HOWTO: Replace thermal paste (Late 2013 Retina 15" DG)

[JPamplin]

Folks,

If you have a Retina Macbook Pro that's a few years old, replacing the thermal paste is one of the best, easiest, cheapest and quickest upgrades you can do. It's the best $10 I ever spent on this thing. I got the thermal paste pictured below for $10 on Amazon.

See below - the original thermal paste was in awful condition (way too much / all over the place, but had caked and lost contact over a lot of the dies), and when CPU load ramped up, the fans would spool up to top speed (the sound was noticeable). CPU temps got over 200 degrees F, and the GPU got well into the 168-175F range.

Fortunately, the heatsink is on the bottom of the logic board - only a handful of T5 screws and it pops right out.

Specs: 15" Late 2013 Retina Macbook Pro 2.6 i7 dual graphics (nVidia 750M) 16/500 - this is the highest end model from late 2013, but this procedure should work for any Retina with variations on the heatsink type.

Here's the procedure:

• Shut down your Mac and unplug it
• Take the bottom plate off (you'll need a Pentalobe 5 (P5) screwdriver)
• Undo the battery connector (center top of battery)
• Unscrew the heatsink screws (one on each end, then 4 around the CPU and 4 around the GPU) with a T5 screwdriver
• Lift the heatsink out on the RIGHT SIDE first (the left side is under a metal edge)
• Clean the CPU/GPU dies and heatsink with a paper towel with rubbing alcohol - clean the black cover around the dies as well
• Use some compressed air to clean out the fans and heatsink fins
• Put a SMALL amount of new thermal paste on each die
• Spread the paste with plastic wrap over your finger evenly
• Reinsert the heatsink carefully on the left side first, then attach the edge screws, then the screws around the CPU and GPU. Tighten around each chip evenly, but you can screw each down all the way.
• Put the bottom plate back on and you're done!

Temps didn't go down THAT much, but OMG THE FANS! SO much more quiet doing the same type of thing. Love love love it. FYI: My fans had been blown out recently, so they weren't that bad. Idle CPU/GPU temps are 97-101F, and up to 200/165 under load.

Launching VMWare Fusion with Windows 10 build 1903 would spool up the fans to 100% - now it's literally half that and so quiet. This whole thing took 15 minutes.

 

[JPamplin]

Thanks for all the views, y'all! Let me know how it works out for you.

 

[uller6]

Nice pictures and glad it worked out! I just changed the thermal paste on my old 2009 iMac and 2012 MacBook Pro. Same result: the fans are way quieter than before the paste replacement. Both are like whole new machines. IMO this is one of the best things you can do to resuscitate an aging Mac.

 

[eleafz]

Thanks for this tutorial!
I've seen some people just put a drop on there and saw you shouldn't spread it before putting on the heatsink, but I see here you have pre-spread it. Any idea which is best? Going to do this next on my 2014 i7 with nvidia 750m gpu.

Also on another note: Do you run High Sierra on the macbook pro with nvidia 750m?
I'm still on Sierra but wondering if it is safe to upgrade to High Sierra and still have good nvidia/CUDA gpu driver performance? Adobe says mac os mojave doesn't support cuda for Adobe apps.

 

[Frieg]

It depends on the thermal paste, good thermal paste has always instructions on the back on how to apply it correctly.

For MX-4 OP actually did it wrong, you are supposed to apply a at best perfectly round drop exactly at the center of the chip surface. The reason behind this is, that the paste gets spread out from the center to the edges when the heatsink is mounted and pressure applied, that way no bubble of air can geht caught in the paste.

if you do it like OP did there will a very high amount of microscopic small air bubbles be caught in the thermal paste, because the paste that's already there off-center will prevent the air from being pressed out. it won't do any harm, but you will miss a few degrees that you could have achieved.

OP did it in every way wrong, but hey, it worked.

 

[maflynn]

For MX-4 OP actually did it wrong, you are supposed to apply a at best perfectly round drop exactly at the center of the chip surface.

Actually it's almost impossible to apply thermal paste incorrectly and there really isn't a wrong way (other then woefully under-applying)

 

[JPamplin]

OP did it in every way wrong, but hey, it worked.

I appreciate your comments! No opinion is unwelcome here, but as maflynn pointed out, all methods generally have the same results as long as you use enough.

So why do I prefer the "pre-spread" method? Simple: you can't really trust that people will put a consistent, correct amount of thermal paste in the center every time. At least with pre-spreading, you can SEE that every bit of the die is covered evenly.

So perhaps my approach is to make it easier to replicate - and the results speak for themselves. There's not going to be huge differences in temps anyway with a mobile CPU/GPU, the real advantage is the ability to transfer heat more efficiently to the heatsink - making for FAR quieter fans. It really helps.

