The Official Thermal Interface Material thread

[Borc]

This is exactly how you measure thermal conductivity. What is the rate of BTU/hr across the paste between 2 copper surfaces. You only need to know the thickness of the paste. You know the wattage input. You know copper thermal properties. We know the conductivity of water and air. You can calculate exactly how many BTU you NEED to move, that's the easy part. My 13700K pulls 235w at a full load. That's about 802BTU/hr, just a little more than half of my window air conditioning unit I chill loops with.

Did you watch the video from Gamers Nexus? Are you aware the number can be "tweaked" with unrealistic high temps? You need a bit more than what you are telling us. And how can you know the number isn't simply made up. It's obvious that most big manufacturers like Shin-Etsu, Honeywell, Deepcool, LairD, Dowsil, Akasa are in a moderate range of 5-8 W/mK. Could it be they are more honest and don't use inflationary numbers just for marketing reasons.

And to make it a little more complicated for you thermal impedance/resistance is more telling for most of our use cases but this number can be tweaked as well.

The most important specification is definitely the thermal impedance, which is measured in degrees Kin2 / W. This is an application specific measure of the ratio
the temperature difference between two mating surfaces to the steady-state heat flow through these surfaces. Due to the additional mounting pressure and the size of the area, the
thermal impedance typically decreases while increasing with TIM thickness.

The ability of a material to conduct heat regardless of its thickness is called thermal conductivity and is measured in W/mK. While TIMs can be compared to thermal conductivity values, this value does not indicate how good the material's ability is at minimizing contact resistance.

Choosing the right thermal interface material: paste, pad or foil

Knowing the needs of your application is key to selecting the correct thermal interface materials.

www.mueller-ahlhorn.com

There are two critical thermal performance characteristics:
Thermal Conductivity (TC) and Thermal Resistance (TR). In low bond line applications, thermal resistance dominates performance.
In high bond line applications, thermal conductivity dominates performance. In medium bond line there is a blended influence.

Thermal Interface Materials (TIM)

Nordson EFD thermal interface materials (TIM) provide reliable heat transfer over a longer time than most industrial thermal compounds.

www.nordson.com

 

[NoiseBox]

So, I'm not stranger to pasting CPUs and repasting them, but in the past I've used Arctic Silver 5 and called it good to go. Based on some websites I've read, AS% has been superseded by better pastes, and specifically Thermal Grizzly Hydronaut. I bought some off Amazon (a whopping 1 gram) and upgraded my CPU on my 2013 Mac Pro, from a E5-1620 v2 to a E5-2667 v2. Same TDP of 130watts, but even at idle the CPU sits at between 38-43 (with a Fan Control speed of 1300, max on these things is 1900) and it'll ramp up to 80 degrees with even modest games (minecraft with Sildur's Lite shaders)

So the question I have, is Hydronaut not that great, or did I somehow re-paste incorrectly? If you've ever seen a teardown of a 2013 Mac Pro (the infamous Trash Can Mac) it's pretty impossible to screw up reassembly and have the system work. Did I get a bad CPU, it was used after all, unsure if such a thing is possible, given the system passes stress tests and Intel's own Processor Diag Tool.

The system is running Windows 10 LTSC via bootcamp.

 

[freeagent]

I have heard it pumps out at 80c, which is the main reason why I never tried it.

 

[oiawshboiwsahnoidaws]

As long as a thermal paste is "good enough" there will be little difference between them at most 2 or 3°C from my testing
More likely the new (beefier) cpu just uses more power

 

[Macro Device]

@TipsyL @lexluthermiester @jayjr1105

Spoiler: PTM7950 has been tested...

View attachment 321149

...and these results are -4C on the junction and -2C on the hotspot. I heard that PTM7950 gets better over time so I'll play around and watch metrics again next week. Gonna lower the pressure (190 → 160 W) for now, 90s don't look so pleasant.

Spoiler: Nothing really changed. At least it didn't get worse.

 

[wNotyarD]

Spoiler: Nothing really changed. At least it didn't get worse.

View attachment 325092

Not getting worse is actually good news. Let's see in another couple of months if it stays stable.

