Wanna delid your Intel CPU? Forget liquid metal, do this

[JoshHendi]

I have an 8086K (w/ Noctua NH-D15) and decided to go with liquid metal and a delid. After a successful delid and LM application, I got better results (though not the 15-25c difference a lot of users reported) and after some more reading, the fact that it can cause some mess and potentially need replacing every couple of years, I opted to try my luck on a different approach.

 

So I popped the lid again, removed the LM and instead used Noctua's NT-H1 for both between the die and IHS, and between the IHS and cooler. The main difference this time was I purchased a $16 pure copper IHS from Amazon and used that instead of the stock Intel. THIS resulted in me getting amazing temp drops, and can rest assured I dont have LM in there waiting to break down and cause potential stains on the IHS and die.

 

Before my max temps on stress tests were hitting 88c, now I cannot get the cpu to touch 70c, 68c is my new high

 

So, just replace the stock IHS with a pure copper one, and you'll still get 15-20c gains. (Plus I get to protect and display my 'super collectors edition' 8086K IHS

 

(Caution, do not use LM on a pure copper IHS)

 

Curious to hear if others have tried the pure copper IHS and their results

 

Edit: Amazon ASIN: B07FFMMMSD

[SavageNeo]

nice find!

[Mister Woof]

Biggest difference isn't so much from the TIM (i mean, it certainly helps) but the distance between IHS and die.

I can't say for sure in your case, but in most reports I read about poor delidding performance, they did not remove the old glue.

 

If you delidded without removing all the old glue, this distance is still too great. Perhaps the copper IHS solved this?

In my case, I did in fact result in a 15-20c reduction in maximum load temperature by removing the old glue and using LM.

[JoshHendi]

1 minute ago, Plutosaurus said:

If you delidded without removing all the old glue, this distance is still too great. Perhaps the copper IHS solved this?

 

I did remove all of the old glue on the first run. Maybe it is possible I didn't use enough LM the first time. 

[Mister Woof]

3 minutes ago, JoshHendi said:

I did remove all of the old glue on the first run. Maybe it is possible I didn't use enough LM the first time. 

LM application can be tricky; i've seen youtubers do it seemingly perfectly and get bad results, do it again, and then its amazing.

 

Maybe the way the LM pools?

[JoshHendi]

I am sticking with my all paste setup. Do not need anymore temp gains and would rather not have to mess with the LM down the road.

 

I'd imagine LM with intel IHS vs paste with pure copper IHS will have close enough results

[Mister Woof]

If its working, i'd leave it alone too

[Flying Sausages]

22 minutes ago, JoshHendi said:

I have an 8086K (w/ Noctua NH-D15) and decided to go with liquid metal and a delid. After a successful delid and LM application, I got better results (though not the 15-25c difference a lot of users reported) and after some more reading, the fact that it can cause some mess and potentially need replacing every couple of years, I opted to try my luck on a different approach.

 

So I popped the lid again, removed the LM and instead used Noctua's NT-H1 for both between the die and IHS, and between the IHS and cooler. The main difference this time was I purchased a $16 pure copper IHS from Amazon and used that instead of the stock Intel. THIS resulted in me getting amazing temp drops, and can rest assured I dont have LM in there waiting to break down and cause potential stains on the IHS and die.

 

Before my max temps on stress tests were hitting 88c, now I cannot get the cpu to touch 70c, 68c is my new high

 

So, just replace the stock IHS with a pure copper one, and you'll still get 15-20c gains. (Plus I get to protect and display my 'super collectors edition' 8086K IHS

 

(Caution, do not use LM on a pure copper IHS)

 

Curious to hear if others have tried the pure copper IHS and their results

 

Edit: Amazon ASIN: B07FFMMMSD

Very good idea and experiment! You also forgot LM is conductive too so if it gets outside the die, your cpu is rip. 

