The Official Thermal Interface Material thread

[KLiKzg]

Here is something I never quite understood; look at the temperature drop for a typical power, area, thickness and thermal conductivity

View attachment 360503
P = 100W
A = 3 cm^2
k = 20 W/(m.K)
Dx = 1/20 mm

How then can thermal pastes differ by say 2 to 3°C degrees?

To prepare for this answer, I started reading some other topics here:

• https://www.techpowerup.com/forums/...opper-ihs-upgrades-for-intel-amd-cpus.292436/
• https://www.techpowerup.com/forums/...delids-the-amd-ryzen-5-2600-processor.243627/

But also, have been working on some other topics on my work for thermodynamics. So to answer your questions: Yes & No.
Why?

1. The deltaX is supposed to be usually thicker then 0,05mm & that thin is ideal. But we should tend to be within 0,1~0,2mm, which is optimal from a thickness perspective, as it would give us the sweet spot of not going too thick with paste (as it is far lower conductive material - will explain later) & not too shallow to have "air gaps".
2. 1/k is not calculated this way, but like this:
   
   . Which means k is not equal to lambda!
   If you do calculate if only for this, you will get only conductance of the paste itself. Which is not good at all & you can't get any info from that!
   
   
3. Also, as you do not need 1/alphas, as you do not have transfer to fluids in case of transfer of heat between CPU & Cooler. Actually, you do - but you do not need to know how to do that on side of the cooler base! Why? As most of temperature is conducted by vacuum tubes, not on the top of the cooler based (Cu heat exchanger).
4. So, you get a formula such as this:
   
   ; where there is a thickness of CPU IHS lid coming into question with d1 with it's Cu based Ni plated material (which lowers lambda from ~380 to ~315W/mK), where you put your paste with 4~13W/mK conduciveness & you put on that cooler with its base usually Cu heatsink with thickness of around 5~6mm.

Why you do run the formula, you get something like 3K difference between CPU IHS (which is not same as temp. on CPU die) & Cooler.

In order to compare the 4W/mK, 6W/mK & 11W/mK pastes on the same calculation with 0,1mm thickness of paste you get deltaT with 4,8K, 3,9K & 3K.
If you do the same comparison of 4W/mK, 6W/mK & 11W/mK pastes on calculation with achievable 0,2mm thickness of paste you get deltaT with 7,6K, 5,7K & 4K.

So that is where you get your difference in temp., especially if you put more paste.
Why?
Well he Ni plated Cu based CPU IHS lid has conduciveness of ~315W/mK, while Cooler usually has Cu based ~380W/mK. So the paste with 4~11W/mK acts more like a resistance, then a transmitter of temperature. But it is, how it is - maybe we will find better suiting materials in the future?!

Also, do note that Ag (silver) has the far most temperature conduciveness of ~415W/mK. But probably it is too costly to make CPU IHS or CPU Coolers from Ag...

Hope this fairly engineering explanations gives you more hint about the issues with changes in temperatures between pastes.

 

[Noci]

Also, do note that Ag (silver) has the far most temperature conduciveness of ~415W/mK. But probably it is too costly to make CPU IHS or CPU Coolers from Ag...

They do exist and are in this price range: https://shop.aquacomputer.de/Water-...T-VARIO-1700-slate-grey-925-silver::4226.html

But check out this video of Roman Hartung (der8auer) why it does not work in reality as calculated/predicted in theory.

 

[KLiKzg]

They do exist and are in this price range: https://shop.aquacomputer.de/Water-...T-VARIO-1700-slate-grey-925-silver::4226.html

But check out this video of Roman Hartung (der8auer) why it does not work in reality as calculated/predicted in theory.

Thanks for the link of silver water cooler!

About the videos, here is what is wrong:

1. intel & AMD has Cu IHS lid, but it is Ni plated...& Ni has ~58W/mK, so the CPU IHS lid has about thermal conductivity ~315W/mK.
2. Cu conductivity is not ~400W/mK, but ~380 +/-10 W/mK! You can't just add 20W/mK of the top of your head.
3. Combining a thermal & electrical conductivity with Wiedermann-Franz law is needed for deltaT larger then 100K...while we have about deltaT of about 60~70K top, so it is negligible!
4. He has confirmed that Ni plating has lower conductivity then 100% Cu, while grinded CPU IHS & Ag plate has about larger conductivity then normal.
5. Also, Ni is also put same as Ag - for surface protection of CPU IHS! He did not mention that at all....
6. Water-block is .925 Ag, which is the problem...water-block would need to be .99 Ag to have better conductivity!!!

Conclusion: he proved all the science still works...it is the "marketing" who is making a false advertising!

