Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Video:
YouTube Video (starting @ 9m 26s) showing the heatsink and compound issue.

Let me preface this by saying the blame is NOT with Apple's engineers. I understand and absolutely applaud their methodology and execution in the Macbook Pro line.

However, Apple's manufacturing and assembly need help.

We're going to go into why:

Apple's machine work on the heatsinks is sub-par, immediately causing poor contact, and thus poor heat transfer from the CPU/GPU to the heatsink.

This is compounded with entirely too much heatsink compound on the dies.

First, a little information that seems to be *sorely* lacking in the Apple community (some materials engineering):

The very best heat transfer method is direct metal-to-metal contact. Unfortunately, once you separate a piece of metal into two pieces, you will always have air between the two pieces, due to surface imperfections.

Now, if you were to look at a flat piece of metal under a microscope, you would see this:
Image

See those valleys, dips, mountains?

Now notice the roughness average (known in the industry as RA) values.

The roughness average of a polished mirror finish is usually 3-5 microinches. That NiTi 80 has an RA of 1 micron (1 micron = ~39 microinches), and you see the valleys and mountains.

Here is an image of a proper heatsink base (Taken from Swiftech's website):
Image

Do you see how reflective the surface is? That means there are very few valleys. I'd say that Swiftech heatsink has an RA of 3-5 microinches.

When I pulled the heatsink out of my Macbook Pro, I was astounded. While I don't have a profilometer available, I'd estimate the RA of the heatsink copper pads in the neighborhood of 40-60 RA. That's freaking pathetic in a system that you pay the kind of money you do for a quality Macbook Pro. I've seen better heatsink bases in a DELL for crying out loud. Now, my lapping was not the best job in the world. I may have gotten the RA down to the 25 range, but it certainly helped.

If you'd like to read more on surface finishes, check out Wikipedia. Plenty of reading, and you can start doing some thinking on your own - Surface Finish

So what the heck does that have to do with all the heatsink goop I see spread around like peanut butter on a sammich? (mmm sammich - sorry, I'm hungry) Glad you asked!

Direct metal to metal is the best heat transfer mechanism, as we've already stated, but we have those surface imperfections that we can't even see to deal with.

Enter: heatsink compound.

See, air is a great insulator; air does not allow heat to transfer effectively. We use heatsink compound of low thermal resistance to 'fill in' those little valleys and imperfections, so we don't have air keeping our heat transfer at bay.

The normal white stuff you see is usually a ZnO (zinc oxide) mixture. And yes, it does the job. However, something like Arctic Silver works even better at facilitating that heat transfer (pure silver is about 10x as thermally conductive as ZnO), so we can move that heat out of the CPU and into the heatsink system.

That's also why it is SO important to have AS LITTLE HEATSINK COMPOUND AS POSSIBLE to fill those voids in the metal-to-metal contact patch. Copper is about 9x as thermally conductive as ZnO, so having too much heatsink compound hurts thermal transfer rate.

To reiterate, this is *not* the fault of Apple's engineering team. The heatpipe assembly is good, the fan control is good, the exhaust system is... well, ok, I think they could have done better there, honestly, but they did the best they could within those design parameters.

Under normal conditions that most use their fashionable Macbooks in, the raw heatpipe transfer will keep the laptop CPU/GPU under ~120F (49C). Heck, I have 11 tabs open in Chrome (one of which is Youtube, my channel actually) and carrying on a couple conversations in both Adium and Colloqey, and I'm sitting at 106F (41C) temperatures at the moment.

The CPU will remain clocked down, the discrete GPU stays off, and things are great.

Now, once you start pushing things, (such as encoding a 40+ minute 1080p Youtube video) CPU clocks come up, discrete GPU comes online and clocks up, and yes, heat starts getting really generated. Fans start ramping up to keep up with the thermal demands, exhaust gets blown out the hinge area, and away we all go.

I have hit 201F (94C) while doing these full power encodes, and once the fans are up and running (5500rpm, not even full 6200rpm) temperatures tend to stabilize around 185F (85C), which I imagine is right where the engineers want it maxed at around. If I pop the Macbook Pro up on a 1" thick book so the exhaust area has more room to 'breathe' and push the fans to 6200rpm, I can get my temps down to 175F (79C) while does those full encodes. I have no thermal throttling at any time for any thing.

