Is Intel’s Thermal Paste ACTUALLY That Bad?

[Levent]

4 minutes ago, JoostinOnline said:

That's not true.  Just a couple weeks ago I replaced the thermal paste on a 6700k (a three year old CPU) with liquid metal.  There was a 15C difference.

I am not sure how much TP you applied and if you cleaned the plastic gunk off the IHS and the cpu pcb BUT from using a delidded cpu for 5 years, I think you applied too much, with cpu dies you gotta apply A LOT less than you think. (for example I get around 15 degree difference between too much paste and very little paste with AS5)

[JR88]

2 hours ago, JoostinOnline said:

-snip-

15c is pretty significant I think and might be worth it...hell something like Mx4 might be a 5-10 degree difference. 

[dalekphalm]

1 hour ago, JoostinOnline said:

-snip-

Right...

 

Anyway, back on topic - I do find the idea of this video highly interesting. I just think their testing methodology was most definitely not ideal. Perhaps they didn't have enough time to plan the video - or simply did not consider the alternate ways of testing?

[Mira Yurizaki]

I could've sworn that the stuff Intel uses is basically like wax: it's hard until the processor's heat melts it, then it starts spreading around. Having it in some liquid or semi-liquid form doesn't make sense to me when the stuff comes pre-applied and there appears to be no warning about how to keep it stored or handle it. That's assuming they're using the same TIM for the heat spreader and die as they do on the heat sinks

[dalekphalm]

1 minute ago, M.Yurizaki said:

I could've sworn that the stuff Intel uses is basically like wax: it's hard until the processor's heat melts it, then it starts spreading around. Having it in some liquid or semi-liquid form doesn't make sense to me when the stuff comes pre-applied and there appears to be no warning about how to keep it stored or handle it. That's assuming they're using the same TIM for the heat spreader and die as they do on the heat sinks

I think we need to understand that people are using the term "liquid" in a very generous way.

 

The pre-applied TIM on Intel sock heatsinks are, very thick - almost like mechanical grease - wax like I think is a good term, yes.

 

But it's malleable, not dried. If you stick your finger onto the heatsink, you can smudge the stuff around.

 

Also I don't really think it melts once you put the heatsink around. I think it just gets spread a bit more based on the pressure.

[JoostinOnline]

2 minutes ago, M.Yurizaki said:

I could've sworn that the stuff Intel uses is basically like wax: it's hard until the processor's heat melts it, then it starts spreading around. Having it in some liquid or semi-liquid form doesn't make sense to me when the stuff comes pre-applied and there appears to be no warning about how to keep it stored or handle it. That's assuming they're using the same TIM for the heat spreader and die as they do on the heat sinks

The pre-applied stuff is definitely goopy.  I'm not sure about what's under the IHS.

[MageTank]

3 hours ago, The Viking said:

true, I understand that lasting a long time is important too, but I'm sure there's another paste out there that is better than this and last longer too.

 

For the soldering, it's just down to money at this point. Until they actually prove, via some independant research (or observers, whatever really), that soldering their cpus is actually impossible, I'll keep calling them cheapos. AMD does it after all, on much more complex cpus. If tiny, 4 billion income a year AMD can do it, then gigantic 63 billion income Intel is more than capable of doing it.

Shin Etsu has some extremely good blends that are ultra durable AND very good at cooling. EVGA uses them on their GPU's. Having removed my cooler quite a few times, I have noticed absolutely zero difference between the stock Shin Etsu paste and my Gelid GC Extreme. It's quite a bit more expensive than the Dow Corning that Intel uses, which is likely why they don't use it. The Shin Etsu paste is easily just as durable in my opinion, as it handles bare die thermal cycles under heavy GPU loads and doesn't pump out over time.

[Mira Yurizaki]

1 minute ago, dalekphalm said:

I think we need to understand that people are using the term "liquid" in a very generous way.

 

The pre-applied TIM on Intel sock heatsinks are, very thick - almost like mechanical grease - wax like I think is a good term, yes.

 

But it's malleable, not dried. If you stick your finger onto the heatsink, you can smudge the stuff around.

 

Also I don't really think it melts once you put the heatsink around. I think it just gets spread a bit more based on the pressure.

 

Just now, JoostinOnline said:

The pre-applied stuff is definitely goopy.  I'm not sure about what's under the IHS.

Well I bring this up because:

• I'm questioning Linus's reconstitution method (I'm almost certain alcohol would eat away at the binder)
• What's the binder that's used to hold the actual thermal conductive material
• How viable is this binder over time

If the binder used in the TIM was perfectly fine, then trying to "reconstitute" the TIM with alcohol could've made it worse.

