CPU Cooling Adventures - (fixing) a delid gone wrong

[silkstone]

This thread is a continuation of the following: https://www.techpowerup.com/forums/threads/lapping-sanding-ihs-and-waterblock.258162/#post-4096571

For TLDR, scroll to bottom.

So after lapping the IHS on my 8700k and waterblock, I'd decided that I really wanted to delid too as the temperatures weren't quite what I wanted.

I ordered my LM Ultra, as well as a delidding kit from AliExpress.

My LM came within a couple of days, but even with the fastest shipping method, the delidding kit would take a least a couple of weeks

Being the impatient SOB I am, I also bought a $.5 pack of thin Gillette razors and made the decision that I was going to 'go manual'.

Now most people reading this will be thinking this was where I went wrong and in 99% of stories, you'd be correct. However, buying the razors might have actually saved my CPU.

I've had the 8700k for about a year now, after upgrading from an 8600k. I sold the 8600k locally for a little bit more than what my current CPU cost. My 8700k wasn't new, and I'd gotten a good deal on it from the forums. When I bought it, I'd thought there might be something wrong with it, but made a gamble, received it quickly, slapped it in my PC and saw that it was working fine. The temps on it were a little high, but nothing unusual for this chip so just assumed I'd not been lucky on the silicone lottery and perhaps that's why the seller was selling it relatively cheap.

Well, today, I discovered the reason it was cheap.

I gently went around the IHS with the thin razor, but a couple of the corners were giving me difficulty. I had a closer look at the CPU and noticed what looked to be white superglue on one of the corners. Barely noticeable.
Damn! It looked like someone had already had a go at delidding this processor and it hadn't gone to plan. I take out a bit of acetone and dip a q-tip in it to try and rub off some of the glue before having a last go with the razor.
A little elbow grease and a heart-wrenching quiet 'crack' afterwards and the IHS pops right off. I was planning on soaking it in acetone after this final, gentle attempt and the 'crack' took me by surprise.

This is what I find:

Someone has gone to the trouble of delidding the processor, butting a bit of TIM on there and then trying to superglue it back on, without removing the sealant!
No wonder the temps were so bad on this chip! If you look closely you can see that the Glue is sitting on top of the sealant which has soaked through forming a hard, crusty layer.

I try to remove this with my fingernail, to no avail. With the CPU in this state, there is little point trying to apply LM and resealing it so I need to find a way to get that glue off.

Acetone it is - The nail polish remover type, not industrial strength.

I soak it for 5 minutes - Very little comes off
I soak it for another 10 minutes - A little more comes off, but the glue seems to have bonded and reacted with the sealant that it's formed a new compound which is a lot more hardy than regular glue.
either that, or the nail polish remover I am using is just too weak in concentration.

At this point, I'm a little worried whether the CPU will boot any more and decide to apply the LM and give it a shot.

My original plan was to reseal the CPU with some silicone sealant, but at this point i'm not even sure it'll work so decide to have the IHS 'free-floating' and lock it down with the CPU retention bracket.

I put everything back together, apply a healthy dose of TIM to my AIO and screw everything down tight.

Moment of truth . . . It boots! success I haven't killed my CPU. That's a win in my book. But, what are the temps?

Worse than they were in the first place. Damn! I need to remove some more of that glue-sealant shit on the cpu so disassemble everything. Take out the CPU and clean it up then soak it in acetone again. This time for another 20 mins.
I take it out again and try to clean up more of the glue. A little comes off, but there is still a stubborn layer, around 0.2 mm thick on two of the corners. . . What is this shit made of?

Sod it. Time to risk it all. I decide to soak it in acetone for 1-2 hrs while I go take a nap. When I get back, I grab myself a coffee before heading outside to the acetone soaked CPU hoping that the PCB hasn't started to melt yet.
Take it out and voila! A good chunk of the glue comes off. After another 10 minute soak, I've gotten the final corner more or less clean.

There are still some stubborn areas, but I have a feeling that I'd only be able to remove them with an angle-grinder and really need to test to see whether the CPU will still work.