[Mister Woof]

1 minute ago, OlympicAssEater said:

Very good idea and experiment! You also forgot LM is conductive too so if it gets outside the die, your cpu is rip. 

 

Depends on the chip, on the 8086/8700 there's nothing ever there for it to short

[brandishwar]

Given you replaced the IHS with an aftermarket one, I'm wondering if your original IHS was the issue. The aftermarket one is likely guaranteed to be flat, meaning guaranteed to get 99+% contact with your heatsink. I'm curious as to what the contact was like with your original IHS and whether lapping it (wet sanding till flat, for those who don't know the term) would've made a similar difference in temperatures. If the contact surface between the IHS and your cooler isn't all that great, your temperatures are going to suffer.

 

I have an i7-5820k, and I've considered lapping it to see if I can get better temperatures (with my triple-360mm radiator setup) since you can't delid it. I'll be checking how flat it is during my next maintenance round.

[Flying Sausages]

24 minutes ago, Plutosaurus said:

 

Depends on the chip, on the 8086/8700 there's nothing ever there for it to short

If the LM got out of the die than you rip your cpu because 8700 die is no different than any Intel cpu die. 

[Falkentyne]

1 hour ago, Plutosaurus said:

LM application can be tricky; i've seen youtubers do it seemingly perfectly and get bad results, do it again, and then its amazing.

 

Maybe the way the LM pools?

People get erratic LM performance because they try to apply it on polished or mirror surfaces which is terrible.

You need to lightly sand (buff) both the CPU and IHS underside (or whatever surfaces LM is contacting) with 1500 grit dry (2000 may work if dry but I think 1500 is best) sandpaper for best results, then wipe the surfaces clean after (91% alcohol+coffee liners work).  This will create a rougher textured surface while still being flat, and LM loves these surfaces.  The coarseness helps LM spread much easier and remain attached instead of pooling up back to itself when it gets very thinned out or compressed.

[1van]

1 hour ago, JoshHendi said:

... and potentially need replacing every couple of years, I opted to try my luck on a different approach. ...

Copper IHS has been tested many times by trusted sources. It improves the temperatures somewhat comparing to the stock IHS.

Your liquid metal application might have been bad, this is why you haven't got the improvement.

P.S. Liquid metal does not require replacing every couple of years. I suspect you might be selling these copper IHSs

30 minutes ago, Falkentyne said:

People get erratic LM performance because they try to apply it on polished or mirror surfaces which is terrible.

You need to lightly sand (buff) both the CPU and IHS underside

Slightly sanding the IHS ribs (where the stock sealant was applied) might be helpful, because there are burrs sometimes. Sanding the die itself is not required and even dangerous if the person doing this is unskilled.

Proper de-greasing/removing the remains of the paste is enough to make the liquid metal spread nicely on the surface. To avoid spilling you just don't use excessive amount of metal, it has to be a really thin film, no drops etc.

Die lapping is not for everyone)

Edited November 17, 2019 by 1van

[Falkentyne]

Again people dont read what I say.

I am not talking about sanding down the die!

I am talking about buffing the surface with your fingers with 1500 grit sandpaper.

 

This is not sanding.  You are not reducing the height.  And there is ZERO RISK to this.

What you are doing is creating a surface where LM adheres much better.

 

If you experiment with trying to apply LM on a fully polished mirror surface and on a buffed 1500 grit wiped same surface, you will see a huge difference in how it spreads.  And when you spread it very very thin, you will notice exactly what causes temps to suddenly "climb" vs "remaining stable."

 

You need to do the work before you comment on posts like this.

 

You also do the exact same thing (1500 grit) to remove hardened oxidized gallium layers from old LM applications on both CPU and IHS.

[1van]

19 minutes ago, Falkentyne said:

Again people dont read what I say.

I've read what you've said. I'm not saying that you recommended sanding down/lapping here.

