[eXtreme-roundup] 26 Thermal Compounds tested

[der8auer]

 

UPDATE: TEST 2015 HERE

 

Extreme-Overclockers are always trying to achieve the highest scores possible. No matter if CPU, GPU, mainboard or memory - everything has to run perfect in order to set up new records. The latest generation of Core i7 - Ivy Bridge - scales brilliantly on deep cold. Without cold bug the CPUs are limited by the pot and thermal compound. I will test 26 different thermal compounds in the following test to see how they perform with air cooling and liquid nitrogen.

 

 

Index:

 

 

1. TIM Roundup 2013
   
   • Setup and description
     
   • Thermal compounds in detail
     
     • Alpenföhn Schneekanone
       
     • Coolink Chillaramic
       
     • Noctua NT-H1
       
     • Alphacool Silver Grease
       
     • Gelid GC-Extreme
       
     • Zalman ZM-STG1
       
     • Revoltec Thermal Grease Nano
       
     • JunPus D9000
       
     • JunPus DX1
       
     • Xigmatek PTI-G3606
       
     • Xigmatek PTI-G4512
       
     • Coolermaster Value V1
       
     • Coolermaster Essentials E1
       
     • Dimastech HTX EE
       
     • Dimastech Thermal Compound
       
     • Prolimatech PK-1
       
     • Prolimatech PK-2
       
     • Prolimatech PK-3
       
     • Akasa 455
       
     • Akasa 460
       
     • Akasa 5022
       
     • Arctic Silver 5
       
     • Arctic Céramique 2
       
     • Arctic MX-2
       
     • Arctic MX-4
       
     • ]Amasan T12
       

 

[*]Test 1: Air cooling

[*]Test 2: LN2 at -50°C

[*]Test 3: LN2 at -150°C

 

[*]Evaluation

1. Evaluation of the results
   
2. Awards
   

 

TIM Roundup 2013:

 

Acknowledgment:

 

First I'd like to thank all the vendors who supported me for this test and provided samples to me. Caseking was the most important as the provided the most products to me. Thanks a lot for that!

 

I also received some samples directly from vendors. Thanks to Akasa, Alpenföhn, Coolink, Noctua, Alphacool respectively Aquatuning, der8auer ECC, JunPus, Dimastech and Prolimatech.

Setup and description:

 

Theory

I'm sure you guys already know why we have to use thermal paste. But just in case if there are some new members out there, I will explain why you need thermal paste. So what is thermal paste and why do we need to use it?

 

 

Let's take a look on this simplified image first. You can see a cut image of a CPU and the cooler on top. Greatly enlarged you will notice that the surface of CPU IHS and cooler are pretty much rough even though it looks very flat with the naked eye. The spaces between CPU and cooler have to be filled with thermal paste to improve the thermal conductivity. That's why the thermal pastes are liquid and ideally never dry out. Most vendors guarantee about 2-5 years of usage between CPU and cooler. After this time you should change your thermal paste in case you still have your system after this time

The majority of the thermal pastes are made of different silicones, oils and additives like silver, aluminum, ceramic or diamonds to transfer the heat. Unless other specified the thermal pastes have no electrical conductivity. This applies to all thermal pastes in this test.

Setup

For the first test I'm using a i7-3770K @ 4,0 GHz and 1,20 vCore cooled by a Scythe Katana 2 at 100 % fan-speed. To stress the CPU I'm using the tool CoreDamage. I'm running CoreDamage for about 15 minutes until the temperature is not changing anymore. Two thermal probes fixed on the CPU IHS and the base of the cooler are used to measure the temperature.

 

 

 

 

As measurement value I'm using the delta-temperature between CPU-IHS and the cooler base. Most thermal pastes differ only by few 0,1 °C so measuring the absolute temperature with changing ambient temperatures would crash the complete test. Using the delta-temperature I'm eliminating this source of error.

 

I'm doing each test with each thermal paste three times to exclude mounting issues of the cooler. After each test the CPU and cooler are cleaned using alcohol to remove all residues.