2nd conclusion:

• If you really want to have better Cooler, check the Cu purity in it. Better suppliers will have it mentioned!
• If you really plan to get Cu CPU IHS, please DON'T. Just take some find grinding paper & grind of the Ni protection (only top layer) from your CPU. This will give you Cu-paste-Cu advantage, for only 2~5€ cost!

 

[Noci]

Thanks for the link of silver water cooler!

About the videos, here is what is wrong:

1. intel & AMD has Cu IHS lid, but it is Ni plated...& Ni has ~58W/mK, so the CPU IHS lid has about thermal conductivity ~315W/mK.
2. Cu conductivity is not ~400W/mK, but ~380 +/-10 W/mK! You can't just add 20W/mK of the top of your head.
3. Combining a thermal & electrical conductivity with Wiedermann-Franz law is needed for deltaT larger then 100K...while we have about deltaT of about 60~70K top, so it is negligible!
4. He has confirmed that Ni plating has lower conductivity then 100% Cu, while grinded CPU IHS & Ag plate has about larger conductivity then normal.
5. Also, Ni is also put same as Ag - for surface protection of CPU IHS! He did not mention that at all....
6. Water-block is .925 Ag, which is the problem...water-block would need to be .99 Ag to have better conductivity!!!

Conclusion: he proved all the science still works...it is the "marketing" who is making a false advertising!

2nd conclusion:

• If you really want to have better Cooler, check the Cu purity in it. Better suppliers will have it mentioned!
• If you really plan to get Cu CPU IHS, please DON'T. Just take some find grinding paper & grind of the Ni protection (only top layer) from your CPU. This will give you Cu-paste-Cu advantage, for only 2~5€ cost!

I won't argue the science aspects and material properties as they are given values and comply to the laws of physics. Atm I don't have the energy to dive back into the theory but assume you did your homework . Despite der8auer, according to you, using the wrong laws/formula's thus method in the end it's the actual measured temps in the chips that matters for pc users.

All these fabuleous thermal conductivity numbers of pure metals or even alloys are of no use if you use a louzy TIM, or a good one but apply it wrongly.
Then you create a thermal barrier, that screws up the efficiency of your perfectely picked cooling parts.
There is no way, maybe outside a specialized materials lab, you can have a perfect direct metal-to-metal or metal-to-die connection for optimal thermal conductivity without the need of a TIM.

So considering TIM being the weakest link of the chain, that is where you can gain the most, actually the very reason of existance of this thread (and quite some other simular ones).
That is also why you see the cooling products manufacterers putting the most effort in improving the TIM's and uniform mounting of cold plates and less in applied metals and other aspects.

Little nuance to your advise to sand the Ni off the IHS; as been proven in many practical tests the effect is neglectible but one constant is that the warranty of the chip is void .

There is also a lot of info (and BS) out there about lapping of IHS/cold plates, it can improve thermal conductivity but is not for everyone.
Better check the pro's & con's before even blindly start with it.

 

[Shrek]

Also, do note that Ag (silver) has the far most temperature conduciveness of ~415W/mK. But probably it is too costly to make CPU IHS or CPU Coolers from Ag...

1g is about $1 so quite reasonable, at least in a thermal paste.

 

[lexluthermiester]

Also, do note that Ag (silver) has the far most temperature conduciveness of ~415W/mK. But probably it is too costly to make CPU IHS or CPU Coolers from Ag...

Diamond is better, but let's not quibble.

Little nuance to your advise to sand the Ni off the IHS; as been proven in many practical tests the effect is neglectible but one constant is that the warranty of the chip is void .

This. Unless someone is going to do extreme OCing, modding one's CPU is a wasted effort. And then the only real solution is proper delidding and block pairing. For that kind of thing, one either goes all in or it's best not to bother.

1g is about $1 so quite reasonable.

This is why Arctic Silver 5 is still being made and is still very effective as a TIM.

 

[R-T-B]

Got a 10g syringe of Noctua NT-H2 bundled with components I bought recently. Nothing to complain and what I've read, Noctua's paste is more than fine.

Honestly I never used anything much more than a big tube of MX-2 on most builds until recently. I just feel the gains to be had here in standard pastes are often overplayed. But longevity etc? That sort of stuff is somewhat interesting to me, so worth looking at at least.

This is why Arctic Silver 5 is still being made and is still very effective as a TIM.

I think that stuff is fine too, just always felt it was overpriced for what you got. But well, silver and all that.

 

[kanmer]

why is the toplamp thermal paste not available to purchase anymore? are they about to release a new product?

 

[Athlonite]

I think that stuff is fine too, just always felt it was overpriced for what you got. But well, silver and all that.

Yeah it maybe a little overpriced for what you get but it's worth it for it's longevity I've had AS5 last 10 years and still be good

 

[Ruru]

Honestly I never used anything much more than a big tube of MX-2 on most builds until recently. I just feel the gains to be had here in standard pastes are often overplayed. But longevity etc? That sort of stuff is somewhat interesting to me, so worth looking at at least.

