The Official Thermal Interface Material thread

[lexluthermiester]

we have AI thermal paste now. please make it stop.

Hold up, there might be some merit to this. AI has the potential to be very good at chemistry. This might be a good thing. I think testing is in order. If these hit the market, I will get some and do some testing here.

 

[Space Lynx]

Hold up, there might be some merit to this. AI has the potential to be very good at chemistry. This might be a good thing. I think testing is in order. If these hit the market, I will get some and do some testing here.

I agree with that, but I think in this case its just marketing stuff. I hope you are right though.

 

[Bones]

we have AI thermal paste now. please make it stop.

Cooler Master introduces colored ‘AI Thermal Paste’ — CryoFuze 5 comes with nano-diamond technology

Claims its can 'maintain stable performance' from -50C to 240C.

www.tomshardware.com

View attachment 348251

I coudn't agree more!

No way I'm using anything with diamond dust to royally scratch up and ruin the contact surface of my coolers/pots, plus it doesn't even work down to the temps I'd sometimes need it for anyway (-192c).

And AI stuff?

What's it gonna do?
Make your machine compute faster, help with framerates, spy and report what you're doing and where you've been, upload your browsing history to Skynet so they can send in the T-1000 liqui-clowns or..... What?
And those colors.... Like you'd really notice it anyway once the cooler is on and the board is in the case, esp if you happen to have your RGB "Bling" goin on in there.

Maybe an AI can come up with a better TIM and do try this stuff if you like - Should be obvious here I'm not.

 

[TPCEA]

Hold up, there might be some merit to this. AI has the potential to be very good at chemistry. This might be a good thing. I think testing is in order. If these hit the market, I will get some and do some testing here.

I just recently watched a video on diamonds, and it explained how good diamonds are at thermal conductivity, but didn't mention AI. I have no doubt about the efficacy of the product - just the marketing nonsense, good meister.

 

[Chomiq]

Hold up, there might be some merit to this. AI has the potential to be very good at chemistry. This might be a good thing. I think testing is in order. If these hit the market, I will get some and do some testing here.

This, 100%.

 

[Borc]

Regardless of the naming AI the regular Cryofuze was quite good, however the regular one has a higher thermal conductivity of 14 W/m.K, the AI one only 12.6 W/m.K, even though the thermal conductivity might not mean much. From the same company it could though.

 

[lexluthermiester]

This, 100%.

I'll be keeping my eye out for those, will test once I get some. I find them intriguing!

Saw this and thought is might be worth a look;

Best Thermal Paste for CPUs 2025: Over 50 Compounds Tested and Ranked

The right thermal paste will improve performance.

www.tomshardware.com

While I won't go so far as to call this round of testing bad, I find it a bit suspicious that they tested MX-5 instead of MX-6, even though MX-5 hasn't been available for over a year now(though in all fairness it did as well for them as my testing showed). The wording of the article makes me think that it has been in the works for over a year, which is reasonably fair, it takes a lot of time to do testing like this. However what really stood out to me was the fact that Arctic Silver was included and seemed to still be performing like a champ in with the best. Very interesting to me.

The article also did discuss it's testing methods which is a bit concerning, but perhaps they wanted to avoid the whining and griping from certain members of the public.

 

[TPCEA]

I'll be keeping my eye out for those, will test once I get some. I find them intriguing!

Saw this and thought is might be worth a look;

Best Thermal Paste for CPUs 2025: Over 50 Compounds Tested and Ranked

The right thermal paste will improve performance.

www.tomshardware.com

While I won't go so far as to call this round of testing bad, I find it a bit suspicious that they tested MX-5 instead of MX-6, even though MX-5 hasn't been available for over a year now(though in all fairness it did as well for them as my testing showed). The wording of the article makes me think that it has been in the works for over a year, which is reasonably fair, it takes a lot of time to do testing like this. However what really stood out to me was the fact that Arctic Silver was included and seemed to still be performing like a champ in with the best. Very interesting to me.

The article also did discuss it's testing methods which is a bit concerning, but perhaps they wanted to avoid the whining and griping from certain members of the public.