 

[Macro Device]

Not getting worse is actually good news.

In all fairness, I overclocked the GPU harder (new driver feels more stable) and got ~10 W more on average. Feels like I'm only limited by my 90 mm fans as they can't be faster than 1950 RPM and they don't really provide much CFM. The GPU was cooler under the previous 120 mm 2000 RPM ones.

 

[ShrimpBrime]

Did you watch the video from Gamers Nexus? Are you aware the number can be "tweaked" with unrealistic high temps? You need a bit more than what you are telling us. And how can you know the number isn't simply made up. It's obvious that most big manufacturers like Shin-Etsu, Honeywell, Deepcool, LairD, Dowsil, Akasa are in a moderate range of 5-8 W/mK. Could it be they are more honest and don't use inflationary numbers just for marketing reasons.

And to make it a little more complicated for you thermal impedance/resistance is more telling for most of our use cases but this number can be tweaked as well.

Choosing the right thermal interface material: paste, pad or foil

Knowing the needs of your application is key to selecting the correct thermal interface materials.

www.mueller-ahlhorn.com

Thermal Interface Materials (TIM)

Nordson EFD thermal interface materials (TIM) provide reliable heat transfer over a longer time than most industrial thermal compounds.

www.nordson.com

Though I appreciate the effort, this does not describe the methods manufacturers use for testing.

And no, I don't watch Linus's rival.

Nor do I watch DerBauers exotic car videos.... I mean his PC videos. Well not on a regular basis.

Was at EOCF when DerBauer was newbie. The shaving razor trick you see in his videos is something I was using back in 08' long before a delid tool was invented.

But to stipulate that manufacturers CAN and DO lie, blatantly for sure, I do not deny.

I'm looking for evidence that paste manufacturers test and the methods they use, or rather it was argued they don't at all and just throw us a number. Which is possible.

So far, we are not proving that manufacturers are not testing pastes......

 

[80-watt Hamster]

Though I appreciate the effort, this does not describe the methods manufacturers use for testing.

And no, I don't watch Linus's rival.

Nor do I watch DerBauers exotic car videos.... I mean his PC videos. Well not on a regular basis.

Was at EOCF when DerBauer was newbie. The shaving razor trick you see in his videos is something I was using back in 08' long before a delid tool was invented.

But to stipulate that manufacturers CAN and DO lie, blatantly for sure, I do not deny.

I'm looking for evidence that paste manufacturers test and the methods they use, or rather it was argued they don't at all and just throw us a number. Which is possible.

So far, we are not proving that manufacturers are not testing pastes......

No one's claiming manufacturers aren't testing, but rather that published results between manufacturers are not directly comparable because the conditions under which the results are obtained are not published. It's like UTQG for tires: you can be confident a 700 Pirelli tire will last longer than a 600 Pirelli, but not necessarily a 600 Michelin. And that's with a published and mandated standard.

 

[ShrimpBrime]

No one's claiming manufacturers aren't testing, but rather that published results between manufacturers are not directly comparable because the conditions under which the results are obtained are not published. It's like UTQG for tires: you can be confident a 700 Pirelli tire will last longer than a 600 Pirelli, but not necessarily a 600 Michelin. And that's with a published and mandated standard.

So what you're saying..... is the Corsair TM30 paste claiming 3.5w/mk is a misrepresentation of what could be outstanding AND better than grizzly paste but it might not last as long??? Wait what??

OK, here's what I am saying.

They test their paste between two copper surfaces.
Testing with X amount of Wattage.
Measuring the time it takes to heat up.
KNOWING the already calculated w/mk of the main ingredient of the paste.
Can take a fairly accurate measurement and advertise it.

I honestly don't think there's any other method of testing outside the intended use of Electronics Thermal Interface Materials.

Paste w/mk
pad w/mk
Ketchup w/mk.

"" quote "" Most pastes are about the same "" End Quote ""

No they ain't - Now I dip.
Peace!

 

[80-watt Hamster]

So what you're saying..... is the Corsair TM30 paste claiming 3.5w/mk is a misrepresentation of what could be outstanding AND better than grizzly paste but it might not last as long??? Wait what??