I reapply the CPU with LM and put everything back in. Damn . . . temps are still high! Now all the guides say to use LM very sparingly and I have, but it's really hard to judge how much is actually on the die with the die being reflective and the LM also being reflective. I saw that there was good contact between the IHS and die before reapplying: each end of the IHS would lift up slightly when pressing on the other.

Time for a little more LM. A bigger blob this time on both the IHS and die along with a bit of electrical tape on the 4 contacts next to the die, just in case. Time to get my PC to boot.

Success!! Finally some good temps.

There are still a couple of cores that are getting quite hot, but this is a win in my book and within expectations for delidding.
I'm not sure whether I will go back at a later date to try reapplying the LM a little better (or adding more) but, this will do for now!

TLDR: Delidded my CPU without knowing someone had done it (badly) before. Got lucky by using a razor, instead of a delidding kit which could have ripped through the PCB and (semi-)successfully cleaned it all up to relid again properly. Not bad for my first attempt at this!

 

[Arctucas]

Have you watched any of the videos on rockitcool about how to apply LM?

 

[silkstone]

Have you watched any of the videos on rockitcool about how to apply LM?

I'm not sure whether I saw the rockitcool ones specifically, but I've watched 4-5 others, all use the same technique. Small amount, grain of rice sort of thing, slowly 'painted' on. This took about double what I've seen before so I'll be going back to watch some more videos.

Edit - I just watched the video - With exception of using tape, that's the amount I used on my first attempt, but found it wasn't nearly enough. Temps were a lot worse with so little LM. Undoubtedly, that last, stubborn, thin layer of glue is having a negative impact on IHS-Die contact, but with the exception of soaking the CPU in acetone overnight, I have no idea how to remove it.

My idle temps are now in the high-30's instead of high 40's and the average temps are between 70-80 with a 4.6 Ghz all core overclock. A couple of cores peak in the 90's when stress testing, but only for a short time and don't cause throttling.

If anyone has any advice on how to remove superglue from cpu die without killing the chip, or has any tips on LM use, I'd really appreciate them.

 

[AsRock]

Well thanks for sharing, it was a interesting read indeed ).

 

[Enterprise24]

Liquid Ultra 2 ?! Anyone know the different between this and original version ?
I know they announce Liquid Extreme for better heat transfer but never heard about Ultra 2 before.

 

[silkstone]

Liquid Ultra 2 ?! Anyone know the different between this and original version ?
I know they announce Liquid Extreme for better heat transfer but never heard about Ultra 2 before.

I'm not sure. I couldn't find a huge amount of info on it myself, but saw it was for sale on Amazon and a few other sites that look legit. It might just be new packaging.

 

[delshay]

The bonding of the LM to both surface is important regardless of which LM end user chooses. Getting the surface tension/bonding right will bring further drop in temperature. This means both surface must be well cleaned. I find rubbing the surface with a soft cloth on both surface for about two mins allows the LM to bond better on die & IHS. This is my method & it may or may not work for other users. I'm sure other users will have other methods in getting the LM to bond with the surface better.

 

[silkstone]

Here's a close-up of my initial LM application. I needed a little more than this the second time to get good contact between the IHS and die. From all the guides I've read, this really should be too much.

 

[dgianstefani]

Conductonaut is the best LM.

 

[TheoneandonlyMrK]

Conductonaught instructions said all that and unusually to do both surfaces (with Tim), this worked well on my vega64.

 

[Solaris17]

Congrats dude, good work!

 

[eidairaman1]

Id sell the chip delidded for LN2 users.

Conductonaught instructions said all that and unusually to do both surfaces (with Tim), this worked well on my vega64.

Use conductonaught on copper only

 

[Sithaer]

That was a pretty good read and something to learn from in a way.

Tbh stuff like this is why I'm really not a fan of used hardware and only do that as a last resort and even then I'm very picky with who I deal with.
My current GPU is used and it was also a miner card for ~6 months,I've suspected that before buying it and when I asked the previous owner he did not try to deny it and since the card was in a fair condition+warranty till 2021 I went along with it.
If he tried to lie abou it or anything 'shady' I would just look for another seller/card.

Some ppl just do ugh 'questionable' things to their hardware and then pretend nothing happened/ghetto repair it and then sell it off.

