Hi everyone, first post here and I joined to share my personal experience in this forum as it seems probably the most serious around and with a lot of "techy-inside" ones that's never a bad thing. :biggrin:
I'm from Italy and english is not my birth language so, please, forgive me if sometimes I'd mispell something.
To cut the story short i found myself with a very bad 3770k which is unable to do 5Ghz no matter the voltage/core/thread config and was capped at 4.8ghz 1,4v with temps reaching the T-junction and sending the CPU screaming PROCHOT#, furthermore it had core#0 topping at 10 degrees less than core#1 and 7 degrees less than core#2 and core#3, quite unbalanced...
Finding a compromise I settled that 4.4Ghz 1.176v LLC-4 (4c8t stable) could have been a good compromise for a 24/7 setup but temps were still way too high topping for the hottest core at 96 degrees (C°) so I took a sabre delidding the sucker, for some kind of madness I HAD to lap the IHS, but not only on top and at the base, even internally :ninja: and after that I have now this results:
I wasn't sure of the nickel hull's thermal properties but I was sure of one thing, the copper has more than four time better thermal transfer properties than the most refined nickel, furthermore I thought a nickel hull could keep somehow the heat inside preventing all the benefits granted by the copper, I had to get rid of it completely, not only off one side.
I must say it's not been an easy task and it took a lot of patience but with only few things I've been able to keep planarity in the inner side of the IHS, at least enough to make nearly a perfect contact using some liquid metal, I didn't take a pic of the inside lapped IHS but I can always add that the first time I'll feel the need to remount, however I found that the inside of the IHS reflects the outside, the outside of mine was nearly the same as looks Idontcare's one on his lapping thread, on other words it was concave in the inner part, the inside instead has the same amount of convexity and this is the reason that leads me to think at a secondary reason for the choice of using a TIM instead of soldering, easy coupling with less expensive IHS, first reason remains that tri-gate 22nm transistors are not suited to whitstand temps of tipical soldering scenarios, for example the 40s at 260C° that is the thermal tolerance for BGA assembly of the socket on the boards could wreak havock on poors 22nm tiny transistors, however I'm divagating, so lets get back to my exp.
Here's what I used:
Sandpaper from 400 to 1500 grit. (better from 320 to 3000 but I didn't have that and was one of those crazy moments had to do with what I had)
Abrasive paste, silicon abrasive, diamond powder 100000 mesh, polishing non-abrasive fluid for metals.
Contacts deoxidizer, isopropyl alcohol.
A toothpick.
A ruler.
A cutter.
kitchen paper.
Cotton swabs.
A small candle.
An hairdryer.
Water.
Optional and a lot useful items are a little vise (bech or mobile doesn't matter) and a small micrometer a 25mm or 1" one is more than enough.
I simply measured the inner of the IHS which is 26mm or slightly more than one inch, cutted stripes of the sandpapers sheets in that measure, wet them with water and wrapped them onto the toothpick till they were perfectly round and tightly packed, then I started using them onto the inside (with a vise this step is much more confortable), with not too much pressure and starting obviously with the coarser one, trying to keep a chessboard pattern and rotating at fixed steps, ten times back and forth on one side, rotate 90 degrees, rinse and repeat, washing frequently and watching to not overuse the same spot of the roll too much till I could clearly see a square of copper, then 5 times and rotate and so on till only a one time pass for every rotation, rinse and repeat for every other grit.
As a last pass I used the polishing stuff to obtain a mirror-like surface and cleaned things up first with deoxidizer (but every solvent that can get rid of the silicone is good as well) then with IPA.
To check planarity a micrometer could be of some use but I used the candle instead (I didn't have one ATM), painting the inside of the IHS with a very thin layer of its wax and heating it up (with a trivial hairdryer) enough to make it "splashable", then waiting a bit to let it cool down for some seconds and placing the cpu on its place, squeezing the two pieces toghether with the right amount of strenght and waiting a bit more for a further cool down, as a result I observed there was no wax on the IHS where the silicon was pushed on and the borders of the IHS were properly aligned with the PCB, that was enough precision for me since I planned to use liquid metal and the results came as expected.
Long story short there were some risks involved but closed the thing with trivial silicone paste I had a 36C° improvement on the hottest core and for me, forgive me, it's a WTF result thinking that ambient temperature is controlled at 21C° costantly and for pure knowledge I used in the second test a couple of fans that are much less performant (and much less noisy), two CM-Xtraflow120Slim (18-58cfm)instead of two Xigmatek XAF-F1255 (50-90cfm) on the same h80i cooler.
Before doing this I thought forever taking in account every termocoupling issue that could stop me from doing that but I came out with this theory, if the IHS has been left willingly loose due to thermocoupling reason I'm probably gonna kill my cpu BUT...if we take in account that a better dissipation could keep the cpu cooler and being cooler it will have lesser thermal dilatation maybe I'm not gonna "crack" even under load, furthermore it couldn't have an "huge" dilatation and the PCB is somehow resilient and could probably balance few microns of silicon expansion.
I'm only sorry for having a life that prevents me to fill this post with technical graphs, comparison, images and so on, unfortunately I don't have all the time I'd like for this kind of things and I didn't it in a scientific way, I know, but I hope it helps anyway.
Please just one thing left, I told HOW I've done the thing, if someone wishes to replicate my experiment he's doing it AT HIS OWN RISK, incidentally I combined the lazyness to go out having a beer with the stuff I had at home AND the PC running hot and making noise but with the consciousness I could have even end with a cpu to fry and post the video on youtube on one side and the sureness to have enough brain an patience on the other, please keep this in mind.
Best regards.