 

For the second test I'm using the new der8auer Fusion rev 3.1 CPU container and liquid nitrogen. The container is cooling down the same i7-3770K which is now clocked at 5000 MHz @ 1,45 Volt to -50°C. Again I'm running CoreDamage to stress the CPU. During the test I keep the CPU as close to -50,0°C as possible. After few minutes if the temperature is not changing anymore I note down the temperature.

The last test is to cool down the CPU to -150°C which is still running at 5000 MHz but on 1,50 Volt to give a little more load on the CPU and produce more heat.

I'm again mounting the pot twice to eliminate measurement errors. If both measurements differ too much e.g. by 3,0 K I'm doing a third test and use the two measurements which are closest to each other.

How to apply thermal paste and mounting pressure

Whereas there are a lot of different ways to apply the thermal paste one thing is clear: the mounting pressure has to stay the same during the test to have consistent results. That's why I'm using the Scythe Katana 2 with Intel Push-Pins for the first test. It's a little more difficult to mount the pot exactly the same but Stummerwinter had a good idea here which I've seen few months ago as he mounted his pot. I'm using 4 springs which all have a normal length of 16,0 mm. Tightening the thumb screws I'm using a sliding caliper to measure the length of the springs until all have the tightened length of 12,2 mm.

 

Applying the thermal paste you have to pay attention not to apply too much. This could have a negative influence on the cooling performance. Still you have to use enough to fill all gaps between CPU and cooler. That's why I'm using a thermal paste applicator to have a very thin layer covering the whole CPU.

 

 

 

 

 

Setup:

 

• GA-Z77X-UP5 TH
  
  
• i7-3770K
  
  
• 2 x 2 GB G.Skill Pi 2400 CL9
  
  
• LEPA 1600 W PSU
  
  
• Scythe Katana 2 (Test 1)
  
  
• der8auer Fusion CPU-Container
  (Test 2 and 3)
  
  

 

Thermal compounds in detail:

 

At the end of this test you will find a huge table including all the technical information and test results. That's why I will not explain the technical details of each thermal paste here.

 

Alpenföhn Schneekanone

Alpenföhn is a well-known manufacturer of cooling components and provides the first thermal paste of the test. Even though the tube is quite big the content is only 1,5 g. I prefer bigger tubes as they are easier to handle so this is a positive aspect. The packaging is very well designed and provides all technical information including the manual on the back. Included in the delivery you can also find an applicator.

 

Coolink Chillaramic

Coolink's Chillaramic straight attracts attention due to the nice packaging. The manufacturer promotes this product to be very good for lower temperatures and the use with Single Stages. An applicator or manual are not included unfortunately

 

Noctua NT-H1

The Austrian cooler manufacturer Noctua provided the NT-H1 for this test. Compared to most of the other thermal pastes this product is very well packed and you can easily open the packaging. I'm not a friend of blister packaging

 

You can find all important technical information on the back and a nice manual. Noctua says it's enough to place a little drop in the middle of the CPU and therefore does not include an applicator.

 

Alphacool Silver Grease

Alphacool provides a quite huge tube with 30 g which is the second biggest in this test. Therefore it's pretty good to use if you're changing the cooler a lot and don't need the best performance all the time. 32 Cent per Gramm is one of the cheapest products here.

Additional stuff like manual, packaging or applicator is not included.

 

Gelid GC-Extreme

Gelid is very well known for the GC-Extreme using extreme cooling methods. The tube which contents 3,5 gram is one of the more expensive products and comes along in a blister packaging which is quite hard to open.

Included in the delivery you find an applicator and all important technical details on the back.

 

Zalman ZM-STG1

I'm sure you all know Zalman for their air- and water cooling solutions. The ZM-STG1 is a pretty interesting product which differs from the normal thermal pastes. It looks like a nail varnish and can be applied using a small brush which is fixed inside the top. Due to the low viscosity of this product you have to apply the thermal paste on both CPU and cooler.

 

Revoltec Thermal Grease Nano

Similar to Zalman's ZM-STG1, Revoltec's Thermal Grease Nano comes along in a small glass bottle and stands out of the test field. It's very easy to apply but is packed in a blister packaging. You can find technical details and a short manual on the back.

 

JunPus D9000

JunPus is a quite unknown manufacturer from Taiwan. D9000 is the first product out of two in this test. It is very well packed and you can find all important technical details on the back including some nice charts. Included in the delivery you find an applicator.