I also had to learn this the 'hard' way with a used GTX 560 Ti a good few years ago when that card was still a decent one.
Whoever used that card did not even peel off the protective plastic/film off the shroud and it was all stuck on it,almost like melted so the person who sold it to me had to clean it off with some strong stuff.
I bet the card ran under very hot conditions with no proper airflow maybe even OCed too.

That card died at me after barely 1 year 'even tho I tried to take care of it' and it had all kind of temperature and fan issues and since I had no warranty on it it was a thrown out money/cost of that lesson so to say.

 

[Vario]

Could you carefully sand the glue down with a high grit sandpaper?
Your temperatures could still be much lower.

 

[silkstone]

Could you carefully sand the glue down with a high grit sandpaper?
Your temperatures could still be much lower.

I have thought about it, but I can't think of a way to not take off some of the PCB material at the same time. If anything, I think it would be better to sand down the IHS on the bottom edges, but that would expose bare copper which would then need to be masked off.

When running my finger around the edges, it all feels smooth, but I can tell there is still a tiny bit of residue in one corner.

 

[phanbuey]

I'm not sure whether I saw the rockitcool ones specifically, but I've watched 4-5 others, all use the same technique. Small amount, grain of rice sort of thing, slowly 'painted' on. This took about double what I've seen before so I'll be going back to watch some more videos.

Edit - I just watched the video - With exception of using tape, that's the amount I used on my first attempt, but found it wasn't nearly enough. Temps were a lot worse with so little LM. Undoubtedly, that last, stubborn, thin layer of glue is having a negative impact on IHS-Die contact, but with the exception of soaking the CPU in acetone overnight, I have no idea how to remove it.

My idle temps are now in the high-30's instead of high 40's and the average temps are between 70-80 with a 4.6 Ghz all core overclock. A couple of cores peak in the 90's when stress testing, but only for a short time and don't cause throttling.

If anyone has any advice on how to remove superglue from cpu die without killing the chip, or has any tips on LM use, I'd really appreciate them.

A razorblade dipped in https://www.acehardware.com/departm...MI5dH2o7SL5AIV9iCtBh2pZwokEAQYAiABEgJy0_D_BwE

just keep dipping and scraping (carefully) and it will come right off. won't affect silicone.

use a q tip to pre apply on layers that need scraping.

i may have had this same problem before... (i used paint thinner but it's basically the same thing).

 

[Divide Overflow]

I'd stop patronizing whomever sold you that CPU.

 

[biffzinker]

A razorblade dipped in https://www.acehardware.com/departm...MI5dH2o7SL5AIV9iCtBh2pZwokEAQYAiABEgJy0_D_BwE

just keep dipping and scraping (carefully) and it will come right off. won't affect silicone.

use a q tip to pre apply on layers that need scraping.

i may have had this same problem before... (i used paint thinner but it's basically the same thing).

Be extra careful the razor blade doesn't dig/cut into the top PCB layer. In your one photo @silkstone you can easily see the traces.

 

[phanbuey]

yes you scrape it like you would if you were getting a bee stinger out of your skin, not the way as if you were shaving, if that makes sense..

 

[eidairaman1]

Be extra careful the razor blade doesn't dig/cut into the top PCB layer. In your one photo @silkstone you can easily see the traces.

Let alone slices flesh

 

[silkstone]

yes you scrape it like you would if you were getting a bee stinger out of your skin, not the way as if you were shaving, if that makes sense..

I've considered it, but I actually think soaking it in acetone for longer would be safer!

The layer that is left is so thin, and the razors edge so hard that I just think there'd be too much risk of scraping the top layer of PCB too. I may take it out again in a few weeks to evaluate the contact between IHS and Die again, if I can figure out a better way to test the contact.

At the moment my OC is 4.6Ghz on all cores at auto-voltage so I can also work on lowering the volts as a next step. That's tested using AIDA which shows there is 0% throttling going on, I could likely go higher if using a slightly less taxing stress test that better reflects everyday usage.

In all honesty, I don't need high clocks and the risk-reward with mechanically removing the glue would be swerving too far into the area of 'OCD' from 'Enthusiast'. However, I will update if I do get it all off in the future.

I'd stop patronizing whomever sold you that CPU.

