Best Thermal Paste for CPUs 2024: 90 Pastes Tested and Ranked

[Smurfles]

Did you let the AS5 or IC diamond cure before testing? These seem to be awfully high numbers for both of them. I know AS5 usually takes like 8 hours before its really working well

 

[rubix_1011]

This article has some serious problem, it has some serious questions unanswered.

First of all: how did you applied the cooling paste?
Do You really know how a cooling paste works?
It seems to me that you don't.

No worries I will explain.

For instance, it is not so much about the type as it is much more important of how.

Don't ever think applying a thick cooling paste will solve a temperature problem! Actually it will worsen this!
The reason is:
- The only sole purpose of a cooling paste is to even out the microscopic uneven and un-leveled surfaces between the component and its cooler.
NOT to create a layer of paste between the parts.

Metal to metal is of course the best cooling and if one put to much paste in between, well that is not an optimal way of how to do it.
Cooling paste by its nature has worse temperature transfer than a clean metal to metal.
That is the reason for my advice below:

Just a small, very thin layer of the cooling-paste over the surface - Not more.
That is enough to fill the gaps in the microstuctrure and every un-even surface.
So, how to do this: Just a small dot in the center of the CPU or preferably use a a credit-card (or similar) and with the same amount of small application, just smoother it out - remember it has to be thin, very thin almost unseen.

Otherwise the cooling paste will serve as a another poor heat spot.

• Also it will take approx 15 to 20 temperature cycles before the cooling paste gets to its optimal performance, meaning for most it will then drop temperatures from 2-5C.
• Why? Because it is a paste and it will take some time to evaporate its solvent to become a dry heat transfer, its fumes has to go away.

Temperature cycles in this regards meaning you run the system at full power and then shut i off and wait to cool down, which will be made by natural usage, no need for excess.

• But for your so called "test" it has to be done for every single paste - No way around.
• Then perhaps You will find something different on the temperature readings for every paste you tested.

But Your article here does not mention anything near of this or how you applied the different paste for the test nor did you take up that the solvent in its paste has to go before you can actually take a reding of the temperature and therefore I strongly suggest that this article should be seeing as a subjective test and Not qualified as a serious objective test.

Best regards from Sweden.
Ps. I am sorry for my poor English and grammar.

I feel that I have likely applied thermal compound more than 99% of readers who visit this site from doing cooling reviews as well as the thermal compound reviews. While not possible to get precise to the exact volume each time, my approach is consistent between all applications for both metal and traditional paste compounds. I have built PCs for nearly 25 years, built watercooled PCs for 20 and have been doing the cooling reviews for Tom's Hardware for nearly 5 years, this includes Intel and Threadripper.

Overall, each compound was evaluated and stressed over a regimen cycle covering six hours with two different coolers and different mounting tensions, making for no less than 90 hours of compound testing for our initial round of tests.

I run a 2-hour automated load cycle via batch file which loads Prime95 for 5 minute intervals with 60 second cooldown between processes for each paste application.

This does not fully account for every single paste (addressing the questions around longer cure times) but in doing so also means variables which are not applied to all pastes in the same way, so the attempt here is to make a single process which is applied to all compounds in the same manner. The minimal outliers should not dictate the overall testing procedure.

 

[DonQuixoteIII]

The one thing missing from ALL thermal paste reviews is the ease of separating the CPU and cooler after a period of time. At some point everyone WILL need to remove their cooling apparatus for one reason or another, and the awful prospect of removing by force the CPU from its socket while trying to remove the cooler makes me cringe, and my wallet to have nightmares.

 

[DSzymborski]

Meh. Most people have dental floss and you can get a spool of fishing line for $6. The problem is people who rip out their cooler like it's 410 and they're sacking Rome with Alaric.

 

[hotaru251]

"best premium" #1 is worse than the runner up....because of price?

what part of "premium" don't you understand? ppl who shop that dont care about the cost.

 

[gamersoul]

Where is TF8 ?

