The Official Thermal Interface Material thread

[RJARRRPCGP]

That was the shit here in Finland about 15 years ago, nobody here uses that anymore.

I've been using Antec Formula 7 in the late-2010s and most recently, Thermaltake TG-50.

 

[lexluthermiester]

Here is the more recent (2018) document (P18) that was also attached above and these more recent articles (also repeated from above)

• 2020 How to Avoid Pump-out and Achieve Efficient Heat Transfer - Electrolube
• 2020 Intel: replace thermal compound “every few years” (ctrl.blog)

Neither of those articles shows anything other than conclusions without the testing science to support those claims. It is effectively hear-say. For example, the photo of the ctrl.blog article does not show anything other than evidence of poor TIM application, which is far more common and plausible than pump-out effect.

Additionally, the pdf you attached is a general discussion piece and does not offer evidence of the effect. The only mention is on page 18, section 7.1 and does not make much of an explanation. This is not evidence. It is only more hear-say.

 

[Shrek]

Additionally, the pdf you attached is a general discussion piece and does not offer evidence of the effect. The only mention is on page 18, section 7.1 and does not make much of an explanation. This is not evidence. It is only more hear-say.

P19 covers the evidence (Fig 27)

https://www.semikron.com/dl/service-support/downloads/download/semikron-application-note-thermal-paste-application-en-2018-01-19-rev-00/

 

[80-watt Hamster]

Neither of those articles shows anything other than conclusions without the testing science to support those claims. It is effectively hear-say. For example, the photo of the ctrl.blog article does not show anything other than evidence of poor TIM application, which is far more common and plausible than pump-out effect.

Additionally, the pdf you attached is a general discussion piece and does not offer evidence of the effect. The only mention is on page 18, section 7.1 and does not make much of an explanation. This is not evidence. It is only more hear-say.

You're sticking to your position with almost religious fervor. Why? It's true that the ctrl.blog and Electrolube articles are just that: articles rather than studies. However, Electrolube is a manufacturer of thermal solutions. It's a little hard to believe they'd spend time on a non-existent phenomenon. Semikron is a provider of power soutions, including IC and MOSFET. While Andy's source isn't meant to address pump-out specifically, the fact that it's included as a concern gives the concept some weight considering the source. Call it all hearsay if you like, but the entities doing the saying are in something of a position of knowledge.

The ctrl.blog article also references a couple of additional sources not mentioned here, one from the National Renewal Energy Laboratory (pp. 4-5), and another from Intel (pp. 7-8). If your claim is that pump-out is of minimal concern in the PC space due to the state of modern components and compounds, sure; I'm on board. To unilaterally declare the effect a myth is a tougher position to support.

EDIT: Corrected article reference to ctrl.blog (erroneously credited Intel originally)

 

[lexluthermiester]

P19 covers the evidence (Fig 27)

https://www.semikron.com/dl/service-support/downloads/download/semikron-application-note-thermal-paste-application-en-2018-01-19-rev-00/

That photo shows little more than paste that has been spread thin by very flat surfaces, nothing more. I have seen this in situations were the heatplate and IHS of a CPU has been laped flat and smooth.

You're sticking to your position with almost religious fervor. Why?

I am a scientist. I do not accept claims which fly in the face of common sense without merit. Nor are papers written to look official that do not cite proper and verifiable data acceptable. So until someone shows evidence that "pumpout effect" actually takes place in situations other than extreme temp swings, I will stand firm and continue to call it what it is, myth.

 

[AlwaysHope]

Neither of those articles shows anything other than conclusions without the testing science to support those claims. It is effectively hear-say. For example, the photo of the ctrl.blog article does not show anything other than evidence of poor TIM application, which is far more common and plausible than pump-out effect.
...

This so called 'pump-out' effect is nothing more than an excuse for poor TIM application. The thinking is, if some is good then more is better. Adequate TIM is all that is needed, not too much! Like a lot of things in life, its a balance concept.

 

[Ferrum Master]

Hey scientists

Take a look on this. Recommend more stuff to someone else, what's yet very unproven.

PR | ARCTIC MX-5 Thermal Compound | ARCTIC

ARCTIC Press Information on MX-5 Thermal Compound | ARCTIC Service | ARCTIC Support | ARCTIC Store

www.arctic.de

 

[lexluthermiester]

Hey scientists

Take a look on this. Recommend more stuff to someone else, what's yet very unproven.