[JoostinOnline]

1 minute ago, M.Yurizaki said:

 

Well I bring this up because:

• I'm questioning Linus's reconstitution method (I'm almost certain alcohol would eat away at the binder)
• What's the binder that's used to hold the actual thermal conductive material
• How viable is this binder over time

If the binder used in the TIM was perfectly fine, then trying to "reconstitute" the TIM with alcohol could've made it worse.

Next time I delid a CPU, I'll try to heat up the paste to see what happens.  It did look worse after the alcohol though.  It was more like powder.

[dalekphalm]

1 minute ago, M.Yurizaki said:

 

Well I bring this up because:

• I'm questioning Linus's reconstitution method (I'm almost certain alcohol would eat away at the binder)
• What's the binder that's used to hold the actual thermal conductive material
• How viable is this binder over time

If the binder used in the TIM was perfectly fine, then trying to "reconstitute" the TIM with alcohol could've made it worse.

Oh yes, I agree. I think that using alcohol to "moisten it" probably severely affected the performance of said TIM in a negative way.

 

Furthermore, I suspect that the dried, flaked off TIM, was probably not really re-usable period. I mean, perhaps there's some way to reconstitute it without negative performance hit (someone mentioned using spirits), but I'm skeptical.

[Mira Yurizaki]

9 minutes ago, dalekphalm said:

Furthermore, I suspect that the dried, flaked off TIM, was probably not really re-usable period. I mean, perhaps there's some way to reconstitute it without negative performance hit (someone mentioned using spirits), but I'm skeptical.

The other thing I want is some kind of data from a formal study on the viability of thermal compounds. Not just people's anecdotal evidence on tech forums because they saw a 1-2C improvement by replacing the TIM every year or something.

[JoostinOnline]

13 minutes ago, M.Yurizaki said:

The other thing I want is some kind of data from a formal study on the viability of thermal compounds. Not just people's anecdotal evidence on tech forums because they saw a 1-2C improvement by replacing the TIM every year or something.

I've seen very big differences, but that's when changing at least 5 year old TIM.

[MageTank]

1 hour ago, M.Yurizaki said:

 

Well I bring this up because:

• I'm questioning Linus's reconstitution method (I'm almost certain alcohol would eat away at the binder)
• What's the binder that's used to hold the actual thermal conductive material
• How viable is this binder over time

If the binder used in the TIM was perfectly fine, then trying to "reconstitute" the TIM with alcohol could've made it worse.

I question the reconstitution method mostly because it's completely unnecessary. We've known for years what paste Intel uses. They even have it listed on their website: https://www.intel.com.au/content/www/au/en/support/articles/000005576/processors.html

Quote

Where can I purchase TIM?

You can order TIM replacement kits through:

• Intel Customer Support, part number G15816-001 for the pillow pack
• Retail sites like Newegg* or other online retailers

G15816-001, otherwise known as Dow Corning TC-1996. https://www.amazon.com/TC-1996-Intel-Recommended-D54816-001-G15816-001/dp/B006Q5MTPM

 

This is the same paste they use on their stock coolers as well as the bare die since Socket 2011-0. Prior to that, they used Shin-Etsu G751: https://www.intel.com/content/dam/www/public/us/en/documents/design-guides/core-i7-900-ee-and-desktop-processor-series-lga1366-socket-guide.pdf (Section 6.2.3).

Quote

Thermal Interface Material

A thermal interface material (TIM) provides conductivity between the IHS and heat sink. The reference thermal solution uses Shin-Etsu G751*. The TIM application is 0.25 g, which will be a nominal 26 mm diameter (~1.0 inches).

@LinusTech if you guys would be willing to redo this test with both TC-1996 and SE-G751, I think people would get a much better insight to the differences in the quality of the pastes themselves. You could even do another video in the future showing paste durability under repeated thermal cycling, to see how the tried and true "durable pastes" hold up against the "performance pastes". 

[Mira Yurizaki]

So one thing in researching this led me to this post on Medium: https://medium.com/@OpenSeason/soldered-cpu-vs-cheap-paste-59fb96a4fca7

 

I should probably read it more thoroughly but it brought up some points:

• The reason why Intel went to a TIM instead of soldering wasn't a cost issue. It was a thermal-mechanical stress issue. The primary concern is this:

  Quote

  Due to the the property of the indium based solder, thermal cycling forms microcracks in the solder layer. The crack formation resembles that of a layered separation leaving voids where thermal energy cannot transfer. The crack pattern is referred to as “Delamination.” Over time these cracks contribute to significant increase in overall thermal resistance of the TIM1 layer. While the largest thermal swing occurs at production time, the life cycle of the processor can also contribute to delamination. Thus, a fresh soldered processor BNIB is at its peak thermal performance prior to running hundreds of heating cycles. Larger dies are less prone to stress based delamination at production time.