 

[McNutts]

The one thing missing from ALL thermal paste reviews is the ease of separating the CPU and cooler after a period of time. At some point everyone WILL need to remove their cooling apparatus for one reason or another, and the awful prospect of removing by force the CPU from its socket while trying to remove the cooler makes me cringe, and my wallet to have nightmares.

I have been building computers since the mid 90s and the best method I have found is to slightly wiggly/rotate the cooler. This will increase the shear strain on the thermal paste and break it free most of the time and allow the cooler to be removed without pulling out the CPUs that use pins on the chip.

If you don’t wiggle the cooler your working against the paste in tension and it requires more force in tension than shear to break free.

min the near future we will not have to worry about this anymore. With AMD switching over to a LGA socket we will be able to just rip off our coolers without fear of the CPU coming out with the cooler when we take it out.

 

[majinsnake]

Cryorig thermal pastes should be added to the lists. CP5, CP7, CP15

http://www.cryorig.com/cp.php

 

[Noah Katz]

Great work, but am I the only one who finds the blurry charts impossible to read?

 

[der___alex]

Be aware the MX5 had QA issues? In germany no MX5 is avaiblable maybe due QA issues?

Arctic press release about defective batches of Arctic MX-5

www.overclock.net

 

[ckelih]

Is it just me or does this article have really unreadable pictures implemented?
If I click on any of the charts to see the ranking the resolution is so bad, I cant read anything on them.

 

[ckelih]

This article has some serious problem, it has some serious questions unanswered.

First of all: how did you applied the cooling paste?
Do You really know how a cooling paste works?
It seems to me that you don't.

No worries I will explain.

For instance, it is not so much about the type as it is much more important of how.

Don't ever think applying a thick cooling paste will solve a temperature problem! Actually it will worsen this!
The reason is:
- The only sole purpose of a cooling paste is to even out the microscopic uneven and un-leveled surfaces between the component and its cooler.
NOT to create a layer of paste between the parts.

Metal to metal is of course the best cooling and if one put to much paste in between, well that is not an optimal way of how to do it.
Cooling paste by its nature has worse temperature transfer than a clean metal to metal.
That is the reason for my advice below:

Just a small, very thin layer of the cooling-paste over the surface - Not more.
That is enough to fill the gaps in the microstuctrure and every un-even surface.
So, how to do this: Just a small dot in the center of the CPU or preferably use a a credit-card (or similar) and with the same amount of small application, just smoother it out - remember it has to be thin, very thin almost unseen.

Otherwise the cooling paste will serve as a another poor heat spot.

• Also it will take approx 15 to 20 temperature cycles before the cooling paste gets to its optimal performance, meaning for most it will then drop temperatures from 2-5C.
• Why? Because it is a paste and it will take some time to evaporate its solvent to become a dry heat transfer, its fumes has to go away.

Temperature cycles in this regards meaning you run the system at full power and then shut i off and wait to cool down, which will be made by natural usage, no need for excess.

• But for your so called "test" it has to be done for every single paste - No way around.
• Then perhaps You will find something different on the temperature readings for every paste you tested.

But Your article here does not mention anything near of this or how you applied the different paste for the test nor did you take up that the solvent in its paste has to go before you can actually take a reding of the temperature and therefore I strongly suggest that this article should be seeing as a subjective test and Not qualified as a serious objective test.

Best regards from Sweden.
Ps. I am sorry for my poor English and grammar.

You sound like a knowledgable person about thermal paste.
I have a question for you.
I added pads and paste on my watercooled 3080TI many times until I got good temperatures.
But I am still not satisfied. Thats why I ordered K5 Pro thermal paste.
https://www.amazon.de/K5-Pro-hochviskose-Wärmeleitpaste-Thermo-Pad-Ersatz/dp/B00UTX7K2E

Which should habe the benefit of having very good cooling stats and because it is a "paste" its adjusting itself to the "gap" between gpu-block and vram/gpu.
What do you suggest in my case?
Risk trying out that K5Pro instead of using thermal pads?

 

[Johnpombrio]

It looks like the ProlimaTech PK-3 Nano Aluminum paste is almost impossible to buy except at scalper prices. None on Amazon or other sites while 1.5 Grams on NewEgg goes for the same price as 5 Grams and ships from China.