PR | ARCTIC MX-5 Thermal Compound | ARCTIC

ARCTIC Press Information on MX-5 Thermal Compound | ARCTIC Service | ARCTIC Support | ARCTIC Store

www.arctic.de

Oh, you mean what's been talked about already here;

ARCTIC Identifies Bad Batch of MX-5 TIM in the Market, Offers Free Replacements

For ARCTIC, it has always been a top priority to supply customers only with products of the best quality and durability. Should there ever be any doubts about products we've delivered, we stand by our responsibility to our customers. Current findings from our permanent quality assurance tests...

www.techpowerup.com

Yeah, you can stop with your condescending trolling smart guy.

 

[80-watt Hamster]

Hey scientists

Take a look on this. Recommend more stuff to someone else, what's yet very unproven.

PR | ARCTIC MX-5 Thermal Compound | ARCTIC

ARCTIC Press Information on MX-5 Thermal Compound | ARCTIC Service | ARCTIC Support | ARCTIC Store

www.arctic.de

Shizz happens in manufacturing. Even Toyota, as close to a paragon of reliability as exists in the automotive world, has recalls.

 

[Bones]

I covered this some time ago, just to recap what I posted:

"Pump-out Effect" Is it real or is it nonsense? Let's discuss...

@plastisch Can you point me to thermal paste that have 18W m/k?

www.techpowerup.com

Hope this helps to clear things up here.

 

[Ferrum Master]

Shizz happens in manufacturing. Even Toyota, as close to a paragon of reliability as exists in the automotive world, has recalls.

Thus you don't recommend a product that's been around recently. That's really scientific ain't it. You can do experiments on your own as you wish. But recommend?

Same as with Pump Effect. Denial of the obvious works for sure. Recent TIM have problems with shelf life and stability, thus ir returned to proven chemistry for the price of 1 degree C but I know that stuff for more than a decade.

 

[Shrek]

I think pump-out can be mitigated by having the right fan profile such that the chip stays hot even when idle; I believe this is already done on video cards to reduce solder fatigue.

 

[Bones]

I think pump-out can be mitigated by having the right fan profile such that the chip stays hot even when idle; I belive this is already done on video cards to reduce solder fatigue.

Actually no - The parts still cool down when the machine is turned off so you still have thermal expansion and contraction taking place.
As for letting a chip get hot I'm not a big fan of that even though they can take it - I've never liked my fan not running while the card is working and that only means less thermal headroom while the card is in use.

 

[Ferrum Master]

Actually no - The parts still cool down when the machine is turned off so you still have thermal expansion and contraction taking place.
As for letting a chip get hot I'm not a big fan of that even though they can take it - I've never liked my fan not running while the card is working and that only means less thermal headroom while the card is in use.

The problem ain't that bad with high viscosity pastes in my experience. The problem seems to be also the filler, the newer pastes are even more prone to it. Looking at graphs the temperature diference actually doesn't need to be that big to provoke the effect. So all things together, bad filling material of the paste, heatsink surface, geometry, pressure and temperature variation. If you want a paste that lives long, you have to take consideration of all these things.

There is only one article online on IEEE.org also viewing recent consumer TIMs, but that's a paid one.

Although no one is denying that the effect is real there.

 

[Bones]

The problem ain't that bad with high viscosity pastes in my experience. The problem seems to be also the filler, the newer pastes are even more prone to it. Looking at graphs the temperature diference actually doesn't need to be that big to provoke the effect. So all things together, bad filling material of the paste, heatsink surface, geometry, pressure and temperature variation. If you want a paste that lives long, you have to take consideration of all these things.

There is only one article online on IEEE.org also viewing recent consumer TIMs, but that's a paid one.

Although no one is denying that the effect is real there.

Yeah, thicker TIMs don't tend to "Move" as readily as thinner TIM's and in some cases a thinner TIM is better, in others it's not.
As for a TIM that stays a thicker TIM is better for that but again, there are cases in which a thinner TIM does just fine and even stays in place too.

It's all a matter of application to what hardware and TIM is used and the conditions the hardware lives under that determines what goes in the end along with proper application of the TIM itself to the hardware for use.

 

[Shrek]

Actually no - The parts still cool down when the machine is turned off so you still have thermal expansion and contraction taking place.

mitigated: make less severe, serious, or painful.

Not eliminated.

The problem ain't that bad with high viscosity pastes in my experience.

The problem is gone with pastes that cure.

 

[Bones]

mitigated: make less severe, serious, or painful.