Honestly. I'm not too bothered considering the price. It wasn't sold as new and I didn't ask. The temps weren't awful, but it did make my delidding experience a lot more 'interesting'. I'm sure I'd feel differently had I ripped off the glue using a delid tool.

 

[killster1]

That was a pretty good read and something to learn from in a way.

Tbh stuff like this is why I'm really not a fan of used hardware and only do that as a last resort and even then I'm very picky with who I deal with.
My current GPU is used and it was also a miner card for ~6 months,I've suspected that before buying it and when I asked the previous owner he did not try to deny it and since the card was in a fair condition+warranty till 2021 I went along with it.
If he tried to lie abou it or anything 'shady' I would just look for another seller/card.

Some ppl just do ugh 'questionable' things to their hardware and then pretend nothing happened/ghetto repair it and then sell it off.

I also had to learn this the 'hard' way with a used GTX 560 Ti a good few years ago when that card was still a decent one.
Whoever used that card did not even peel off the protective plastic/film off the shroud and it was all stuck on it,almost like melted so the person who sold it to me had to clean it off with some strong stuff.
I bet the card ran under very hot conditions with no proper airflow maybe even OCed too.

That card died at me after barely 1 year 'even tho I tried to take care of it' and it had all kind of temperature and fan issues and since I had no warranty on it it was a thrown out money/cost of that lesson so to say.

true true, i never have considered someone deliding a cpu before i purchased (i try to buy used workstations with the cpu i want and give away the box) , i think this is pretty rare and more likely if you have a problem its because someone is TRYing to commit fraud instead of a mishap delid, (8700k lid on a pentium G or something). i enjoy buying used hardware however gfx cards are a little different after the mine crazy boom. i will buy dead hardware off you if you have it for sale (i have a wall of gfx / mobo's alot of them still work)


and to the OP wow great post very much enjoyed it and im sure we are all curious what the price on the cpu was

 

[Grog6]

As far as removing the residue, I'd get real acetone from the hardware store, and use a q-tip to gently rub the surface.

It should take the silicone and superglue off; if it only takes the superglue, use a blue scotchbrite pad to gently remove the silicone.
It shouldn't damage the epoxy or soldermask on the die carrier.

I'd try to find a silver or solid copper shim the thickness of the heat spreader, lap them and the die carefully, and use the liquid metal to bond that to the die; it will up the heat transfer quite a bit.

There will be more pressure on the chip, so you want to be careful with a heavy cooler.

That's the way I put together my old Barton system, and it overclocked to 2800MHz from 1800MHz after.
2500 was the best I could get before the shimming/gallium.

Looks Good!!

 

[forman313]

Liquid metal can be very tricky. Polished surfaces tend to make it much harder to get it right. Its like pouring water on a geese (norwegian saying). Very poor wetting.

I use 1500 grit SiC paper until matt, on both cooler and die/IHS. Its extra important on copper because of the oxidation, but really helps on dies and nickel plated surfaces also. Its almost like applying paint afterwords.


A picture of my cooler _after_ cleaning, Only 24 hours after installing it.

 

[Sithaer]

true true, i never have considered someone deliding a cpu before i purchased (i try to buy used workstations with the cpu i want and give away the box) , i think this is pretty rare and more likely if you have a problem its because someone is TRYing to commit fraud instead of a mishap delid, (8700k lid on a pentium G or something). i enjoy buying used hardware however gfx cards are a little different after the mine crazy boom. i will buy dead hardware off you if you have it for sale (i have a wall of gfx / mobo's alot of them still work)


and to the OP wow great post very much enjoyed it and im sure we are all curious what the price on the cpu was

I also like to keep dead hardware 'not that I have many' but I do hope that I won't have more from my own PC.

Yea I know that some ppl like used stuff and honestly its a good way to save money,its more like a personal preference that I try to buy new hardware or at least with 1-2 years warranty on them.
Otherwise it bothers me knowing that if something dies in my PC I can't do anything about it other than replacing completely +have to spend money again.

My brother is also one of those who prefers/like used hardware,like his new PC hes building I can't even tell anymore where he bought the parts from.
So far it worked out well for him,barely any dead hardware over the years.