• Thermal conductivity is only part of the story. In a CFD simulation they did, they found that a TIM with 20W/mK did not perform any better than a TIM with 10W/mK
• They mentioned having low thermal resistance also plays a role. However, thermal resistance is not the same as the reciprocal of thermal conductivity. The reciprocal of thermal resistance is thermal conductance.
  • I don't claim to be an expert in thermal engineering, just pointing things out here.

So basically... it appears a lot of us (myself included) have more to learn than just numbers on a spec sheet. But hey, at least that post shed some light on things.

[JoostinOnline]

46 minutes ago, MageTank said:

@LinusTech if you guys would be willing to redo this test with both TC-1996 and SE-G751, I think people would get a much better insight to the differences in the quality of the pastes themselves. You could even do another video in the future showing paste durability under repeated thermal cycling, to see how the tried and true "durable pastes" hold up against the "performance pastes". 

Don't hold your breath on that.  It would be admitting a rather glaring mistake.  I don't mean to say that Linus has never admitted mistakes, because he seems like a pretty humble guy (especially considering how successful he is), but releasing a more accurate video that totally negated the old one wouldn't be great for their image.  They made the anti-delid video, including every mistake you could make to "prove" that delidding didn't improve temps in a meaningful way, and never made a statement on it (that I'm aware of).

 

It's ironic that they probably spent more time trying to turn 6 year old paste into something usable than the amount of time it would have taken to search for Intel's paste and order it.

[JR88]

5 minutes ago, JoostinOnline said:

Don't hold your breath on that.  It would be admitting a rather glaring mistake.  I don't mean to say that Linus has never admitted mistakes, because he seems like a pretty humble guy (especially considering how successful he is), but releasing a more accurate video that totally negated the old one wouldn't be great for their image.  They made the anti-delid video, including every mistake you could make to "prove" that delidding didn't improve temps in a meaningful way, and never made a statement on it (that I'm aware of).

 

It's ironic that they probably spent more time trying to turn 6 year old paste into something usable than the amount of time it would have taken to search for Intel's paste and order it.

 

SE I hear is suppose to be good stuff, But he always liked to use MX4 and I do like the stuff over AS5. So I put some of that on there. He was all Anti-delid and a lot of the same reason's a chance People tearing up there cpu's. 

[MageTank]

20 minutes ago, JoostinOnline said:

Don't hold your breath on that.  It would be admitting a rather glaring mistake.  I don't mean to say that Linus has never admitted mistakes, because he seems like a pretty humble guy (especially considering how successful he is), but releasing a more accurate video that totally negated the old one wouldn't be great for their image.  They made the anti-delid video, including every mistake you could make to "prove" that delidding didn't improve temps in a meaningful way, and never made a statement on it (that I'm aware of).

 

It's ironic that they probably spent more time trying to turn 6 year old paste into something usable than the amount of time it would have taken to search for Intel's paste and order it.

Oh, I am very aware of their history of not really following through once called out until months/years later, I was very vocal about their memory tests in the past. I just figured it would be worth a shot now, given they seem to be getting more and more interested in thermals as of late. 

 

54 minutes ago, M.Yurizaki said:

So one thing in researching this led me to this post on Medium: https://medium.com/@OpenSeason/soldered-cpu-vs-cheap-paste-59fb96a4fca7

 

I should probably read it more thoroughly but it brought up some points:

• The reason why Intel went to a TIM instead of soldering wasn't a cost issue. It was a thermal-mechanical stress issue. The primary concern is this:
• Thermal conductivity is only part of the story. In a CFD simulation they did, they found that a TIM with 20W/mK did not perform any better than a TIM with 10W/mK
• They mentioned having low thermal resistance also plays a role. However, thermal resistance is not the same as the reciprocal of thermal conductivity. The reciprocal of thermal resistance is thermal conductance.
  • I don't claim to be an expert in thermal engineering, just pointing things out here.

So basically... it appears a lot of us (myself included) have more to learn than just numbers on a spec sheet. But hey, at least that post shed some light on things.

It's difficult to consume information like this without knowing the exact testing methodology. It might be true to say a 20W/mK paste and 10W/mK paste perform identically if used on say, a stock Intel cooler when the cooler itself is the bottleneck, but that could change if using a different cooler, where the bottleneck is no longer tied to the thermal solution. The phrasing of "perform any better" is also ambiguous. Are we talking actual processor performance upon using different pastes? Or just the difference in temperatures when using the different pastes? Again, it's important to know the cooler used, the thermal load used, the ambients (and controls during those ambients) and the properties of the thermal pastes they used, along with controls in mounting pressure and application method.

 

I too am no engineer, and I certainly would not claim to be an expert in this subject, but one does not need to be an engineer to request a testing methodology in order to recreate a test to compare the results.

