posting the article and pics, you Sir are a gentleman, amazing how a little quality application of Artic Silver (btw, the best out there, yes it does make a difference) went a long way to curing the problem ,thanks for the post

 

[NC MacGuy]

I just did a secure empty trash of 500 meg file on USB drive and mine got up to 83º. Got some interesting and unexpected results I'll try and throw together later.

 

[wordy]

NC, from your pics I noticed that the amount of thermal paste on both your CPU and GPU seem to be less than was on mine, and it seems to have been better distributed too.

Might explain why you haven't seen much, or any, of an improvement.

 

[NC MacGuy]

NC, from your pics I noticed that the amount of thermal paste on both your CPU and GPU seem to be less than was on mine, and it seems to have been better distributed too.

Might explain why you haven't seen much, or any, of an improvement.

Might very well be the case.

I actually did this because of a super heating MBP. The Air is worlds easier to take apart and I wanted to see if temp. change would justify cracking the MBP open again.

I tried to do a semi-controlled test:
Temps., fan speeds, proc. data pic's taken at close to same intervals before & after tinkering.
1) Restart from cold
2) Start iStatPro - no other apps. at all
3) Battery charging at 80% charge or higher - trickle charge
4) Insert USB flash drive and transfer one 120MB episode of MP4 encoded episode of The Office
5) Watch Office episode in QT as large a screen as possible, screen full brightness, volume max.
6) Office finishes, trash it and do a secure empty trash
7) Load four more episodes of Office on flash drive at 500MB total
8) Secure empty trash the 500 MB files transferred

Shutdown. Tinker w. Arctic Silver 5. Have cocktail. Repeat tests.

 

[wordy]

I trust myself more than some assembly line worker in China, and while AS5 might not be the best paste in the world, from PC overclocking I know it is more capable than the average paste a manufacturer would use.

If you think your application was good, and you got good contact, your next step should be conditioning. Meaning you turn on your MBA, and stress it... i.e. 10min+ YouTube video or CPU Test etc then shut off MBA and leave it for several hours, then repeat. Someone quoted the break in period as being ~ 20 hours. I'm not sure if that's true, but from PC experience, I do know it takes a little while to see the full benefits of replacing thermal paste.

 

[NC MacGuy]

I trust myself more than some assembly line worker in China, and while AS5 might not be the best paste in the world, from PC overclocking I know it is more capable than the average paste a manufacturer would use.

If you think your application was good, and you got good contact, your next step should be conditioning. Meaning you turn on your MBA, and stress it... i.e. 10min+ YouTube video or CPU Test etc then shut off MBA and leave it for several hours, then repeat. Someone quoted the break in period as being ~ 20 hours. I'm not sure if that's true, but from PC experience, I do know it takes a little while to see the full benefits of replacing thermal paste.

Yeah, I know You probably couldn't see in the pic's but I have plenty of different heat sink compounds on hand and do this kind of crap for my livelihood so not quite a noob.
I wanted to see what happens after a few thermal cycles before posting I had good or bad results. Going through cycle three.
22 minutes total job time and 1.5 hours total test times.

Hopefully I'll see 1º difference & if so, I can operate on my MBP which I loathe opening.

 

[Eric.]

I have a Week 1 MBA, with Coolbook running. But it wasn't running that hot even before Coolbook; i.e., fan didn't kick in under normal web browsing, and running Microsoft Office apps. Although Coolbook did lower the max temp from >100C (before) to ~ 92C when cpu is stressed (after).

Today, I cleaned the excess thermal paste. Did I notice any difference? Not really. Cpu temp still stayed around mid-50 C under normal use, and it still climbed to over 90C under stress.

IMO, don't bother with this, unless your MBA fan does spin up to 6000 rpm even under normal use.

WHOA!

100C? That's getting into GPU territory. I'd have taken that back ASAP.

 

[instant hit]

tried reapplying paste to my week 5 mba 2 weeks ago, thanks to this thread. at the first go-around i did the dab-on-the-chip method and managed overall to reduce my temps by 2C. needless to say, i wasn't getting the dramatic decrease in fan rpms after say browsing youtube for 5 minutes.

going back to this thread i saw some were getting better results using the spread-with-razor-blade method so i thought i'd give that a go. so i opened up the mba again yesterday, removed the as5, then reapplied with a razor blade.

with that method, what i previously thought i'd never go through happened: a core shutdown. this happened while streaming a 640x480 divx file in mplayer. normally the fans would just spool to 6200rpm and keep the cpu in check at around 58-62C. with the reapplication, the cpu hit 83C and shut down a core.

guess i'll open it up a third time. YMMV

 

[netddos]

I didn't expect this thread would garner this much attention.

