Thermal Paste Round-up: 85 Products Tested

[mindbreaker]

I'd be very interested to know which thermal compounds do the best when the CPU has been lapped well first. I would think the thick ones would be much worse. But that is just a guess without a test.

There are a hundred things I'd like to test. They make thin foils of various metals: gold leaf, silver leaf, copper leaf etc. I'd like to see how that zinc stuff for your nose works. Silver polish might be interesting. Carbon nanotubes would be interesting. Some buckyballs...

 

[InvalidError]

I'd be very interested to know which thermal compounds do the best when the CPU has been lapped well first. I would think the thick ones would be much worse.

The key determinant to how well something will perform as a TIM (after not using junk as the transfer media) is how thin the final layer will be once everything has settled. Between two perfectly lapped surface such as the faces on gauge blocks, nothing would be the best TIM but then you'd run the risk of the two surfaces cold-welding together.

Foil wouldn't work at all since unless you bring it up to melting temperature or crush it with sufficient force to extrude the foil from peaks between surfaces and force it into valleys, it won't help fill microscopic gaps between surfaces and may add some of its own.

 

[Stetson-Harrison]

Great article!

At the first page you mentioned Fisher Elektronik and their professional pastes. We've been using their pastes several years in high power electronic maintenance and repairing. I kind of expected to find it in your tests since you mentioned it, but maybe next time?

This is the paste we mainly use and it's non-conductive:

http://www.fischerelektronik.de/web_fischer/en_GB/heatsinks/E01.08/Thermal transfer compound and thermal interface film/PR/WLPF50_/index.xhtml

 

[sandaker]

Did you use the manufacturer recommended application methods in testing or did you use the dot application method for all of them?

https://www.youtube.com/watch?v=TRytGgmdeQM

I'd like to know wether grizzly's recommended application method is superior to using the dot application in a professional test enviroment.

 

[The Paladin]

I have been using Artic Silver 5 for years, yes there is a "burn in time" and no it isn't the "best of the west", its efficient, does not eat up your heatsink, and I don't know how you were testing it, but in water cooling, air-cooling testing of my own, I get better temps and results than you do with it so..

grain of salt.

using edge load page then click on the ... top right corner and select open in internet explorer. it works fine, because the web site needs to update somethings to be edge compatible.

 

[Pixel13]

Images? What images? I tend to think those might be useful, if one could see them.

 

[Spankymcbob1]

Not like someone hasn't made something like this before

 

[bit_user]

Not like someone hasn't made something like this before

But it needs to be updated periodically, as new products come onto the market and existing compounds undergo changes in production or sourcing. Plus, if multiple reviewers test the same compounds, you can get even better data by aggregating their results and rejecting the outliers. Lastly, I don't recall seeing anyone do viscosity and ease-of-use charts, before. Altogether, this was a worthwhile effort - even if there are a few controversial aspects.

Again, I thank you Igor. It seems like quite an undertaking.

Now, what would be really nice is to have more quality long-term testing.

 

[IInuyasha74]

Not like someone hasn't made something like this before

Now, what would be really nice is to have more quality long-term testing.

That would be very interesting to see, but I'm not sure it is something Igor is set up to do. Even if he limited the test to five or so thermal pastes and planned to test them for a period of two years and then re-check them, he would have to be careful that each system was placed under roughly the same amount of usage for the same amount of time every day for two years. Would take up quite a bit of time. Not to mention the financial investment to have five identical systems running like this, and the space it would need. I'm sure if Igor wanted to make it happen, he could, but I think he has far too many other important tasks on his hands.

 

[mindbreaker]

I'd be very interested to know which thermal compounds do the best when the CPU has been lapped well first. I would think the thick ones would be much worse.

The key determinant to how well something will perform as a TIM (after not using junk as the transfer media) is how thin the final layer will be once everything has settled. Between two perfectly lapped surface such as the faces on gauge blocks, nothing would be the best TIM but then you'd run the risk of the two surfaces cold-welding together.