OK, here's what I am saying.

They test their paste between two copper surfaces.
Testing with X amount of Wattage.
Measuring the time it takes to heat up.
KNOWING the already calculated w/mk of the main ingredient of the paste.
Can take a fairly accurate measurement and advertise it.

I honestly don't think there's any other method of testing outside the intended use of Electronics Thermal Interface Materials.

Paste w/mk
pad w/mk
Ketchup w/mk.

"" quote "" Most pastes are about the same "" End Quote ""

No they ain't - Now I dip.
Peace!

That is not what I am saying; it was merely an analogy. To spell it out directly, we can be confident that a 4.0 W/mK Corsair compound has superior thermal transfer to a 3.5 W/mK Corsair compound, but a 4.0 W/mK TG product could be equal, better or worse. Thermodynamics is complicated, and different test criteria can easily provide different results.

 

[Shrek]

I suspect the metal to thermal grease interface plays a role.

 

[ShrimpBrime]

That is not what I am saying; it was merely an analogy. To spell it out directly, we can be confident that a 4.0 W/mK Corsair compound has superior thermal transfer to a 3.5 W/mK Corsair compound, but a 4.0 W/mK TG product could be equal, better or worse. Thermodynamics is complicated, and different test criteria can easily provide different results.

Thermodynamics is one of the many things humans have mastered.

So 4w/mk from both manufacturers should produce very similar results. And so the saying goes... all pastes are about the same....

If that saying is true, then it doesn't matter how the measurement was taken. Or the number provided doesn't matter either. And it was never tested.

That's what science is about though. Taking a measurement that gives results.

I'm fairly certain that 4w/mk would have a small margin of error between 2 pastes, but both are going to perform the same within that margin of error.

Not saying that it is not possible to have 2 pastes say the same w/mk but perform different, it just doesn't seem to happen like you are saying. I don't see people exclaim a 4w/mk paste performs better than another paste. What happens, people come and show pastes advertising a much higher w/mk, they have used and tested it for themselves against lower w/mk pastes and found the higher number works better.

 

[Shrek]

Thermodynamics is one of the many things humans have mastered.

So 4w/mk from both manufacturers should produce very similar results. And so the saying goes... all pastes are about the same....

Thickness plays a very important role.

 

[ShrimpBrime]

Thickness plays a very important role.

Yes actually it does. Part of the equation certainly.

So as I had stated back there.

Best to lap cpu and cooler and use liquid metal. It would be the equivalent to soldering the cooler to the cpu.

Lapping to reduce space and use a smaller amount of thermal interface material.

The suggested amount is a pea sized drop. That's because 2 surfaces that are not flat require that much to ensure proper fill of spaces. Be it a concave IHS plate or a bumped cooler surface.

 

[Macro Device]

a pea sized drop.

I'm not 100% certain how big peas are where you live but in Russia, one pea worth of thermal paste is enough to serve about 3 and a half CPUs. Way too excessive for one and totally too inefficient if used on an average GPU (6700 XT/3070).

Not a huge fan of liquid metal (yet had no other option with my CPU) since it can short circuit the snot out of your components but yeah, it's darn efficient.

 

[ShrimpBrime]

View attachment 325184

I'm not 100% certain how big peas are where you live but in Russia, one pea worth of thermal paste is enough to serve about 3 and a half CPUs. Way too excessive for one and totally too inefficient if used on an average GPU (6700 XT/3070).

Not a huge fan of liquid metal (yet had no other option with my CPU) since it can short circuit the snot out of your components but yeah, it's darn efficient.

Haha, well I can't say it's an accurate description for everyone, but we could say a 2mm ball of paste instead.

Used 25mm wheel spacers on my mk4 jetta.

Watch all my temps in celsius....

But normally, I use a dial indicator to measure liner protrusion which measures thousands of an inch. Engine guys know what I'm talking about...

 

[Count von Schwalbe]

Best to lap cpu and cooler and use liquid metal.