 

[JPamplin]

Also on another note: Do you run High Sierra on the macbook pro with nvidia 750m?
I'm still on Sierra but wondering if it is safe to upgrade to High Sierra and still have good nvidia/CUDA gpu driver performance? Adobe says mac os mojave doesn't support cuda for Adobe apps

I'm on Mojave, and I don't use the professional Adobe apps, so I had not heard of the issues with CUDA, sorry. However, if I were you, I would upgrade to High Sierra for the Apple File System support, plus all the new features including Airplay 2 in 10.13.6.

According to this page on nVidia's site, the 418.163 driver has CUDA and 10.13 support - and this was released on May 10, 2019. So give it a shot and let us know!

JP

 

[eleafz]

I went ahead a replaced the thermal paste and cleaned out the dust, wow the fans run much less noticeably. However I am noticing cpu core 1 is running cooler than the other 3 at full load, is that normal?
At idle each core is about the same somewhere between ~43-50C

Under full extensive load, cores 2-4 top out at 100C, while core 1 never seems to max much above 90C. Maybe it is just the internal design of the chip that the furthest away core runs a bit cooler than the ones next to each other? or.. the sensor isn't accurate?

ex: I just ran the 'prime95' stress test tool, cores 2,3,4 hovered around 98C while core 1 stayed at about 88C.

I used the vertical line method, for each die, as that was what is recommended on the arctic silver site, but after reading the arctic silver instructions again online, they say use a 1mm line. My line must have been at least about 1cm not 1mm (but it still ended up being a slightly less than what is shown in the photos on this post)...
With the metal clip thing on the left where the heat sink has to clip under, I wasn't able to go directly down with the heatsink but instead from left to right, could that spread it unevenly?

Or maybe this is typical behavior.
Ambient air temps here are about 85-90F this time of year.

I ran geekbench4 and the single core/multicore results on this mac seemed to be slightly above the average scores on there for this processor.


Processor: MacBook Pro (15-inch Retina Mid 2014) Intel Core i7-4870HQ 2500 MHz (4 cores)

 

[JPamplin]

So a thermal compound expert (who is an owner at Thermal Grizzly) stated that the bubble issue is not an issue (fast forward to 11:32):

 

[JPamplin]

Yet another validation of the importance of replacing thermal paste on older Macbooks:

 

[_Kiki_]

for model 2016+ you need to remove whole logic board, because the screws for heatsink are on different side, procedure is more difficult

 

[TheSl0th]

Reviving this ancient thread in case anyone is watching and has ideas/knowledge.

One of the things that doesn't come in these reapplication threads is the question of what paste to use given the extremely high temperatures that Apple runs their laptop processors at.

Most reviews and tests of thermal paste are in systems where the CPU temps don't get much above 80ºC, whereas Apple lets their laptop CPUs get close to 100º. This matters because thermal pastes dry out and degrade much faster with higher temperatures.

I can't find much info on what paste would be best suited for a MBP in terms of longevity and not just performance. I'd happily trade a little thermal performance for something that will last and have consistent performance over time. This is one of the reasons Apple use the much-maligned stuff they do - it has to run indefinitely with consistent performance characteristics over time, and (reasonably IMO) compromises thermal performance to do so.

If anyone has any knowledge or experience here, it would be appreciated.

 

[JPamplin]

> If anyone has any knowledge or experience here, it would be appreciated.

As the original poster, I’m happy to assist, @TheSl0th

I agree, Apple’s algorithm tends to favor noise over temperature. So they let things get up to redline before they start cooling. With that in mind, I prefer Arctic MX-5 - it has a very different consistency from MX-4 (which is still good), and it appears that MX-5 won’t dry out as fast.

But I’m no expert - just speaking from experience.

 

[TheSl0th]

> If anyone has any knowledge or experience here, it would be appreciated.

As the original poster, I’m happy to assist, @TheSl0th

I agree, Apple’s algorithm tends to favor noise over temperature. So they let things get up to redline before they start cooling. With that in mind, I prefer Arctic MX-5 - it has a very different consistency from MX-4 (which is still good), and it appears that MX-5 won’t dry out as fast.

But I’m no expert - just speaking from experience.

I ended up using Honeywell phase change thermal pads because I don't want to have to think about redoing it in the future!

But just to mention my experience, changing the paste temporarily to Noctua NHT-2 (needed to perform other surgery before I had the thermals pads), and then the Honeywell phase-change pad had more or less no effect on temperatures, performance or clocks over stock on my 2015 2.8ghz. Seems it was doing just fine with Apple's 'bad' paste. In a Cinebench stress test the all core clock does jump up a tiny bit higher to 3.2 (from 3.1) momentarily but it's for such short duration that it's not a meaningful change from the stock at 3.1 steady.

I suspect a lot of the benefit people get from these 'mods' is that they end up cleaning out the fans and ducts through the heatsink and end up with much better airflow.