Let me also mention that I have the BTO 2.5Ghz Quad i7 - NOT the normal 2.2 or 2.4 Ghz versions.

Whomever Apple has manufacturing their heatsinks, seriously needs to get on the ball with providing a properly machined surface to attach to the logic board.

Whomever Apple has assembling the laptops, needs to learn how to properly apply heatsink compound.

Combine too much heatsink compound with poor heatsink machine work, and you know what you have?

A Macbook Pro that runs too damn hot.

I hope this was helpful for you all. Questions? Comments? Keep your blind fanboy-ism, but constructive discussion, I'm all for.

*edit* Micron to microinch conversion - my fault. Used to dealing with µin, not µm;)

Possibly on of the most intelligent posts anywhere. Thank you. There is simply NO EXCUSE for this shoddy crappy "it still works OK, sort of, sometimes" type of work on an item like this.

Lower heat = fewer issues = lower support costs = better consumer experience/satisfaction.
 
Possibly on of the most intelligent posts anywhere. Thank you. There is simply NO EXCUSE for this shoddy crappy "it still works OK, sort of, sometimes" type of work on an item like this.

Lower heat = fewer issues = lower support costs = better consumer experience/satisfaction.

I for my part find iMacC2D's response on page 2 to be one of the most intelligent I've read on this forum in years. Check it out if you haven't already.
 
I must have re pasted 15 2011 models, and a few dozen more of the other mbp range over the past year but I go a bit further than Doward by using a lapping stone polishing up the contact plates thoroughly like mirrors and using gelid gc extreme paste and a few pairs of surgical gloves too.

I average idle temp drops of about 15c on the 2011, retinas with their much better cooling system 8-10c and I've even done it to the new Mac Pro when fitting an 8 core xeon with a much faster clock and got an 8c drop. iMac's too with 15c drops.

The heatsink plates are machined awfully on the 2011, the palm tree heatpipe which can cope with 2008-2010 and 2012 in the 15 can cope pasted the shoddy factory method but this sandy bridge CPU and AMD GPU combo its simply not the scratch cos these dies run too hot and in terms of mm squared the die areas are too large, you only need a magnifier to see that the surface is hardly even, more like a hundred microns uneven. Not good to bind a die to a heatsink one bit. And despite the rash of GPU failures in the 2011 and having referred many clients to get a reball not one single GPU has failed after I repasted and optimised the thermal system as purely a preventative measure. My own late 2011 17 had a 20c idle drop after treatment.

All I do know is empirically that it's about time apple engineering took a look at these methods in stopping these gpu failures in the 2011 by retrospectively changing the engineering guidelines so that they don't fail anywhere near as much.

Getting better machined parts, better paste helps a lot. I know cos I've been repasting CPUs and GPUs since 1996, voodoo 3d era if anyone here can remember!
 
Last edited:
The 2011 is a ****ing joke. I have the MBP. Late 2013, I replaced the fans after some time, and it turned out ok. Now, a year later, it's overheating again, preventing me from using Ableton.

I'm geekbenching 3200 on the single cores, which isn't TOO bad compared to retinas 2014...but, it's just constantly heating.
 
As an early poster in this thread who was dismissed as being silly, I just thought I'd say - even if I am late to the punch here - toldjaso!

http://www.tuaw.com/2014/10/29/apples-2011-macbook-pro-graphics-woes-lead-to-class-action-laws/

That's all for today

----------

That's obviously only your opinion, as notebooks are configured for a variety of intended uses. They are not all alike and not all configured to handle every possible task. That's why a MBP may be considered overkill for some users' needs, while it may be completely inadequate for others. It's up to each user to determine what their own usage requirements are, then shop intelligently for the computer that meets those needs. Buying the wrong computer for your needs and then blaming the manufacturer displays a lack of experience and unwillingness to accept responsibility for one's own inappropriate buying decisions.

In this case shopping intelligently would be looking into the future to find the 20,000+ users who have a useless brick. Quad core sandy bridge + hot AMD GPU sharing the same tiny tiny tiny heatsink with no heatpipe and junk thermal paste. Was anyone NOT expecting this??
 