Foil wouldn't work at all since unless you bring it up to melting temperature or crush it with sufficient force to extrude the foil from peaks between surfaces and force it into valleys, it won't help fill microscopic gaps between surfaces and may add some of its own.

Gold leaf is very thin you brush it on with a paint brush. It could easily fill gaps it shreds into pieces with the brush.

Metal typically will not weld itself unless the temperatures are high or the metal had had no contact with the air. Once exposed to the air there will be enough bonding with oxygen to prevent that. In space or in a vacuum, it would be possible. Most "cold welding" is not very cold. They use friction to "cold weld" typically.

 

[Guest]

I have been using Artic Silver 5 for years, yes there is a "burn in time" and no it isn't the "best of the west", its efficient, does not eat up your heatsink, and I don't know how you were testing it, but in water cooling, air-cooling testing of my own, I get better temps and results than you do with it so..

grain of salt.

using edge load page then click on the ... top right corner and select open in internet explorer. it works fine, because the web site needs to update somethings to be edge compatible.

I've also used Artic Silver 5 for years, and always had great results. But this year I re-evaluated, and the technology has really evolved.

I'm a no-nonsense guy that likes tested and proven solutions. I still use the noctua nh-d14, even if it's a decade old. Show me something better (air).

But the pastes have changed and evolved. Artic Silver 5 isn't the best, not by a long shot.

 

[randomizer]

It might not have the best performance, but IC Diamond is a multi-purpose TIM. You can use it for cleaning oxidation off both surfaces. I do it every time

 

[grozzie]

Why don't manufactures adopt a similar scheme to food stuff's....
Why can't they put a manufacturing date, a sell by date and an expected life span once used. It would only take a bit of pressure from sites like this and from retailers for them to adopt such a scheme. You can bet once one manufacturer does it they all will.

 

[InvalidError]

Gold leaf is very thin you brush it on with a paint brush. It could easily fill gaps it shreds into pieces with the brush.

The goal of TIM is to fill the gaps in uneven surfaces. I'm not convinced that gold leaf that breaks down into countless little solid chunks would do a remotely good job at filling those gaps.

Cold welding doesn't require a vacuum: if you have two perfectly flat surfaces and wring them together, practically all of the air gets displaced from the interface.

 

[cryoburner]

Now, what would be really nice is to have more quality long-term testing.

That would be very interesting to see, but I'm not sure it is something Igor is set up to do. Even if he limited the test to five or so thermal pastes and planned to test them for a period of two years and then re-check them, he would have to be careful that each system was placed under roughly the same amount of usage for the same amount of time every day for two years. Would take up quite a bit of time. Not to mention the financial investment to have five identical systems running like this, and the space it would need. I'm sure if Igor wanted to make it happen, he could, but I think he has far too many other important tasks on his hands.

I thought about this while reading the article, and you would not necessarily need multiple test systems to perform such a durability test. One option would be to set up a large temperature-controlled hotplate, and simply apply each thermal compound fastened between two small metal blocks, or between a metal block and an identical heatsink. This would allow for potentially dozens of thermal compounds to be tested at once. The hot-plate could be automatically controlled with a timer, and maybe even have it cycle the heat on and off repeatedly to simulate varying workloads. Each day, the position of the blocks could be rotated to prevent any potential hot-spots from affecting the results. At the end of any number of rotation cycles, the blocks could be tested again to see whether their performance has changed at all from when they were first tested. This would likely require a second layer of fresh thermal compound between the metal block and the CPU when you go to test the temperatures, but as long as that were performed identically each time, probably using a single compound, you should be able to get meaningful results.

It still might not be very practical to test this way though, as having a hotplate running all day could potentially pose a hazard, and it would still require a fair amount of long-term attention by the tester. It would probably be easier to test multiple compounds in actual computers, in a computer lab type of environment where a large number of systems will be running all day anyway. The only thing then is that you might see some computers get significantly more use than others, so you would ideally want to track the CPU load level in some way, and test each compound on more than one system.