Not eliminated.



The problem is gone with pastes that cure.

Yes and no - Concerning mitigation, It still goes up to the temp it does while operating and once you shut the machine down it still cools down fully to room temp regardless of how the fans are set up.

If it's normally seeing 85-90c while running under load, that's where your thermals will be even if there is a lesser thermal variance while in operation once it reaches temperature.
You can't keep it from cooling down fully once you shut off the machine which brings the other end of the thermal range into play, that means it still sees the full range of temp variance anyway.

It may take longer to happen (Mitigation), this is true but the end result is still about the same once all is said and done.

 

[Shrek]

I'm a believer!

 

[VeqIR]

I installed my CPU last April (i9-10900KF) and used a fresh container of TG Kryonaut from Microcenter. Since then I've been worried whether my batch was one of the bad scratchy ones, but haven't felt like repasting until there was need with rising temperature. Lately the temperatures have not been great at all--today spiking up to upper 80s... I decided it was really time for the change.

Upon removing the air cooler tower I found that the Kryonaut was pretty liquidy (it was recently warmed by CPU activity) and there were lots of spaces that were pretty much empty of the paste. I had bolted down the CPU cooler very tightly before, so it wasn't really poor contact. But that certainly explained the performance degradation.

Cleaned everything up and thankfully didn't really find scratches on the cooling plate or the CPU except one 2mm long one on both, which I know was not there before. It's minor enough that I just won't worry about it at all.

Changed to Noctua NT-H2 applied with an X. Normally I do a thin spread but decided to try Noctua's officially recommended method for this paste. Bolted down very tightly again. Excellent temperatures right away! I feel like even lower than TGK used to do when I first used it last year. In one game where I was seeing mid 40s to mid 50sC and spikes to mid-upper 70s with TGK in the recent months and to upper 80s today. Now the CPU has been sitting at around 33-36C on all cores, going up to 45-47C in spikes while in a game! The highest temp on the CPU since I turned it on a few hours ago is 56C for package temp and 54C per individual core (I've had HWiNFO64 running in the background), and I'm doing all the normal stuff so far that I usually do.

Glad my Kryonaut was a tiny tube, I won't be using it again... I know 10 months of use is not terribly short, but in the past my paste jobs lasted much longer than that without such drastic deterioration. My other contender instead of NT-H2 was SYY-157, but I found that one super hard to use in the past. NT-H2 has been very easy to use--I've had it going on my family member's desktop for a few months so far, also with excellent temperatures and no noticeable degradation so far.

P.S.: my previous go-to paste prior to last year was AC MX-4. It was alright, didn't really have any issues with it. Longevity was fine, but not great. Much much better than TGK though. Hoping NT-H2 holds up. Now I have to open up a desktop I listed on Craigslist and remove TGC in favor of NT-H2 or SYY-157... Don't want someone else experiencing this performance deterioration too fast and not knowing what to do about it.

 

[-batab-]

This so called 'pump-out' effect is nothing more than an excuse for poor TIM application. The thinking is, if some is good then more is better. Adequate TIM is all that is needed, not too much! Like a lot of things in life, its a balance concept.

The so-called "pump-out" effect exists and it is actually well-known among scientists with expertise in similar fields. It is even mentioned in very recent scientific literature such as "Thermomechanical Degradation of Thermal Interface Materials: Accelerated Test Development and Reliability Analysis" by Carlton et al. (DOI link: https://doi.org/10.1115/1.4047099 ).
They even refer to a recent scientific review "Novel nanostructured thermal interface materials: a review" by Hansson et al. (DOI link: https://doi.org/10.1080/09506608.2017.1301014 ).

So yeah, the phenomenon is well-known and also perfectly reasonable to anyone with a background in solid mechanics and thermomechanical interactions.

Also, even if i didn't take photos I just repasted my 7820HK + 1060 GTX laptop after 5 years of Kryonaut. Kryonaut was not dried out and still has a nice viscosity but was completely pumped out. I was basically running with no paste and I lowered my temps by 25-30°C by applying (horribly, since last time I applied a thermal paste was 8 years ago) a Thermalright TFX.

Actually no - The parts still cool down when the machine is turned off so you still have thermal expansion and contraction taking place.
As for letting a chip get hot I'm not a big fan of that even though they can take it - I've never liked my fan not running while the card is working and that only means less thermal headroom while the card is in use.