  • Like
Reactions: Jubadub
As an early poster in this thread who was dismissed as being silly, I just thought I'd say - even if I am late to the punch here - toldjaso!

http://www.tuaw.com/2014/10/29/apples-2011-macbook-pro-graphics-woes-lead-to-class-action-laws/

That's all for today

----------



In this case shopping intelligently would be looking into the future to find the 20,000+ users who have a useless brick. Quad core sandy bridge + hot AMD GPU sharing the same tiny tiny tiny heatsink with no heatpipe and junk thermal paste. Was anyone NOT expecting this??
I got bummed by Radeongate as well.

And I will blame Foxconn (or whoever else that did a bloody crappy job applying the thermal paste) and also AMD for producing inefficient chips.

Even with the dGPU activated full-time on my late-2013 rMBP, the NVIDIA chip never gets as hot.
 
I got bummed by Radeongate as well.

And I will blame Foxconn (or whoever else that did a bloody crappy job applying the thermal paste) and also AMD for producing inefficient chips.

Even with the dGPU activated full-time on my late-2013 rMBP, the NVIDIA chip never gets as hot.

I blame Apple for not polishing those two copper plates like mirrors, foolishly thinking that it could cope dissipating 85 watts of thermal energy when it was initially designed for much less, a Core2Duo and a Nvidia 9600/9400GPU in fact. Polished plates and pasted just right it can cope, which costs a fraction of a redesigned system and shame on Apple for not doing exactly that.
 
That's also why it is SO important to have AS LITTLE HEATSINK COMPOUND AS POSSIBLE to fill those voids in the metal-to-metal contact patch.

This is false. Clamping force will force most of the excess out of the die contact area. The amount of TIM between heatsink and die will remain minimal no matter how much you use. Only thing that will increase is the amount that's squeezed out and remains around the die.

This example shows that extreme overapplication of TIM degrades thermal performance by only a few degrees. I'd wager the amount of TIM on that CPU nears the mass of the CPU itself.

Comparatively, the "normal" concept of overapplication would be a little more than a rice grain. This would have minimal thermal differences to a "proper" application. Again, this comes from the fact that the amount of TIM between the CPU and die will always be minimal due to clamping force squeezing out the excess.

In fact, using too little TIM will cause even worse performance than too much TIM. Why is this so? Because the TIM will not cover the entire die and fill the microscopic grooves in the heatsink. Spreading the TIM will do the same: it causes unequal application of TIM across the die causing air bubbles as well as reducing the amount of TIM that will be squeezed into the microscopic grooves.
 
This is false. Clamping force will force most of the excess out of the die contact area. The amount of TIM between heatsink and die will remain minimal no matter how much you use. Only thing that will increase is the amount that's squeezed out and remains around the die.

This example shows that extreme overapplication of TIM degrades thermal performance by only a few degrees. I'd wager the amount of TIM on that CPU nears the mass of the CPU itself.

Comparatively, the "normal" concept of overapplication would be a little more than a rice grain. This would have minimal thermal differences to a "proper" application. Again, this comes from the fact that the amount of TIM between the CPU and die will always be minimal due to clamping force squeezing out the excess.

In fact, using too little TIM will cause even worse performance than too much TIM. Why is this so? Because the TIM will not cover the entire die and fill the microscopic grooves in the heatsink. Spreading the TIM will do the same: it causes unequal application of TIM across the die causing air bubbles as well as reducing the amount of TIM that will be squeezed into the microscopic grooves.


As an experienced PC overclocking and cooling fan I can confirm that :) The importance of mirror finish and TI application are heavily overrated. I've seen a lot of "expert" videos where people remove Macbook (or any other OEM PC) heatsink and rave about the excessive thermal paste around the CPU die. In my opinion there're two main factors in MBP heatsink performance:

1. The quality of thermal paste. I think MX4 or AS5 are simply better than stock.

2. Good contact between the heatsink and the CPU/GPU dies (no skew). Make sure that the heatsink isn't bent because the spring force isn't enough to compensate it reliably.

Surface quality is important, but you won't see a big difference between a very good and perfect finish. I even tried liquid metal thermal compound (which I don't recommend unless you know what you're doing) and all I got was a 2-3C drop compared to MX4 at the maximum CPU load.