Actually my friend said that the tube was lying for ages so he just gave it to me. Though still seems like new so I guess there's no problem with longetivity.

 

[R-T-B]

Actually my friend said that the tube was lying for ages so he just gave it to me. Though still seems like new so I guess there's no problem with longetivity.

I've heard the Noctua stuff is pretty good for shelf life, though tbh, most paste is if you just stir it a bit.

By longevity I mean more post-application, like how many thermal cycles it can take before performance degrades. THAT is a relevant stat to me. Some may not care as much but it does kind of interest me. That's where these new PTM7950 honeywell pads are supposed to really shine.

 

[lexluthermiester]

I've heard the Noctua stuff is pretty good for shelf life, though tbh, most paste is if you just stir it a bit.

NT-H2? Yes. NT-H1? Not some much. Once applied, the stuff dries out fast(2 to 3 months) and loses a good 10% of it effectiveness, which is entirely unacceptable. I have seen and measured this personally. I won't touch it again. If you're going to go with a Noctua TIM, go with NT-H2 and avoid the first one like the plague.

 

[R-T-B]

NT-H2? Yes. NT-H1? Not some much. Once applied, the stuff dries out fast(2 to 3 months) and loses a good 10% of it effectiveness, which is entirely unacceptable. I have seen and measured this personally. I won't touch it again. If you're going to go with a Noctua TIM, go with NT-H2 and avoid the first one like the plague.

Well I was commenting exclusively on shelf life in the first paragraph, but yeah that "application longevity" (I guess that'd be the term?) would be an issue for sure.

 

[freeagent]

Is NT-H1 like creamy mashed potaters or is that Ceramique that I am thinking of

 

[KLiKzg]

I won't argue the science aspects and material properties as they are given values and comply to the laws of physics. Atm I don't have the energy to dive back into the theory but assume you did your homework . Despite der8auer, according to you, using the wrong laws/formula's thus method in the end it's the actual measured temps in the chips that matters for pc users.

All these fabuleous thermal conductivity numbers of pure metals or even alloys are of no use if you use a louzy TIM, or a good one but apply it wrongly.
Then you create a thermal barrier, that screws up the efficiency of your perfectely picked cooling parts.
There is no way, maybe outside a specialized materials lab, you can have a perfect direct metal-to-metal or metal-to-die connection for optimal thermal conductivity without the need of a TIM.

So considering TIM being the weakest link of the chain, that is where you can gain the most, actually the very reason of existance of this thread (and quite some other simular ones).
That is also why you see the cooling products manufacterers putting the most effort in improving the TIM's and uniform mounting of cold plates and less in applied metals and other aspects.

Little nuance to your advise to sand the Ni off the IHS; as been proven in many practical tests the effect is neglectible but one constant is that the warranty of the chip is void .

There is also a lot of info (and BS) out there about lapping of IHS/cold plates, it can improve thermal conductivity but is not for everyone.
Better check the pro's & con's before even blindly start with it.

1st of all, the calculus is right & give out outputs as it should, according to physical law. Not theory, a law!

It also proves that you can have lousy numbers or 1~2°C temperature more on CPU, if you use TIM which has about ~5W/mK. So it is advisable to use ~10W/mK or more.
& yes, I call TIM a "heat barrier", as it has 1/30 the amount of thermal transfer from CPU IHS to Cooler. The science will go forward & we will have better materials (TIMs)...

& yes, we can improve the thermal management with using better TIM. While we hope, as @lexluthermiester said, some day use something far better conductive materials. As diamonds are still taken by girls for their rings. Maybe some other sort of material, as Graphite is of no use for this application.

My suggestion still stands about getting that Ni of the CPU IHS to have better thermal conductivity! Warranty is something that people need to check for themselves & is applicable only on single person level, as it differs from case to case. Yes, people need to check their warranty, but many of us use "older computers" without warranty - so no harm in that!

About the lapping: will it increase the surface area of CPU IHS that much, that it would decrease the negative effects of thicker lapping material?
That is the main question here. As the physical law clearly states that using thicker CPU IHS is negative, compared to thinner CPU IHS - which also makes one question a big one: why is AMD making so much thicker CPU IHS, compared to Intel? Especially considering that it has negative results within physical law.

 

[lexluthermiester]

Is NT-H1 like creamy mashed potaters or is that Ceramique that I am thinking of

You might be thinking of Ceramique. NT-H1 goes on fairly smooth.

 

[stahlhart]

Got a 10g syringe of Noctua NT-H2 bundled with components I bought recently. Nothing to complain and what I've read, Noctua's paste is more than fine.