I was surprised by the omission of products like PTM7950 and its brethren, amongst a few others. "Why?"

 

[Shrek]

Ask the author?

palcorn@outlook.com

 

[lexluthermiester]

I was surprised by the omission of products like PTM7950 and its brethren, amongst a few others. "Why?"

That's easy to answer and understand, he said it in the article:

Even though a dizzying array of different thermal pastes is available

No one can be expected to know about, let alone test, all of them.

 

[Borc]

I'll be keeping my eye out for those, will test once I get some. I find them intriguing!

Saw this and thought is might be worth a look;

Best Thermal Paste for CPUs 2025: Over 50 Compounds Tested and Ranked

The right thermal paste will improve performance.

www.tomshardware.com

While I won't go so far as to call this round of testing bad, I find it a bit suspicious that they tested MX-5 instead of MX-6, even though MX-5 hasn't been available for over a year now(though in all fairness it did as well for them as my testing showed). The wording of the article makes me think that it has been in the works for over a year, which is reasonably fair, it takes a lot of time to do testing like this. However what really stood out to me was the fact that Arctic Silver was included and seemed to still be performing like a champ in with the best. Very interesting to me.

The article also did discuss it's testing methods which is a bit concerning, but perhaps they wanted to avoid the whining and griping from certain members of the public.

Because this test is really old, it's all recycled from years old tests and it's flawed anyways:

https://www.reddit.com/r/overclocking/comments/mk2ap8

A real new test is in the works from THG.

 

[Hyderz]

Cooler Master introduces colored ‘AI Thermal Paste’ — CryoFuze 5 comes with nano-diamond technology

Claims its can 'maintain stable performance' from -50C to 240C.

www.tomshardware.com

okay....

 

[KLiKzg]

Because this test is really old, it's all recycled from years old tests and it's flawed anyways:

https://www.reddit.com/r/overclocking/comments/mk2ap8

A real new test is in the works from THG.

Of course they are "misleading". Any test with CPU mounting, cleaning & applying new paste is misleading!
Why?
- How did author apply the paste?
- Did he/she use the proper applicator?
- Did he/she take exactly the same procedure for each application? Same amount of paste? (not possible)
- How did the cooler fit on CPU on each assembly?
- How did cleaner react to paste? Did author use same cleaner or special cleaner for each paste?
etc.

In order to have viable results, there should be some kind of test equipment - which is NOT COMPUTER.
Mounting must be mechanical.
Putting paste must be automated.

Then & only then, you can have "some good & comparable results"!
All others is just: testing & we will get some (non comparable) data.

 

[lexluthermiester]

Any test with CPU mounting, cleaning & applying new paste is misleading!

No it isn't. That is exactly how you test. The testing involves a single base system using the exact same testing conditions. This is the only was to get results that compare fairly to each other. That's how I test anyway.

 

[TPCEA]

No it isn't. That is exactly how you test. The testing involves a single base system using the exact same testing conditions. This is the only was to get results that compare fairly to each other. That's how I test anyway.

I think they are referring to high precision testing vs manual, and the variability that the human introduces to the paste application...

Personally, I feel that it is exceedingly hard to produce results that are equally valid for everyone because we're talking about computers and all the variables involved. It is basically impossible to test anything in such a way that it'll apply to everyone and, even if it were possible, by the time all testing was completed, which would take huge amounts of time and money unless it could be automated (including recalibration) and that would require robotics that are currently unavailable, well, by that time the stuff tested would probably be "old" in tech.

Still, I don't entirely trust TH's TIM (and some other) results, but I'm old, jaded and cynical.

 

[KLiKzg]

I think they are referring to high precision testing vs manual, and the variability that the human introduces to the paste application...

Personally, I feel that it is exceedingly hard to produce results that are equally valid for everyone because we're talking about computers and all the variables involved. It is basically impossible to test anything in such a way that it'll apply to everyone and, even if it were possible, by the time all testing was completed, which would take huge amounts of time and money unless it could be automated (including recalibration) and that would require robotics that are currently unavailable, well, by that time the stuff tested would probably be "old" in tech.