 

JunPus DX1

DX1 is the second product from JunPus. Packaging and content of delivery are same to the D9000.

 

Xigmatek PTI-G3606

Xigmatek is a well-known cooler producer and provided two products for this test. The PTI-G3606 comes along with an applicator and a nicely designed packaging which is also quite hard to open.

 

Xigmatek PTI-G4512

The PTI-G4512 is the high end product of Xigmatek and has a slightly higher performance than the PTI-G3606. Packaging and content of delivery are pretty equal to the PTI-G3606. All important information is printed on the back of the packaging.

 

Cooler Master Value V1

The famous case producer Cooler Master also provides two products for this test. Value V1 is the midrange product which is packed in a blister packaging. On the back you can find technical details.

Included in the delivery you find an applicator.

 

Cooler Master Essentials E1

Essentials E1 is Cooler Masters high end product which is shipped in the same packaging like the V1. The E1 has one interesting feature, though: a cleaning cloth to remove old thermal paste.

Like the V1 the E1 also has an applicator included.

 

Dimastech HTX EE

We all know Dimastech for his cool and helpful Bench Tables made in Italy. HTX EE is Dimas first thermal paste and comes along with an applicator and a well-designed packaging. I would have opted for something different than blister packaging.

 

Dimastech Thermal Compound

For testers with a high consumption of thermal pastes Dimastech offers a big box with 60 g content - the biggest in this test. Included you find two applicators

 

Prolimatech PK-1

PK-1 is the entrance product of Prolimatech which is well known for its cooling solutions like the MK-13. Prolimatech offers the best packaging of all products. Very easy to open, well designed and very detailed technical information including the ingredients. An applicator is also included.

 

Prolimatech PK-2

PK-2 is the midrange product has a higher performance than the PK-1.

 

Prolimatech PK-3

PK-3 is Prolimatechs high end product and available in different sizes like the other thermal pastes of Prolimatech.

 

Akasa 455

Akasa 455 is the smallest product in this test. Technical information is printed on the tube itself - a manual is missing though.

Akasa 460

Akasa 460 is a midrange product of the Taiwanese cooler manufacturer. You can find all necessary information on the packaging even though it's quite hard to open. Included in the delivery you can find a multi-language manual and an applicator.

 

Akasa 5022

5022 is the high-end product of Akasa and the packaging is pretty much equal to the Akasa 460, although the performance is higher.

 

Arctic Silver 5

Artic Silver 5 is one of the oldest products in this test and has been on the market for quite a long time. The 3,5 g tube gets delivered without any packaging, manual or applicator.

 

Arctic Céramique 2

Same as the Arctic Silver 5 the Céramique 2 comes along without any additional stuff. It has been used by a lot of different extreme overclockers in the past years and is a kind of classic product to me.

 

Arctic MX-2

Arctic - don't mess up with Arctic Silver - provided two products for this test. The MX-2 is the cheaper version and comes along in a blister packaging. You can find all important information on the back of the product. The hazard symbol on the back should actually not disturb you as this applies for all pastes in this test.

 

Arctic MX-4

MX-4 is the high-end product of Arctic and like the MX-2 available in different sizes. Except for ingredients and performance both products are pretty much equal.

 

Amasan T12

Amasan T12 is a very cheap industrial product which I used over the past years for quick tests and therefore I'm including it into this test.

Test 1: Air cooling

 

 

Test 2: LN2 at -50°C

 

 

Test 2: LN2 at -150°C

 

 

Evaluation:

 

Testable:

Click to enlarge

 

Evaluation of the test results

 

 

 

After changing the cooler and applying the thermal pastes about 150 times and 80 hours of work - let's go over to the conclusion. Basically there are no bad thermal pastes - it just depends on what you are using it for and your expectations. There is no problem to use all of these products for 24/7 systems. Although you might be ending up with relatively cheap 70 EUR air- or water cooler, don***8217;t save on other parts of your cooling configuration. So which thermal paste for which application?

 

Normal 24/7 Gaming- or Office systems:

You can use all products here without any problems. Cheap products like the Cooler Master IC Value V1 or the Akasa 455 are enough here.