I guess there is overclock's blood flowing among many Air users.

With that said, me coming from the background of overclocking PC since.....forever, thermal paste application comes as a second nature.

That might not be the case for many who haven't ventured into the overclocking scene so I will try to be as clear as possible with my steps.

1. Open up your Air till you see the heatsink and the fan.

2. Remove two cables (HDD & plug cable) from the mainboard for clearance.

3. Remove heatsink by unscrewing it from the mainboard.

4. You will notice that Apple overused the paste. My guess is that by overusing the thermal paste they "insure" that it will at LEAST make a contact with the core. Although that might be the case, it is not the best way of transferring heat as the excess will act more like an insulator that traps heat.

5. You don't have to invest in products designed specifically for cleaning up thermal paste. All you need is pure alcohol (more concentrated the better as it will leave less residue..I recommend more than 75% at least) that you can get from any department store.

6. Before you go on with the cleaning up process, I recommend that you scrape off the excess thermal paste around the core with a thin sheet of paper.

7. Once it has been moderately cleaned, use lint free cloth+ alcohol to clean up the remaining thermal paste from the heatsink and the core.

8. This is probably the hardest part of this process. You have to apply just the right amount of thermal paste to the core. For this, yourself have to be the judge because generic instruction will not match up with the surface area of Air's CPU/Bridge core.
Apply a thin layer paste while maintaining a good contact with the heatsink. One way to check is to screw the heatsink back on and unscrew it back again to see if the core has made a good contact with the heatsink. Getting it right the first time is best because it will minimize air pockets getting stuck in between the core and the heatsink.

9. You don't necessarily have to do this but I recommend it for the best result. You probably have noticed there is a "L" shaped bracket that acts as a clamp. You might want to bend the bracket a little so that it will apply even more pressure when it is screwed back on. It will push the heatsink with greater force so naturally the thermal conductivity will increase even more.

10. That's it. Hopefully you will see the similar result as I have. Remember that there is a break-in period of about 24 hrs for maximum result. Good luck.

 

[hellobenny]

I just did the application the 2nd time. I think I did a horrible job the first time because this time the machine stays much cooler.

This time I used arctic silver 5 as opposed to tc-1996. Also I spend a good 20 minutes with cuetips and a bottle of ethanol alcohol getting out all the chunks of leftover paste from the previous applications. After everything was good and clean I applied the paste, dab style instead of using a razor to spread it thin. I put on just enough so that when the heatsink is pressed down on the cpus, the paste spreads to the edges, but doesnt spill off. To check that I applied the right amount, I laid back down the heatsink, then lifted it up again to see how the paste spread. After getting the right amount, I laid back the heatsink and screwed everything back up (btw, when laying on the back panel again, is the the back edge side closer to the right (usb port side), stick up a little higher than the back edge side closer to the left (power adapter side) for you?).

Now everything is running cooler in my opinion and I'm finally happy with my macbook air. (I really hope this machine lasts at least 2 years without problems, if so I'd be very happy) Yay!

Oh yeah I forgot my final step was to remove the istatpro panel from my widgets, otherwise I'd continue to waste too much time paying attention to the temp of my machine as opposed to actually using the machine.

 

[hkq37]

I just did the application the 2nd time. I think I did a horrible job the first time because this time the machine stays much cooler.

This time I used arctic silver 5 as opposed to tc-1996. Also I spend a good 20 minutes with cuetips and a bottle of ethanol alcohol getting out all the chunks of leftover paste from the previous applications. After everything was good and clean I applied the paste, dab style instead of using a razor to spread it thin. I put on just enough so that when the heatsink is pressed down on the cpus, the paste spreads to the edges, but doesnt spill off. To check that I applied the right amount, I laid back down the heatsink, then lifted it up again to see how the paste spread. After getting the right amount, I laid back the heatsink and screwed everything back up (btw, when laying on the back panel again, is the the back edge side closer to the right (usb port side), stick up a little higher than the back edge side closer to the left (power adapter side) for you?).

LOL! I know what you mean by constantly checking istatpro. But I reckon I am going to have to do the same as you - re-re-re-apply. I am doing mine for a 3rd time when I get home from work tonight.

Chuck out what I previously know about applying thermal paste to CPUs and heatsinks and dab it a little more liberally.