The bottleneck in all MacBooks is small radiator with barely enough surface to handle 100% load.
 
Last edited:
As an experienced PC overclocking and cooling fan I can confirm that :) The importance of mirror finish and TI application are heavily overrated. I've seen a lot of "expert" videos where people remove Macbook (or any other OEM PC) heatsink and rave about the excessive thermal paste around the CPU die. In my opinion there're two main factors in MBP heatsink performance:

1. The quality of thermal paste. I think MX4 or AS5 are simply better than stock.

2. Good contact between the heatsink and the CPU/GPU dies (no skew). Make sure that the heatsink isn't bent because the spring force isn't enough to compensate it reliably.

Surface quality is important, but you won't see a big difference between a very good and perfect finish. I even tried liquid metal thermal compound (which I don't recommend unless you know what you're doing) and all I got was a 2-3C drop compared to MX4 at the maximum CPU load.


The bottleneck in all MacBooks is small radiator with barely enough surface to handle 100% load.

I'm also well versed in system building, having run a delidded 4770K bare-die just like the forum post I posted. :)

Currently I'm using Antec Formula 7 on all my machines (desktop CPU, GPU card, laptop). It's just so much better than any stock compound I've come across, and it also seems to outperform AS5 as that used to be my go-to compound.

However, the OP's point that the grooves are too big on the MacBook Pro heatsinks are warranted. Mine had a big old gash going through the middle, with the pits between copper heat pipe and aluminum block base not completely filled. So instead of looking like this:
shagonahc118_det3.jpg

it looks more like this, minus the solid square of copper:
HeatPipe-Pipe-Heat-Sink-2-lg.jpg


This is a design flaw from the start and no amount of polishing can fix that. In fact, any attempt at it would completely break the heat pipe.


As for the amount of thermal grease, the best amount would be a thin line that goes right down the middle of the die. That way, when the grease spreads out, it covers the entire die evenly.
 
I'm also well versed in system building, having run a delidded 4770K bare-die just like the forum post I posted. :)

Currently I'm using Antec Formula 7 on all my machines (desktop CPU, GPU card, laptop). It's just so much better than any stock compound I've come across, and it also seems to outperform AS5 as that used to be my go-to compound.

However, the OP's point that the grooves are too big on the MacBook Pro heatsinks are warranted. Mine had a big old gash going through the middle, with the pits between copper heat pipe and aluminum block base not completely filled. So instead of looking like this:

it looks more like this, minus the solid square of copper:

This is a design flaw from the start and no amount of polishing can fix that. In fact, any attempt at it would completely break the heat pipe.


As for the amount of thermal grease, the best amount would be a thin line that goes right down the middle of the die. That way, when the grease spreads out, it covers the entire die evenly.

I agree on the application technique.

I also think that due to a very low spring force it's probably more important to put less TI because it may not be squeezed away efficiently (and could result in a heatsink skew)
 
Imac overheating : Long problem

Hey,:apple:

We defenetly know its the video board. Try to ebay the video card and change it. YOu can also try to go to a local computer shop and see if they have the same one, so they can try and swap just to test. Some people charge for that , but if ur going to have them repair it at times they do not charge for the check.

WE fix a lot of macs, and the real problem is since they are all in one heat is really the problem. SOme people have drilled holes in the back, but the problem with that is if you do not cover it with vents sheet you will get dust in it.

Also applying thermal paste to the new vide card when you get it, is the right thing to do. As well as applying new thermal paste to the CPU. I have added a link to a video we made on this topic and how it can be fixed. Tell us what you think in the comments.

Link 1 of 3
http://youtu.be/mRUfboAki5o

Link 2 of 3
http://youtu.be/dliRrEzEh2w

Link 3 of 3
http://youtu.be/ZJ9CPMWaeJk


Artic Sylver 5 for the cpu, Some have also not used artic sylver 5 but its 99% sylver. The video GPU i was told to use the white cpu. NOt sure how accurate this is, but its worth reading and asking about.. HOpe this works .
Goodluck,
:)
 
Last edited:
Agreed.

People freak out about temperatures so much after coming from desktops, not knowing that other laptops (non-Apple) actually have it much worse.

