Cannot get a good repaste on 2014 15" MacBook Pro

[Flyview]

Hi!

I have re-pasted my late 2013 logic board's CPU and GPU numerous times over the years. After finding out about PTM7950 thermal "pads" I switched over to that because it's supposed to last much longer and not need repasting every 6 months to a year. It worked great and I was very happy with it.

A month ago I swapped logic boards between my mom's (ratty) mid 2014 15" MBP and my late 2013 15" MBP, to get 200MHz more CPU *facepalm*. To be fair, my 2013 logic board would also freeze during sleep in a humid environment.

After swapping logic boards, I cannot for the life of me get a "good" repaste on the 2014" logic board. The temperature of some of the cores is up to 15c higher than some of the others (with 100% CPU load on all cores). I went through 2 sets of PTM7950 and 2 repastes with Arctic MX-6. I thought maybe the heatsink from the 2014 was bent or cracked so I swapped that over from the old 2013 board as well, and it's still the same. The temps on my 2013 board, now in my mom's laptop, are fine!

Has anyone seen this before? Is there a trick to screwing down the heatsink evenly on these (older) machines?

 

[Flyview]

The temps get progressively hotter from core 1 which seems good up to core 3 and 4 which are very bad. Maybe there is not enough heatsink pressure on one side of the chip. Can someone point me to a diagram for the core layout of this Haswell Intel core i7 chip (i7-4870HQ)?

 

[Frieg]

that's the Haswell die map oc-haswell-die-map.jpg

but it's normal that not all cores behave the same thermally, some cpu's are better in that way, some are worse. and it's also normal that the cores near graphics area are hotter than the ones farer from it.

 

[Flyview]

that's the Haswell die map oc-haswell-die-map.jpg

but it's normal that not all cores behave the same thermally, some cpu's are better in that way, some are worse. and it's also normal that the cores near graphics area are hotter than the ones farer from it.

Thanks, I found that diagram as well but it doesn't really match what I see. On the CPU side (I also have a dedicated GPU), there are 2 separate dies, a larger square, and a smaller rectangle. I always assumed the square was the CPU and the smaller rectangle is the integrated GPU. Is it the other way around?

Yes I would imagine there can be some variability, but right after repasting the difference was 3c - 4c. After a few weeks it had jumped to a difference of 15-16c!

I actually just repasted it again a couple of days ago with a thicker amount than I usually would to see if that helps. The difference between the coldest core and the hottest started at 3c. Yesterday it was 5c. Today it's up to 7-8c difference. I'll try pasting very thin next.

I did realize I didn't swap over the 2 little springy bars that clamp down the CPU heatsink from my 2013 logic board. I looked at them closely and one has slightly more curvature to it than the other. Not sure if this could be the reason for what I'm seeing. I might try bending them to be the same or have more or less curvature next. I can't get to the 2013 laptop right now to see what those look like in comparison to these.

 

[Frieg]

The smaller square is the platform controller hub PCH chip. don't put paste on that one. the larger die is the CPU with the integrated GPU

 

[Flyview]

The smaller square is the platform controller hub PCH chip. don't put paste on that one. the larger die is the CPU with the integrated GPU

Oh interesting. They're both under the copper heatsink though.

 

[Flyview]

I re-applied as thin as I could without seeing through to the shiny dies. The difference is back at 2-3c. That's the best I've seen it on this board. The temps seem a couple of degrees lower overall as well. I am not hopeful that this will last but will keep the thread updated with what I find in the coming days and weeks. I do intend to go back to the PTM7950 so I hope I can get the core temperatures to be somewhat close with that.

Combined with undervolting, some thermal pads between the heatsink and case, and a notebook cooler, I am able to maintain the max 3.5GHz indefinitely at CPU package temp of 96c - 97c at 26c ambient in the "Endurance" app's quad core CPU stress test. Prime95 gets much hotter and starts to throttle. To be fair, if I limit the CPU package power to 44W in order to keep the CPU freq at 3.3GHz like my old board, then the temps drop about 10c and is about on par with what I was seeing for that board. The faster CPU pulls 51.5W, that's about 8W more power and heat for 200MHz!

 

[f54da]

what idle temps are you seeing?

 

[Frieg]

Just anecdotal, that's what my 13 inch 2015 MBP is able to maintain in Cinebench R23. That's PTM7950 after 2 months.
With some pastes I was able to maintain below 80°C! But only for a week and then temperatures quickly rose up to 100°C and throttling. for me PTM7950 is the best choice atm because it prevents throttling in the long term even if it doesn't perform the best in Max temperatures. but even the ptm7950 shows some pump out on my system and a negative trend in temperatures over time. I've started with 84°C when applied new.
i have a delta of 8°C between both cores.