 

I ran some pretty basic tests after I got into delidding. I've tried a few different brands of liquid metals, and a ton of different "conventional" pastes. While it's true that LM gives you the best thermal performance, it should also be mentioned that simply removing the silicone adhesive and using the stock Dow Corning paste also yields a pretty substantial drop in temperature. I've seen a 5-7C drop in doing that alone, along with also improving per-core variance in temperatures substantially. I've seen about a 10C drop using Gelid GC Extreme, which is consistently superior to TC-1996 as far as thermal performance goes, but there is no question that Gelid wouldn't last as long under intense thermal cycles. I've actually observed this with my GPU's, having already needing to re-paste a GPU that was only a year old after noticing it getting warmer under normal loads. This was certainly not the case with the stock EVGA Shin-Etsu paste, which performs nearly identically from the day I first got it on my GTX 1070 that my brother now uses.

 

If Linus and crew are not able to do a video on the subject, I might try my hand at it just to get the information out there. Wouldn't really hurt to have additional sources on this if others would like to try as well. Might finally end that "Intel Toothpaste" meme once and for all.

[Mira Yurizaki]

27 minutes ago, MageTank said:

It's difficult to consume information like this without knowing the exact testing methodology. It might be true to say a 20W/mK paste and 10W/mK paste perform identically if used on say, a stock Intel cooler when the cooler itself is the bottleneck, but that could change if using a different cooler, where the bottleneck is no longer tied to the thermal solution. The phrasing of "perform any better" is also ambiguous. Are we talking actual processor performance upon using different pastes? Or just the difference in temperatures when using the different pastes? Again, it's important to know the cooler used, the thermal load used, the ambients (and controls during those ambients) and the properties of the thermal pastes they used, along with controls in mounting pressure and application method.

The test they did was a CFD simulation. And by "performing better" bit, it was explained in this paragraph:

Quote

The effect of increasing the material thermal conductivity results in a lower thermal resistance just as we would intuitively expect. However it seems that as the conductivity increases, the bold line thickness of the paste plays a less significant role in the heat transfer. MAT4 and MAT5 behave almost the same and show almost no change with increasing thickness! Also see that a larger die has a much smaller Theta-JC, 10x less in this case, since larger die sizes allow for better heat distribution. Cup Lid IHS allows for more permissive parameters for mechanical properties of the thermally conductive epoxy material. IHS thickness made little difference in the results.

Since we're dealing with thermal engineering here, I don't think what the processor does is relevant. All that matters is how much heat can be transferred.

 

1 hour ago, JoostinOnline said:

It's ironic that they probably spent more time trying to turn 6 year old paste into something usable than the amount of time it would have taken to search for Intel's paste and order it.

I got the impression they were testing the viability of the TIM after 6 years. But then that wouldn't be a fair test anyway against new TIM.

[Arika]

this just in; Crusty old thermal dirt doesn't work as well as brand new thermal paste from the tube. more at 11

 

If they really did do it for longevity, i would prefer what they use now and never had to delid and reapply new paste for the sake of a few degrees when it's not even overheating.

[JR88]

Exactly you'll need to change it anyways, why being should of stayed with the Solder. Or just go overkill on water that's still gonna cool most things.

[wikipooh]

use some lighter fluid(https://www.amazon.de/Zippo-Benzin-f%C3%BCr-Feuerzeuge-125-Inhalt/dp/B000F37YKQ/ref=sr_1_4?ie=UTF8&qid=1521971158&sr=8-4&keywords=zippo+benzin) to make the paste fluid again. and test that way

[JoostinOnline]

7 hours ago, M.Yurizaki said:

I got the impression they were testing the viability of the TIM after 6 years. But then that wouldn't be a fair test anyway against new TIM.

He said they used old TIM because they didn't want to buy new CPU's, so I don't think so. 

[Sfekke]

I'm also on edge about replacing the TIM on my 4790K since temperatures have been climbing steadily since I got it.

Even with a beefy Scythe Mugen Max with Push/Pull 140MM fans on it C°80 isn't unusual.

Problem being if I break the thing, lets just say I'm SOL on multiple levels.

[Trik'Stari]

Based on my experience delidding two intel CPU's and disassembling a few thousand laptops from a manufacturer that rhymes with "Hell", I'd say they're using the same thermal paste.

[JR88]

11 minutes ago, Sfekke said:

I'm also on edge about replacing the TIM on my 4790K since temperatures have been climbing steadily since I got it.

Even with a beefy Scythe Mugen Max with Push/Pull 140MM fans on it C°80 isn't unusual.

Problem being if I break the thing, lets just say I'm SOL on multiple levels.

Airrr not helping either...airrr a little harder time with Load temps imo. need a loop..even if not a fascinating one and a pump that's like 3 or 4 on a d5. Really should have a d5.