Why don't manufactures adopt a similar scheme to food stuff's....
Why can't they put a manufacturing date, a sell by date and an expected life span once used. It would only take a bit of pressure from sites like this and from retailers for them to adopt such a scheme. You can bet once one manufacturer does it they all will.

One problem with that is that some of these thermal compounds can have very long shelf lives. How do you test the longevity of a new thermal compound that could last for a decade or more? Likewise, the durability could be impacted by storage or use conditions. A particular thermal compound might last much longer on a well-cooled home computer that gets used for a few hours a day than it would on a server processor under high load and temperatures. At best, they could include a date of manufacture and a rough estimate of how long they expect it might last for.

 

[larkspur]

I use MX4 specifically because it doesn't have a burn in period and because it lasts FOREVER.

No, it doesn't deteriorate. I've seen reports a decade after the fact showing less than three degrees celsius difference from when it was first applied.

So. Either you're biased because of ignorance, or both Artic's warranty and every long term test done before this has been lying. Gosh, lemme think which is more likely...

Now, is something like MX4 the best thermal paste out there? Of course not. But it IS way better than a lot of the market, super easy to apply and maintenance-free, and very reliable. If you're going to be a snob about your thermal pastes, at least be accurate about it.

I agree with this wholeheartedly. Been using Arctic MX-4 on CPUs and GPUs since the end of 2010 and have never had to reapply it or bother with burn-in time. I'm actually shocked that Igor specifically called it out as having a long burn-in time and low durability. Neither is true - it's almost like he has it confused with AS5...

MX-4 has "no" burn-in time and it is at least as durable as any affordable all-purpose TIM available. I have an i7-920 @3.5ghz with MX-4 and a Noctua nh-u12p that has been running almost 24/7 under heavy loads (SETI@home and Lightwave ScreamerNet renders) since the end of 2010 that has shown zero signs of TIM deterioration. I was very skeptical at Arctic's claim that MX-4 would last at least 8 years, but my own experience with about a dozen homebrew builds has shown their claims are coming true. Before 2010 I used AS5 and had to deal with very long burn-ins and TIM replacement every 3 years or so...

MX-4 is affordable, easy to apply, has no burn-in and is very durable over time - this is not just my opinion, it is a proven fact. I realize that reviewers can inject some subjectivity into their reviews, but Igor is just plain wrong about the burn-in and durability of MX-4. I'd suggest a bit of research before making such claims. Thanks for the otherwise excellent review!

 

[Ne0Wolf7]

I just realized:
We have a CPU cooler roundup.
We have a thermal paste roundup.
The only things missing for us to have a total suite of roundups for thermal performence is a case roundup and a fan roundup. I havn't seen a fan review here before, but I know we have quite a collection of case reviews... Just a thought.

 

[bit_user]

Now, what would be really nice is to have more quality long-term testing.

That would be very interesting to see, but I'm not sure it is something Igor is set up to do.

It was meant as neither a request nor a critique. Just a statement to be taken at face value.

I was actually thinking it'd be neat for someone to do in some sort of datacenter or a reasonably large server room.

 

[bit_user]

Why don't manufactures adopt a similar scheme to food stuff's....
Why can't they put a manufacturing date, a sell by date and an expected life span once used.

You mean while it's in the tube, or once it's been applied? Either way, it'll depend a lot on the environmental conditions. Once applied, its lifespan will also be affected by how much it's stressed. So, this makes it difficult to predict how quickly it will degrade.

And... nobody dies if your TIM degrades (well, just talking about home-use desktop PCs, at least). What happens is that you'll get throttling and probably a temperature alarm. If you OC, maybe you'll start getting more crashes and will then think to check your temps. Either way, it'll be pretty clear when it's time to replace your TIM.

