Changed Thermal Paste in 15" rMBP

[kelebra]

I wanted to share my experience with changing the thermal paste in my 15" Macbook Pro Retina. Using iStat Menus (currently 4.03) I've been closely monitoring my temps over the past few months. My "CPU Die - Digital" temp idles at around 50 C. This is the steady state temp, after a couple of hours, when there is little or no activity. When I use any app which pushes the CPU utilization to 100%+ my temp will go to around 102 C with the expected increase in fan RPM (around 3600 RPM). Given that the ivy bridge core is designed with a max temp of 105 C I decided to see if I could improve the temps by changing the thermal paste.

Picture 1 shows the screen capture of my typical idle temp. Picture 2 and 3 shows the thermal paste application by Apple. After removing the heatsink I cleaned the factory paste off the die and heatsink with ArtiClean. I applied a thin layer of Artic MX-4 using the controversial "spreading" method (picture 5). Personally I don't believe that the application technique makes that much of a difference.

My new thermal temps are now around 42 - 44 C at idle (compared to 50). The biggest difference I see is under load. With CPU utilization maxed out to 100%+ my temps are now stable around 80 C with the fan RPM around 2300. This is using the Macbook in the same ambient room temp where I had previously had temps go to 102 C).

Quite honestly I don't think that the temp difference is due to just one factor. There are likely to be multiple things which contributed to the higher temps of the factory paste:

1) As expected they used too much paste. This alone isn't the going to kill the temps since much of the paste just get squeezed out by the pressure of the heatsink. However...

2) The screws on my heatsink were not tightened properly. They were too loose. This contributes to a less than perfect mating between the heatsink and die

3) The quality of the thermal paste probably made a difference. If you read all the benchmarks on the Internet you see that the different pastes have a different performance. However, they are generally close.

I believe that all three factors above contributed to higher temps from the factory. Regardless, I'm satisfied with the new thermals as my 3D rendering apps no longer seem to be pushing my Macbook to the edge.

BTW, changing the paste is pretty easy. You can reference iFixit but basically you just remove the bottom plate and you have access to the heatsink (don't forget to disconnect the battery). Just unscrew the standard torx screws securing the heatsink and you're all set. I'm sure someone will bring up the warranty but quite frankly I don't care what Apple thinks I should or should not be doing to my own $4000 laptop.

 

[Brian Y]

I usually change mine as soon as I get the computer, and then every year or so.

 

[madsci954]

Doesn't this void the warranty?

 

[krravi]

I usually change mine as soon as I get the computer, and then every year or so.

Just curious? Why every year?

 

[Brian Y]

Just curious? Why every year?

Just habit - every year I take them apart and clean them out properly (i.e. dust behind the MLB, clean out fans, etc) - as part of that process I replace the thermal paste. When I got my rMBP, I changed the crappy paste they put on at the factory with AS, and it dropped about 5 degrees.

 

[GermanyChris]

Replacing the thermal paste on Apple portables is too often necessary. The last 2 bought 1 new 1 used it was the first step before first boot. There is always to much and it's never quality. This time I used liquid metal I have idle temps in the high 30's loaded in the 70's.

 

[duervo]

Doesn't this void the warranty?

Damage done as a result of the modification is not covered by warranty.

Here are a few examples:

Breaking the logic board, because too much torque was used when screwing the heat sinks back on, would be something that would not be covered by warranty.

If the CPU or GPU failed as a result of the modification, it would not be covered under warranty.

If the hard drive failed after the modification, it will be covered under warranty, as there's little chance that its failure would be caused by re-application of thermal paste.

If a stick of RAM went bad after the modification, it would be covered under warranty.

If the SuperDrive (i.e.: in non-Retina MBP's) failed after the modification, it would be covered under warranty.

If the display failed after the modification, it would be covered under warranty.

 

[kelebra]

That seems within the ballpark of what I'm seeing after re-pasting. It's hard to compare CPU / GPU temps from different people. What people sometimes forgot is that the ambient temp makes a huge difference to the CPU temp. When I use my laptop outside, where it's cooler, I also currently have idle temps in the high thirties. That's why I was careful to measure my before and after temps, posted above, in the same room.

When benchmarking temps it's also important to note the RPM of the fans. Also keep in mind that the diode used to measure the CPU temp is not perfect so it's hard to compare absolute temps between die. I've been designing x86 processors for over 15 years and have never seen ISSCC papers for a truly accurate methodology for measuring die temps.

This time I used liquid metal I have idle temps in the high 30's loaded in the 70's.

 

[GermanyChris]

That seems within the ballpark of what I'm seeing after re-pasting. It's hard to compare CPU / GPU temps from different people. What people sometimes forgot is that the ambient temp makes a huge difference to the CPU temp. When I use my laptop outside, where it's cooler, I also currently have idle temps in the high thirties. That's why I was careful to measure my before and after temps, posted above, in the same room.

When benchmarking temps it's also important to note the RPM of the fans. Also keep in mind that the diode used to measure the CPU temp is not perfect so it's hard to compare absolute temps between die. I've been designing x86 processors for over 15 years and have never seen ISSCC papers for a truly accurate methodology for measuring die temps.

I just use SMC fan control, in the end utmost accuracy is not my concern just keeping heat out of the case is. The difference between 35-36 and 37-38 in the end don't mean much the difference between 37-38 and 48-50 do. Liquid Metal also keeps my 8 core MP (5355's) at 42-45 at 800% over hours on end..