Actually yes since this is what fatigue damage is about. Fatigue is a cumulative damage phenomenon and while peak values are important, the number of cycles at that specific temperature amplitude is also important. If they are really reducing the cooling to reduce the operative temperature range they are most likely doing something very benificial from a structural point of view. The number of cycles caused be power ON/shut down operations is vastly inferior than the number of cycles during a single "powered ON" time window and high cycle and ultra high cycle fatigue are usually in the range of millions or billions cycles.

 

[Shrek]

Absolutely

I was happy to have the opportunity to present data, and I respect a person's right to question that data.

 

[watzupken]

This so called 'pump-out' effect is nothing more than an excuse for poor TIM application. The thinking is, if some is good then more is better. Adequate TIM is all that is needed, not too much! Like a lot of things in life, its a balance concept.

I don’t agree. I think you kind of over simplified the issue. I believe the issue caused by “pump out”effect can be replicated even with people that have been applying thermal compound the ”right“ way (whatever right way means). Assuming the same amount of TIM used between say Arctic MX4/5 or even Thermal Grizzly‘s thermal compound, versus say SYY-157 or TFX, over time you will noticed that the latter 2 holds up better when they are applied on bare die where temps can swing drastically.

I think any experienced hardware enthusiast will know not to spam TIM, but apply an amount that they think is sufficient (with may be a little more just in case). Even if they do over apply, it will just get squeezed out to the sides when there is pressure from the mount, which should not negatively impact cooling tangibly.

 

[Shrek]

From an old thread

Re-paste my 2080Ti, idle and load temp significantly down

So my 2080Ti has been overclocked mildly most of the time. As it is approaching 2yrs old I decided to check the thermal paste as I have been seeing some weird fan behavior. Basically when core temp gets to around 80C under load the fan will suddenly spike to 100% no matter what fan curve I set...

www.techpowerup.com

Pump out? reason I suggest this is because there is paste all along the outside edge.

I'm asking; not trying to start a dispute.

 

[AlwaysHope]

The so-called "pump-out" effect exists and it is actually well-known among scientists with expertise in similar fields. It is even mentioned in very recent scientific literature such as "Thermomechanical Degradation of Thermal Interface Materials: Accelerated Test Development and Reliability Analysis" by Carlton et al. (DOI link: https://doi.org/10.1115/1.4047099 ).
They even refer to a recent scientific review "Novel nanostructured thermal interface materials: a review" by Hansson et al. (DOI link: https://doi.org/10.1080/09506608.2017.1301014 ).

So yeah, the phenomenon is well-known and also perfectly reasonable to anyone with a background in solid mechanics and thermomechanical interactions.

Also, even if i didn't take photos I just repasted my 7820HK + 1060 GTX laptop after 5 years of Kryonaut. Kryonaut was not dried out and still has a nice viscosity but was completely pumped out. I was basically running with no paste and I lowered my temps by 25-30°C by applying (horribly, since last time I applied a thermal paste was 8 years ago) a Thermalright TFX.



Actually yes since this is what fatigue damage is about. Fatigue is a cumulative damage phenomenon and while peak values are important, the number of cycles at that specific temperature amplitude is also important. If they are really reducing the cooling to reduce the operative temperature range they are most likely doing something very benificial from a structural point of view. The number of cycles caused be power ON/shut down operations is vastly inferior than the number of cycles during a single "powered ON" time window and high cycle and ultra high cycle fatigue are usually in the range of millions or billions cycles.

I don’t agree. I think you kind of over simplified the issue. I believe the issue caused by “pump out”effect can be replicated even with people that have been applying thermal compound the ”right“ way (whatever right way means). Assuming the same amount of TIM used between say Arctic MX4/5 or even Thermal Grizzly‘s thermal compound, versus say SYY-157 or TFX, over time you will noticed that the latter 2 holds up better when they are applied on bare die where temps can swing drastically.

I think any experienced hardware enthusiast will know not to spam TIM, but apply an amount that they think is sufficient (with may be a little more just in case). Even if they do over apply, it will just get squeezed out to the sides when there is pressure from the mount, which should not negatively impact cooling tangibly.

Let's get into the semantics of what "pump out" means with TIM right here
I agree with the OP of that thread, its a nonsense term.

 

[erocker]

Pump out? reason I suggest this is because there is paste all along the outside edge.

I'm asking; not trying to start a dispute.

Nah. The pressure distribution during mounting and placement of the paste most likely resulted in where paste spread out over the IHS.