Still, I don't entirely trust TH's TIM (and some other) results, but I'm old, jaded and cynical.

Exactly.

+ you would need to have constant way of room temperature, which means you need to turn on climate control at least some 2h before the test in order not to interfere with the room temperature with the test! It has to be stable & keeping the room constant, not cooling it down to some value.

& again, automation does not need to cost much. I have done with the 1/2 of the price or a standard robot to seal application. So if it can be done with production on larger scale, it can be done on smaller scale also. But you need to have some constant test jig with some automated applicator & constant delivery system in order to have viable & comparable results.

 

[remixedcat]

What about RGB AI thermal paste??!!

 

[KLiKzg]

What about RGB AI thermal paste??!!
View attachment 348626

Frankly, it is too much LGBT for my taste.

 

[overclockedamd]

AI Thermal paste!!!! Have they not seen Terminator 2!!!!


*edit* sarcasm does not convey well.....twas a joke.

 

[KLiKzg]

AI Thermal paste!!!! Have they not seen Terminator 2!!!!

Guys, it is "not AI thermal paste", but AI generated thermal paste. Big difference!
But their marketing department is full of incompetent people. (or not, as we argue about it now)

 

[wNotyarD]

Guys, it is "not AI thermal paste", but AI generated thermal paste. Big difference!
But their marketing department is full of incompetent people. (or not, as we argue about it now)

Not even that. According to an errata on CM's social media, it was a mistranslation from their Chinese pages. It should read as "a paste for the AI era requirements", or something along these lines.

 

[Veseleil]

Not even that. According to an errata on CM's social media, it was a mistranslation from their Chinese pages. It should read as "a paste for the AI era requirements", or something along these lines.

That's even worse.

 

[ShrimpBrime]

I coudn't agree more!

No way I'm using anything with diamond dust to royally scratch up and ruin the contact surface of my coolers/pots, plus it doesn't even work down to the temps I'd sometimes need it for anyway (-192c).

And AI stuff?

What's it gonna do?
Make your machine compute faster, help with framerates, spy and report what you're doing and where you've been, upload your browsing history to Skynet so they can send in the T-1000 liqui-clowns or..... What?
And those colors.... Like you'd really notice it anyway once the cooler is on and the board is in the case, esp if you happen to have your RGB "Bling" goin on in there.

Maybe an AI can come up with a better TIM and do try this stuff if you like - Should be obvious here I'm not.

You and I have different ideas about diamond pastes lol.

There's absolutely no thermal paste that makes it that cold. You would even freeze acetone at that temperature.

If you value the laser etching, I get it. People delid these days. So maybe diamond paste isn't good for bare die, but I've never once had a problem.

Copper. Coolers. Lapped IHS plates.... When you clean the paste off with a towel, it shines the copper nicely like rubbing compound. It does take Lazer etching off, but nickel plating should be lapped off in my opinion. It's much worse than just saying I won't use this paste because scratches. Nickel conducts like shit. I've never seen lines like scratches, but maybe some diamond pastes have clumps of diamonds that could cause this effect. My take was when I wiped it off, it clumped in the towel and maybe left a visible line.

Not to take away from surface area. Scratches aren't horrible. I didn't have horrible results scuffing plates with 60 grit emery cloth. It seemed to work well with thicker pastes. IC diamond is thick. Difficult to work with cold. Becomes a little pliable when warmed up. Antec Formula 6 and 7 are very decent diamond pastes. I've always had good thermals, and not super thick. Worked fine both with lapped, scuffed and direct die cooling.

Been using diamond pastes about 15 years straight. I'm satisfied enough. Just like guys that stick with artic silver 5. To each their own

 

[micropage7]

What about RGB AI thermal paste??!!
View attachment 348626

actually green or red looks great than boring grey

 

[Athlonite]

CoolerMaster are saying their CryoFuze 5 has a thermal conductance of 12.5W/mK so it may be good we'll just have to wait and see