 

[Asifasim22]

Hey I just had a few questions. What year is your rMBP? And how long have you had it? The original paste was so sloppy. Also how much roughly would you say this improved your battery life? Also thanks for posting I didn't know how easily available this was to do on the retinas! And one last question do you think if I complained the apple store would reapply this for me? If mine was still under warranty

 

[alphaod]

Hey I just had a few questions. What year is your rMBP? And how long have you had it? The original paste was so sloppy. Also how much roughly would you say this improved your battery life? Also thanks for posting I didn't know how easily available this was to do on the retinas! And one last question do you think if I complained the apple store would reapply this for me? If mine was still under warranty

There is only one model of the rMBP and how would this improve your battery life? You'll have as much luck cleaning the battery connectors as this to improve your battery life.

 

[B...]

There is only one model of the rMBP and how would this improve your battery life? You'll have as much luck cleaning the battery connectors as this to improve your battery life.

Well, there are 2012 and 2013 models.

 

[utech]

There is only one model of the rMBP and how would this improve your battery life? You'll have as much luck cleaning the battery connectors as this to improve your battery life.

Well, there are 2012 and 2013 models.

And cooler temps means less fan usage, resulting in less battery consumption.

 

[Asifasim22]

And cooler temps means less fan usage, resulting in less battery consumption.

Exactly what I was thinking , I just wanted to see how much it changed it. I didn't think it was this easy to do I've never opened up a retina after I heard how eveything was hard to get to I assumed things like this would be extemely difficult, but I guess it's accessible on both models.

 

[Queen6]

Exactly what I was thinking , I just wanted to see how much it changed it. I didn't think it was this easy to do I've never opened up a retina after I heard how eveything was hard to get to I assumed things like this would be extemely difficult, but I guess it's accessible on both models.

15" Retina is the most accessible by far, as the heatsync is downward facing just need to remove the base plate. As for the fans, a very few minutes at best as the fans are designed to be highly efficient.

 

[fatlardo]

Can you tell me what screwdriver size you need you need?

 

[throAU]

1) As expected they used too much paste. This alone isn't the going to kill the temps since much of the paste just get squeezed out by the pressure of the heatsink. However...

Open to discussion...

Too much thermal paste will result in thermal paste around the edges of your CPU die.

Thermal paste is a heat condutor. It will thus be transferring (some) heat there, and not into your heat sink, thus keeping your CPU hotter around the die edge (?).

I'm guessing 2300 vs 3600 rpm fan speed is a significant difference in noise and at 80c you won't be seeing any thermal throttling

 

[dusk007]

There is only one model of the rMBP and how would this improve your battery life? You'll have as much luck cleaning the battery connectors as this to improve your battery life.

Actually many desktop cooler tests show that the temp of the a cpu or gpu chip under full load does affect power consumption quite a bit.

It won't be a huge difference as it mostly is felt at load and with the kind of difference a thermal paste swap would achieve I doubt it is more than 3% but still. It can be up to 10% of total system power with a big enough temp delta.
http://forums.anandtech.com/showthread.php?t=2200205

GPU cooler tests show similar results as between reference designs that aim for 95C+ and aftermarket stuff that are much better there is often a quite significant temp delta.

 

[bkribbs]

Can they even tell if you do this? I'm fairly technical (I've completely dissembled old laptops and done everything but thermal paste). I've seen my computer get really hot and wouldn't mind doing this but don't want to regret it later.

And if they can tell, is the only result basically that I claim responsibility for if the CPU fails?

 

[Ichabod.]

Can they even tell if you do this? I'm fairly technical (I've completely dissembled old laptops and done everything but thermal paste). I've seen my computer get really hot and wouldn't mind doing this but don't want to regret it later.

And if they can tell, is the only result basically that I claim responsibility for if the CPU fails?

If you replace the compound with something else (e.g., arctic silver 5) and apply only a thin layer, it is pretty obvious that it isn't the work of the factory technicians. To actually check this they have to pull off the heat sink anyway (which they probably don't want to do).

Depending on the model, you often have to pull out the entire logic board, so you may be liable for more than just the CPU.

I just replaced the thermal compound on three of my MacBooks with fairly good success.(only forgot one heatpipe thermocouple hookup)

Good luck, it's fun to do.

 

[bkribbs]

If you replace the compound with something else (e.g., arctic silver 5) and apply only a thin layer, it is pretty obvious that it isn't the work of the factory technicians. To actually check this they have to pull off the heat sink anyway (which they probably don't want to do).

Depending on the model, you often have to pull out the entire logic board, so you may be liable for more than just the CPU.

I just replaced the thermal compound on three of my MacBooks with fairly good success.(only forgot one heatpipe thermocouple hookup)

Good luck, it's fun to do.

On rMBP its just remove bottom case, then the heatsink is right there. Not hard at all.

Just not sure what sort of problem to expect from apple if I go back in to have it serviced.

 

[dusk007]

I doubt any. If they don't deliberately check for it, they won't notice. I doubt they would expect it, as 99.9% of people would never even think about doing such a thing.

 

[bkribbs]

Ok then I'll probably do this! How much paste does everyone recommend? Ive seen to use a piece of rice size, BB size, and the arctic guide seems to say a good bit more.

 

[sofianito]

I think you should apply a thin layer over all the chip surface...

 

[dastinger]

I swapped mine using Arctic Cooling MX2 and didn't spread it. On the CPU, I used a small pea sized bit of compound and on the GPU a bit more than a grain of rice. Then I just put the heatsink on top and that's it. The heatsink does the spreading job. Got 5-10ºC difference results.

On a 15" rMBP btw.