 

[Your Username]

The amazon link to Thermal Grizzly Kyronaut links to Hydronaut. Wanted kyronaut bought hydronaut instead, but i should have read what i was buying.

 

[LukeHK45]

If possible, I would like to see Chinese OEM thermal compounds in these tests, such as:

• Halnziye HY883
• Halnziye HY-T1
• Halnziye HY-A9
• Halnziye HY-P16

• Foshan GD900
• Foshan GD900-1
• Foshan GD007

• Maxtor CTG8
• Maxtor CTG9

• Thermagic ZF-12
• Thermagic ZF-EX

• Alseye T9+ Platinum
• Snowman MTG10
• Wovibo WB-4

Also would like to see Thermalright TFX, TF9 and TF8 on the list as well.

In the past I also used a thermal compound called Thermalcote II from an American company known as Aavid Thermalloy, I can't find it anymore, but it used to perform extremely well with my overclocked 4690K back in the days, I wonder how it performs against the high-end thermal compounds.

All of these are excellent performing thermal compounds but due to them being foreign or unknown no one actually believes they're good or useful, so having them competing in a Tomshardware article would make all the difference.
Halnziye for example is an OEM that makes thermal compounds for Gelid (GC-Extreme, GC-Pro), Phobya HeGrease Extreme , Thermaltake, Zalman, Antec, Foxconn production lines, Lenovo computers, AVC production line, Aerocool, Uphere PH-10, etc, I think that having OEMs in the competition would be pretty interesting to see how their own products stack up, also most companies source their thermal compounds from the same OEM, so might as well just test the products from the factory itself. It was already pretty nice to see a Dow Corning TIM in the list, more OEMs like Halnziye and Shin Etsu on the list would be great as well.

 

[rbuffetta]

Arctic MX-6. test this one too please.

 

[watzupken]

While this is not an easy test, but I feel it is hard to say how well a thermal compound really is without looking at the long term performance. Some of these incredibly expensive paste don't hold up over time, resulting in declining performance over time.

 

[watzupken]

Be aware the MX5 had QA issues? In germany no MX5 is avaiblable maybe due QA issues?

Arctic press release about defective batches of Arctic MX-5

www.overclock.net

Defective batch is not impossible. The batch of Thermal Grizzly Kyronaut that I purchase quite a number of years back was causing very bad scratches/ scarring on the baseplates of all the coolers I applied the thermal compound with. I only realized its a defect when I checked online to find others having the same problem. So all the 3 coolers that I used had bad scarring after removed the thermal compound. After that, I've never purchased an overpriced Thermal Grizzly product.

 

[Digital~Dreams]

Please add Thermal Grizzly Kryonaut Extreme !!

 

[10tacle]

Great readup, and it about jives with my experience over the past 10 years of PC building. Prior to ten years ago, I was sold on Arctic Silver Ceramique. Then for a 2014 build (still used as a backup gamer), I used the Noctua paste that came with my NH-D14. That was better.

Finally for my most recent build last year I decided to plunk down for Grizzly Kryonaut and could not be happier. It most impressed me when applying to my 3080 Ti which dropped load temps 7-8C over whatever EVGA used at the factory. That's a big margin when dealing with GPU temps in the summer on an air rig reaching 80C with a 65% fan curve setting. Grizzly is expensive but beware that there are counterfeits being sold out there. The real stuff has the German information inside.

 

[USAFRet]

which dropped load temps 7-8C

From what to what?

 

[10tacle]

From what to what?

~80C - now running low 70s. Some of that could be from the application process not being ideal at the factory.

 

[USAFRet]

~80C - now running low 70s. Some of that could be from the application process not being ideal at the factory.

And performance exactly the same.

 

[enlyus]

Didn't see TG Kryonaut Extreme on the list. Would love to see how that stacks up!

 

[JohnDoh337]

Any reviews on Honeywell 7950 phase change thermal pad? I've read individuals giving it high praise but have not seen any performance comparisons or anyone really talking about this stuff.