Look up idle temperatures on an Alienware M17x. And if that isn't enough to scare you already, look at load temperatures.

Then go back and compare that to the same readings on a MacBook Pro 17".

That is pretty scary.
 
Oh, the irony... it hurts. ;)
There's no irony. Unlike some, I've never claimed or pretended to be an expert. I do, however, have basic observation skills and a healthy dose of common sense. If you had quoted my post in context, you would see that I, unlike some, am not pretending to know more than Apple about their computers.
 
Oh, the irony... it hurts. ;)

I elected not to come back with an 'I told you so'... well, I suppose I just did ;)

It's still quite a few months off, but I'm working on building an electropolishing machine, and intend to start developing specialized laptop cooling systems.

I've successfully built my first home-grown heatpiped heatsink, so that's my start ;)

As bill-p mentioned, this is not just an issue with Apple systems, but with all major manufacturers.

Prodo123 - sorry for re-entering the discussion so late, but I wanted to state that your application technique is spot on, but your statement that my statement is false is, in itself, false ;) Why? On desktops the amount of clamping force is sufficient to completely expel excess thermal compound, but laptops do NOT have the same level of clamping force, unfortunately. That said, I believe it better to err on the side of caution, and use only as small of an amount as necessary to achieve 100% coverage of the die.
 
  • Like
Reactions: Jubadub
Just applied new thermal grease to my Mid 2009 MPB 15" C2D 2.8ghz.
With stock grease cpu temp went up to 105°C (the mac played chess against himself :D)
After applying Arctic MX-4 grease, temps are down between 10-15 °C
Max temp is now 93-95°C & light browsing is around 53°C :)
 
On desktops the amount of clamping force is sufficient to completely expel excess thermal compound, but laptops do NOT have the same level of clamping force, unfortunately. That said, I believe it better to err on the side of caution, and use only as small of an amount as necessary to achieve 100% coverage of the die.

That is true if one relies only on the puny metal clamps, but that is not how TIM is applied in laptops. One should (1) apply the TIM, even if it's a thicker layer, then (2) clamp the heatsink on using the laptop's clamp, and then (3) apply generous pressure on top of the heatsink -- granted, one should do this when the motherboard sits on a soft cloth or rubber mat, not when it's screwed in the laptop case (though to be honest I had plenty of success both ways). I use stuff like this: http://j.mp/1aBU2yf
 
Can MacBook Pro run cooler?
Yes.How?Make the cooling system more aggressive or bulkier for better airflow.
Do we want that?
No,we don't want either,we want to work in peace and we want it light and beautiful.
Thermal paste is not make a significant difference either way in massive production is not possible for every piece to carefully apply high quality thermal paste.
 
Last edited:
Are the latest MacBooks Pro (such as Late 2015) still hot?

I'm planning to upgrade from MacBook Pro Mid 2010, which is uncomfortably hot if discrete GPU is used. Also, due to overheat over the course of the time I had to repair it twice. And now the GPU is dying. Also, the OS X is crashing even on integrated graphics card under CPU load.

I was always jealous to MacBook Air which is really cold compared to Pro. My dream laptop would be MBP 15' without discreet GPU (do people play games on their macbooks, why is it needed anyway?).
 
Last edited:
You can always go to the Apple store on one of the running Macbook Pros and download my tool from https://github.com/qnxor/macoh/releases then run a x264 test there on the spot.

See this thread and also the readme page for more details and comparison graphs to know what you should epxect
Hi qxnor-

Coming into this discussion late. Your work looks pretty awesome. Thanks for making this test.

Just to make certain - you are strictly measuring CPU frequency and temp, correct? So even when you are running a test that is meant to tax the GPU, you ARE NOT displaying GPU frequency and temp, correct? You are looking at the GPU's heat contribution to the CPU and subsequent throttling.

Are you aware of a test or app in the Mac environment that shows GPU frequency? I am trying to determine if my GPU is throttling on my 2015 15" MBP.
 
How come it includes expensive GPU which takes up space, drains the battery, makes extra noise by having dedicated fan and produces extra heat for no real value to me?
The one that he linked to DOES NOT has a discrete GPU, just the integrated Iris Pro. The entry level 15" MBPs do not have the discrete R9 M370X GPUs.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.