My first application - a fine film of thermal paste didn't even make a connection- so I found out after re-opening and examining..
My second application - I did it 3 days ago and have been running tests since.
Right at this minute I have CPU 53c, Heatsink 50c and 2500rpm ( web browsing, Azureus , chat i.e. light load)
I ran CPUtest app. for a few minutes and I was getting CPU 90c whilst the heatsink was only 60c and the fan 6200rpm --> seems to indicate a poor CPU-heatsink contact. I did check the cpu/heatsink connection before I fastened the screws, it seemed to make a 'connection in places'
How I wish the CPU and heatsink would contact more flushly!!

But it was good to hear about your results. If you still have istatpro maybe you could edit your post to include some tests and stats/temps?
I am getting a bit worried I might bend the heatsink which would really throw things out of whack. I just hope AS5 is the good stuff and still gives good conduction through a relatively thick layer. I'll repost my findings, if you are like me you will be interested to see other users results and temps.

UPDATE:Opened it up again applied more AS5, checked the contact. (Sidenote: Oops.. Forgot to shutdown! just closed the lid off i went.. not recommended !!!).
Ran CPUtest huge 10 cycles + all other normal apps. inc. Azureus and Dashboard as well as firefox playing a youtube video!! both CPU's were constantly pushing100% for 20 minutes :result CPU 80MAX ( most of the time it was 66-72!)c, Heatsink 63c, 6200RPM steady and stable
*tests done on battery power*

One other poster suggested bending the boomerang heatsink clamp to apply more pressure. I actually put a couple of pieces of lightweight cardboard under the pressure points to help push down over the chips. Great result, very happy.
Check out the pic.. actually taken whilst 1 core had shut down !! Cores shutting down at 66c ! possibly due to this -time to try Coolbook
I have attached a screen shot at the 13 minute mark you can see the apps running

 

[Stratus Fear]

I went ahead and did this as well on my 1.8/80, but I couldn't find my old tube of AS5, so I bought some Tuniq TX-2 from newegg.com that a number of people recommended (electrically non-conductive, less time to set than AS5, a little bit better cooling performance). I didn't get a dramatic decrease in the fan speed -- it still kicks in towards 6200rpm once the MBA's enclosure thermal sensors get up high, but the CPU core temps are lower when doing light work (low-mid 40Cs) and gets as low as 32C-35C when completely idle for a while. Temperatures drop fast after stopping a CPU intensive operation, and things like H.264 video run 10C cooler than they used to (~65C vs ~75C before). Heatsink temp usually stays close to the core temp, after 60C they diverge a bit since the fan turns on, sucking hot air away from the heatsink. As I write this, iTunes is throwing my iPhoto library on my iPhone for the first time, 2.2k pictures, the optimizing photos step has the cores at 80C, heatsink at 62C, but unlike before, it's running 1800mhz about 98% of the time without throttling down very often, and no core shutdowns. My results both before and after are under the use of coolbook, though, for reference. Between coolbook and redoing the thermal paste, this machine has become highly useable under some more intense processes. The fan is still annoying sometimes, but the core shutdowns and throttling are far less than they were previously (mostly non-existent). Videos and YouTube, etc. are no longer the problems they were before. And it runs nice 'n cool on light loads

Attached some pictures of the heatsink disassembly. Apple's definitely got this kind of messy. But you can make them nice and shiny clean to apply the new paste with a little bit of work

 

[wordy]

You probably have noticed there is a "L" shaped bracket that acts as a clamp. You might want to bend the bracket a little so that it will apply even more pressure when it is screwed back on. It will push the heatsink with greater force so naturally the thermal conductivity will increase even more.

I think that L is there not as a bracket, but rather as a way of conducting heat. Typical formations are found on the BACK side of video cards, etc.

 

[netddos]

I think that L is there not as a bracket, but rather as a way of conducting heat. Typical formations are found on the BACK side of video cards, etc.

That L might "help" to conduct heat better but it is definitely not just there as a way conducting heat. This part is just a bonus you get from thermal conduction.

If you screw everything back on without the bracket the aluminum plate does not have any real pressure to make good contact with the cores. The formation you mentioned on the video cards also does the same thing. It provides a surface for the heatsink to "grab" on to so the PCB can withstand the pressure when it is screwed on. The PCB will eventually warp if this kind of force is applied for a prolonged time without any support.

 

[wordy]

If you screw everything back on without the bracket the aluminum plate does not have any real pressure to make good contact with the cores.

There's a screw that goes through the L and heatsink, so I thought this point was a given. I do see how it could be used to distribute the force more evenly across both cores though.

 

[ayeying]

Here's the air flow. the L bracket should have some saying in how heat is dissipated.. sorry for the crappy paint job.. lol.