I'm having really good luck with this on CPUs lately as well, since I don't delid. I used to prefer Gelid GC Extreme for GPU dies, but they're all under PTM7950 now, mainly to reduce maintenance downtime.

 

[Veseleil]

makes one question a big one: why is AMD making so much thicker CPU IHS, compared to Intel?

Two apparent reasons. (Damn) Compatibility with AM4 cooler mounts, and "we don't wanna be Intel, bending IHS..."

 

[Noci]

& yes, we can improve the thermal management with using better TIM. While we hope, as @lexluthermiester said, some day use something far better conductive materials. As diamonds are still taken by girls for their rings. Maybe some other sort of material, as Graphite is of no use for this application.

There is a form of Carbon that works, the manufactering & application techniques are just not mature enough.
No doubt scientists will find some groundbreaking cooling applications in the (near) future.

Just Google on Graphene.

That's why I'm willing to give Thermal Grizzly KryoSheet a try, the results of several practical tests look promising.

About the lapping: will it increase the surface area of CPU IHS that much, that it would decrease the negative effects of thicker lapping material?

I have to correct you here, it decreases the surface size, if you go from a 'rough' surfaces to a perfectely flat surface . But different than heat dissipiating surface as in cooling fins to air, here the main factor is thermal conductivity. What you do increase (only when perfectely lapped and no effect from convex or concave variables) is, the contact surface for better thermal conditivity but after applying a TIM, even a good one the gain in efficiency is only marginal.

That few micron thickness layer of (porous) Nickel has a neglectible effect in practical scenario's. Theoretical you're right., but it's not worth the effort and risks for the average user.

 

[Nordic]

That's why I'm willing to give Thermal Grizzly KryoSheet a try, the results of several practical tests look promising.

I have tested MX-4, kryonaut extreme, PTM7950, and kryosheet on my 6750xt. The kryosheet has worked best achieving the highest benchmark score with Furmark. Kryosheet also worked well on my CPU but everything was within 2c of each other.

 

[Shrek]

At 2°C is it worth the trouble, when long term endurance is more important.

 

[Nordic]

At 2°C is it worth the trouble, when long term endurance is more important.

Is 2c worth the cost of buying multiple pastes and effort to apply them? Absolutely not. I am enjoying this process though. My next test is with TFX which Igor showed having top tier thermal performance but was really hard to apply.

 

[Shrek]

Fun is a good enough motive.

 

[lexluthermiester]

Fun is a good enough motive.

It really is. Tinkering is as much of a hobby as anything else can be. Tech tinkering can be as much great fun as it is interesting!

 

[KLiKzg]

Two apparent reasons. (Damn) Compatibility with AM4 cooler mounts, and "we don't wanna be Intel, bending IHS..."

I understand the compatibility issue...but not sure about the "we don't wanna be Intel, bending IHS...", like there is an issue with that? Once you put it in, you do not remove it. It is a CPU, not a pecker.

But maybe AMD has some other issues with mounting. (I am still an intel fan!)

There is a form of Carbon that works, the manufactering & application techniques are just not mature enough.
No doubt scientists will find some groundbreaking cooling applications in the (near) future.

Just Google on Graphene.

That's why I'm willing to give Thermal Grizzly KryoSheet a try, the results of several practical tests look promising.

Well, not sure...as Graphene I have looked about & found it is electrically conductive (so that is a no go) & also is very thermally polarized in its conduciveness (more conductive in line with the sheet or straws, less conductive perpendicular to sheets or straws). So although this is a good example & it is an interesting material, it can't solve anything!

Like it can't also solve Space elevator also, but that is another story.

I have to correct you here, it decreases the surface size, if you go from a 'rough' surfaces to a perfectely flat surface . But different than heat dissipiating surface as in cooling fins to air, here the main factor is thermal conductivity. What you do increase (only when perfectely lapped and no effect from convex or concave variables) is, the contact surface for better thermal conditivity but after applying a TIM, even a good one the gain in efficiency is only marginal.

That few micron thickness layer of (porous) Nickel has a neglectible effect in practical scenario's. Theoretical you're right., but it's not worth the effort and risks for the average user.

You are right, polishing will decrease the surface area. So please, do not polish the Cu, after you get through Ni! As that would negatively effect with thermal conductivity with lesser etta (emission coefficient).

But I do need to differ, as Ni has thermal conductivity (later TC) of ~58W/mK, Cu has TC of about ~380W/mK, while CPU IHS has TC of ~315W/mK.
Why has CPU IHS so much less TC then Cu? Well, because of the Ni plating, which has about 6~7 times less TC.
So putting pure Cu CPU IHS will get you something of more TC, but it is better to scrape down the Ni & to have rougher surface (not polished).
Why? Again, you can so more damage with replacing CPU IHS, then using fine sanding paper & getting rid of that Ni from the top.