Demanding gaming setups incl. overclocking

Overclocker and gamer with high-end setups should spend few euros on the best thermal paste. You can easily gain 2-3°C upgrading from cheap products. I recommend: Gelid Extreme, JunPus D9000, Noctua NT-H1, Prolimatech PK-3, Akasa 5022, Artic MX-4, Cooler Master IC Essentials E1, Xigmatek PTI-G4512 and Alpenföhn Schneekanone.

Extreme-Overclocker at -50°C (e.g. DICE or Single Stage)

As you will be pretty limited by the temperature you should use one of the top thermal pastes to get the maximum out of your system. Good products are: JunPus D9000, Prolimatech PK-3, Alpenföhn Schneekanone, Noctua NT-H1, JunPus DX1 and Dimastech HTX EE.

Extreme-Overclocker at -150°C and lower (Liquid Nitrogen!) - the supreme discipline

Especially using cold bug free hardware like the Ivy Bridge CPU you should only use the best products available. I'm recommending Gelid Extreme and D9000 from JunPus.

Awards

 

 

 

 

Let's start with the best products of this test. Both

JunPus D9000

and

Gelid GC-Extreme

are great products and reach the lowest temperatures possible. Both products come along with applicators but I prefer the packaging of the D9000 however it's a little harder to apply than the GC-Extreme. For the great performance both products receive the GOLDEN SYRINGE award.

 

 

 

 

Alpenföhn Schneekanone, Prolimatech PK-3 and Noctua NT-H1 also have a pretty solid performance and therefore receive the SILVER SYRINGE award.

Alpenföhn's Schneekanone

means snow cannon in English which pretty much applies to its performance. It's quite expensive but the performance is good even on very low temperatures. The content of the tube is quite low but this can be good if you are using it just for one system.

Prolimatech

PK-3

is a solid product for air cooling and deep temperatures up to -50°C with a great packaging.

Noctua***8217;s NT-H1

convinced me in all three categories and the packaging is also very well designed. Both 24/7 overclockers and extreme-overclockers can use this product without compunction.

 

 

Cooler Master IC Essentials E1

offers a great price/performance. The cleaning cloth which is included in the delivery is a very useful tool and I hope other manufacturers will add this to their products.

Akasa 5022

also offers a very good performance in all three tests and I can recommend this products for air- and water cooling as well as for 24/7 overclocking.

JunPus DX-1

is also very interesting for 24/7 overclocking and extreme overclocking at temperatures of about -50°C. Taking a closer look on the test table you will notice that this product has the highest thermal conductivity with 16 W/(m*k) even though it did not achieve the best temperatures in this test. According to JunPus the DX-1 has a very long burn in time and will evolve the high performance after a few days of usage. Due to lack of time I could not test this unfortunately.

Edited August 31, 2015 by der8auer

[der8auer]

Finally got it online. Don't kill me if there are too many spelling mistakes

 

Please report back if any graphics are not working.

[ME4ME]

wow, so much work. Thanks for the review!

[I.nfraR.ed]

Strange, but according to my testing GC-Extreme and Ceramique 2 were on par with 3770K at -186 and 6+ GHz, while the Noctua seemed to be completely frozen and the clocks were pretty bad.

And I tested all of them twice each.

[Turrican]

nice review roman.

[der8auer]

wow, so much work. Thanks for the review!

 

nice review roman.

 

Thx

 

 

Strange, but according to my testing GC-Extreme and Ceramique 2 were on par with 3770K at -186 and 6+ GHz, while the Noctua seemed to be completely frozen and the clocks were pretty bad.

And I tested all of them twice each.

 

Mhhh. How did you test? Link to review?

[Farjam]

you did a great work on this review,it is good to see the test of different kind of thermal Compounds

[I.nfraR.ed]

Mhhh. How did you test? Link to review?

 

No review, just a personal testing in order to decide which thermal compound I should stick to. Unfortunately I was only measuring the pot base temperature with the probe I usually use. No IHS readings, but I used the maximum clocks as a measure, don't even remember what was the bench. Can't provide any real data

 

Tried GC-Extreme and Ceramique 2 with Deneb at full pot as well and didn't see any difference there, too.