 

[Memento]

Did Tuniq’s TX-4 get tested? This i the one I've been using because of an older test similar to this showed it was good, wonder how it stacks up now.

 

[InvalidError]

And... nobody dies if your TIM degrades

Most paste TIMs don't degrade until temperatures in excess of 200C for silicone oil. Excess oil does get pressed out of the TIM over thousands of thermal cycles and after that happens, the TIM becomes susceptible to mechanical shock. In principle, the pressed "dry" TIM should perform better as its particles are packed tighter together.

For tubes of paste though, there is the problem of particles precipitating to the bottom of the tube over time. This can result in a much thicker paste that is more difficult to work with - I have one old tube of zinc oxide paste which leaked a bunch of oil and now I need pliers to squeeze paste out of it. The main two reasons to have oil in pastes is ease of application and provide surface tension to hold particles together. The oil itself at ~10X the thermal resistance of zinc oxide and other common paste materials doesn't contribute much to heat transfer.

 

[FormatC]

On small thing:
The translation is too harsh, in the Original you can read "Sehr günstige Silikonpasten wie Arctics MX-2 oder MX-4", that not means cheap as bad word but something like "not so expensive pastes like...". The problem is, that we have no equal English word for this and this two pastes were described als example. But the MX-2 and MX-4 are outdated products und to be honest - if the PR said something about no burn in, this is not automatically a fact. I doesn't mean, that this pastes are unusable and if I build a entry-level system, it makes really no sense to use an expensive 8 USD TIM for an 15 USD cooler on an 60 USD CPU

 

[bit_user]

that not means cheap as bad word but something like "not so expensive pastes like...". The problem is, that we have no equal English word for this

We usually say something is "inexpensive" if we only mean to say it doesn't cost much. Or you could just say "low-cost".

As you correctly state, "cheap" can imply poor quality, as well as low cost.

If you want to say that something is good for the price, you could use the term "cost-effective". The opposite of this would be something that's "overpriced".

 

[FormatC]

The translations are every time a problem and the readers reacts very often simply too emotionally.
It is like the running gag from South Park: Oh my God, they killed my (place here the company name) paste!
If a thermal compound corresponds with my skills and I'm satisfied with the result, the world is ok.
But I have also to accept, that other products may perform simply better. If I don't need this small plus, it's also ok

 

[cryoburner]

The translations are every time a problem and the readers reacts very often simply too emotionally.
It is like the running gag from South Park: Oh my God, they killed my (place here the company name) paste!
If a thermal compound corresponds with my skills and I'm satisfied with the result, the world is ok.
But I have also to accept, that other products may perform simply better. If I don't need this small plus, it's also ok

The wording isn't the problem. It's more that you called out one specific brand of thermal compound with claims that it has some problems which others here suspect is incorrect. It sounds more like you're describing the characteristics of another thermal compound from a different company with a similar name. And you seem to be basing this entirely on hardware you get from other reviewers, where you probably don't even know for certain which compound they actually used, especially since that compound might have been applied years prior. Maybe it was Arctic MX-2/4, or maybe it was Arctic Silver, or maybe it was something else entirely. It seems unfair and potentially inaccurate to call out one brand of compound from the dozens tested in the article, that you apparently don't even use and only suspect you've encountered on hardware installed by others. Even if there happened to be some truth to it, I don't see the point in naming one brand when you didn't perform any sort of durability testing, and when other benchmarked compounds likely hold up worse.

And on that topic, it's great to see all these thermal compounds tested, but the performance after 24 hours might vary significantly compared to the performance after a year or more. Just because a compound performs well in these charts on a fresh installation, doesn't mean it will hold up well over time. I'd love to see a test that tracked the long-term durability of these compounds being subjected to heat cycles for a few years. It might be impractical to test all of those found here, but perhaps a limited selection of the best performers alongside some of the more popular brands could be feasible.