I'm saying that people who can't even apply liquid metal properly will do even more harm if they bring sandpaper anywhere near the CPU

 

As I said before, good de-greasing makes it spread very well, the temps don't climb up, too.

Never had to re-apply it though, maybe you have a point.

[Mick Naughty]

I’d rather do direct die with regular tim. If I’m deliding anyway. 

[Falkentyne]

17 minutes ago, 1van said:

I've read what you've said. I'm not saying that you recommended sanding down/lapping here.

I'm saying that people who can't even apply liquid metal properly will do even more harm if they bring sandpaper anywhere near the CPU

 

As I said before, good de-greasing makes it spread very well, the temps don't climb up, too.

Never had to re-apply it though, maybe you have a point.

That I honestly agree with.

I was basically just explaining to @Plutosaurus why the youtubers keep getting different results with 'perfect' applications. And yes he's right, it is about how it 'pools', as well as in what manner pressure is applied.  Buffing the surface helps eliminate this (assuming of course that LM is applied to both IHS and CPU surfaces).  

 

Maybe I'll polish a spare laptop heatsink and show the difference in 'new' (10 strokes) and 'thinned out' LM on both the polished surface and buffed surface, but I'm running out of LM and polishing paper atm.

[JoshHendi]

So my original IHS was ALMOST perfectly flat... A good way to test this is put the IHS/chip on a flat smooth surface (like glass) and try to spin it like a top. If it spins then it isn't perfectly flat and should be lapped/replaced.

 

I didnt want to lap the IHS of my 'special edition' cpu which prompted me to buy the copper, and was the best $16 spent

[Labeled]

5 hours ago, JoshHendi said:

(Caution, do not use LM on a pure copper IHS)

Why? I use it on one I got from rockitcool, while it does stain it, it doesn't affect cooling performance.

[JoshHendi]

On 11/17/2019 at 2:21 PM, Labeled said:

Why? I use it on one I got from rockitcool, while it does stain it, it doesn't affect cooling performance.

For a long term application, IMO this would just be tempting fate. YMMV

[Labeled]

On 11/19/2019 at 3:16 PM, JoshHendi said:

For a long term application, IMO this would just be tempting fate. YMMV

I've used it for over a year now....

[Otto_iii]

neat findings, i like that its more conventional and probably more reliable approach 

[Bitter]

From what I can recall reading about those aftermarket copper IHS is that they are indeed often flatter than the Intel parts are AND they have a shorter 'skirt' around the outside so they sit lower and there's tighter contact to the die and possibly also the CPU rentention lid and CPU cooler mounting pressure also push it tighter to the die. The BEST conductor of heat is the least amount of TIM, be it paste or liquid metal the less gap you're taking up the better the heat transfer. The better the heat transfer the cooler the CPU runs and the less wattage it consumes as well, because some of you more astute overclockers might notice that the hotter a CPU gets the more wattage it consumes, lower watts mean less amps which means more room to pump up the voltage for a more stable OC if you're having a hard time. It's not always just about running cooler for the sake of running cooler.

[Falkentyne]

1 hour ago, Bitter said:

From what I can recall reading about those aftermarket copper IHS is that they are indeed often flatter than the Intel parts are AND they have a shorter 'skirt' around the outside so they sit lower and there's tighter contact to the die and possibly also the CPU rentention lid and CPU cooler mounting pressure also push it tighter to the die. The BEST conductor of heat is the least amount of TIM, be it paste or liquid metal the less gap you're taking up the better the heat transfer. The better the heat transfer the cooler the CPU runs and the less wattage it consumes as well, because some of you more astute overclockers might notice that the hotter a CPU gets the more wattage it consumes, lower watts mean less amps which means more room to pump up the voltage for a more stable OC if you're having a hard time. It's not always just about running cooler for the sake of running cooler.

This is true.

[amdorintel]

On 11/17/2019 at 10:12 AM, JoshHendi said:

Maybe it is possible I didn't use enough LM the first time. 

yea that is probably the case