I suspect the person a couple of pages back that had bad results with thermal sheet would do well to follow this advice. Thermal sheet or even nothing at all can outperform the best of pastes if the surfaces are sufficiently smooth, though this is in practice very difficult to achieve.



For those arguing/discussing, this can be a good read: https://en.wikipedia.org/wiki/Thermal_conductivity_and_resistivity#Influencing_factors

Remember that pastes are a mixture of solid particles (metals and non-metals), liquids, and dissolved substances. My understanding of the K-value was that it was primarily for homogeneous substances; if so, then attempting to market a thermal paste based on its K-value (expressed in W/(M*K)) is at best misleading.

 

[ShrimpBrime]

I suspect the person a couple of pages back that had bad results with thermal sheet would do well to follow this advice. Thermal sheet or even nothing at all can outperform the best of pastes if the surfaces are sufficiently smooth, though this is in practice very difficult to achieve.



For those arguing/discussing, this can be a good read: https://en.wikipedia.org/wiki/Thermal_conductivity_and_resistivity#Influencing_factors

Remember that pastes are a mixture of solid particles (metals and non-metals), liquids, and dissolved substances. My understanding of the K-value was that it was primarily for homogeneous substances; if so, then attempting to market a thermal paste based on its K-value (expressed in W/(M*K)) is at best misleading.

True.

Would (should?) be stated as a coefficient of thermal conductivity. W/mk.

We aren't even touching on expansion or anything yet.

Does your mount get tighter after a good burn in?? It just might!

 

[damric]

Finally used up my last tube of mx-4. Was it really from 2019? I don't remember but the other tube before that didn't say 2019.

I use the MX stuff for most general purpose projects and save my expensive TIM for my main rig. Alas I ran out of MX-4 so now I ordered an 8g tube of MX-6. That stuff was cheap too.

 

[Falkentyne]

I've been using JLJ TP-133 thermal paste between my AMD CPU & block and it seems to be working fine. Very affordable in 6G & up quantities too.

13.9w/mk. That's really high for a paste! Grizzly I think is 12 or more. Nice. Might have to order me a tube! Yes, I think I'm going to!

Um....
Anyone notice anything similar between:

Amazon.com

and
https://www.amazon.com/JLJ-Compound-Processor-Performance-Interface/dp/B09MHVN3G2 ?

.......

Is it me or is that the exact same company?

 

[ShrimpBrime]

Um....
Anyone notice anything similar between:

Amazon.com

and
https://www.amazon.com/JLJ-Compound-Processor-Performance-Interface/dp/B09MHVN3G2 ?

.......

Is it me or is that the exact same company?

I don't see it??

Manufacturer is different, make and models are different. What is the same?

 

[Macro Device]

Noticed dangerously high VRAM temperatures and decided to find out what's up with the thermal pads. One was misaligned and ultimately fixed, yet that screwed up my thermal paste on the die.

Found liquid metal and decided to go ham on GPU cooling. Results are... questionable.

GPU: RX 6700 XT, XFX SWFT309, 1900 RPM 90 mm case fans instead of too worn out stock fans.
Mean wattage: 190 W (probably about 240 W in total if we count all components of my video card, y'know these RDNA2 readings).
Ambient temperature: unknown, but tried to keep as much of the same as possible. Probably +15...17C (59...63F).

Way before, semi-noname cheapo paste: 79C junction, 100C hotspot.
Before, PTM7950: 76C junction, 94C hotspot.
Now, no-name liquid metal: 72C junction, 88C hotspot.

These readings were recorded after 30 minute long Cyberpunk 2077 gaming sessions at 4K High with FSR at Quality mode.

 

[lexluthermiester]

Um....
Anyone notice anything similar between:

Amazon.com

and
https://www.amazon.com/JLJ-Compound-Processor-Performance-Interface/dp/B09MHVN3G2 ?

.......

Is it me or is that the exact same company?

Might be a sister company or a copy-cat company. Given the descriptions, it could go either way.

I don't see it??

Manufacturer is different, make and models are different. What is the same?

Take a closer look.

 

[68Olds]

At the very least, the two companies are using the same marketing/advertising company.