[Olesius]

nice job dear roman

[Mr.Scott]

Great review and test method.

Surprised no Diamond paste was tested.

[der8auer]

you did a great work on this review,it is good to see the test of different kind of thermal Compounds

 

Thanks

 

 

No review, just a personal testing in order to decide which thermal compound I should stick to. Unfortunately I was only measuring the pot base temperature with the probe I usually use. No IHS readings, but I used the maximum clocks as a measure, don't even remember what was the bench. Can't provide any real data

 

Tried GC-Extreme and Ceramique 2 with Deneb at full pot as well and didn't see any difference there, too.

 

Damn okay I didn't do any testing about MHz. Just temperature difference.

 

 

nice job dear roman

 

Thanks mate

 

 

Great review and test method.

Surprised no Diamond paste was tested.

 

JunPus D9000 and JunPus DX-1 are diamond particle pastes I didn't find any information about the ingredients of the GC-Extreme but just because of the look I'm quite sure it's also diamond based or at least there are some diamond particles included.

[I.M.O.G.]

Nice work, thanks for sharing. Good to have more information to consider.

 

No review, just a personal testing in order to decide which thermal compound I should stick to. Unfortunately I was only measuring the pot base temperature with the probe I usually use. No IHS readings, but I used the maximum clocks as a measure, don't even remember what was the bench. Can't provide any real data

 

Tried GC-Extreme and Ceramique 2 with Deneb at full pot as well and didn't see any difference there, too.

 

There are a ton of variables which are difficult to control that make this sort of testing on a scientifically accurate scale somewhat arbitrary, unfortunately. When dealing with pastes, Roman mentioned near the conclusion that the differences are small enough that pretty much any paste "will do" under casual 24/7 conditions ... The differences are small. On air, from best to worst the difference was 2.6 degrees. On mild subzero, the difference was 1.95 degrees. At -150, the difference was 6.6 degrees... The inconsistencies in this variance from air, to subzero, to way subzero also imply there is difficulty in accurate measurement. Logic suggests that the spread should be smallest on air cooling, medium on light subzero, and largest on real subzero as the pastes freeze and some take on far worse thermal characteristics.

 

Surface tinting from multiple paste applications are one factor. The goal of paste is to fill the microscopic gaps, but not interfere with direct contact. You can clean a base between applications, and it can look clean, but on a micro level that isn't quite the case. Each test is influenced by those before it to some degree. Probably not enough to matter, but, it can't be easily quantified or reported to know for sure.

 

Then there's also application technique. Over the past decade, a lot of people have analyzed this - both in the TIM market and independent reviewers. Many of us have made up our own mind about how we best like to apply paste, but to look at the applicator/spreading technique used in this review specifically, it has been covered numerous times but the most prominent in my mind are 2 articles by Joe Citarella. Generally, a pea in the middle has been shown to produce the best spread characteristics without trapping air inside the interface and with consistent coverage across the surface.

 

This one is about 5 years old:

http://www.overclockers.com/applying-thermal-grease/

 

And this one is sort of an updated follow up to that work:

http://www.innovationcooling.com/applicationinstructions.htm

 

We also shouldn't overlook the calibration and resolution of the temperature reading equipment. Without calibration of the probes one can't be certain how accurate they are to reality. They may be rated for a certain degree of accuracy but this isn't that dis-similar to the issue we have with Kill-A-Watt meters - depending on the specific unit accuracy can be pretty different and its not an absolute measurement, but more of a ballpark figure. If they were calibrated, what temperatures were they calibrated to? Accuracy varies a lot at room temperature to -150. And resolution comes into play here as well - the temperature unit may display in increments of .1 degrees, however does it really have the capability to reliably measure and report variances on that level? I dunno. With looking at paste where the differences can be as small as a few degrees or even tenths of a degree from one paste to another, it is hard to be absolutely conclusive.

 

I don't think that makes the effort or results any less admirable, or even any less useful to look at - these are just things to keep in mind, and a testament to the likely impossibility to be able to reproduce the results in another independent test bed (as if anyone else is going to put 80 hours and hundreds of mountings into retesting them all). For pastes that are close in this list, anything could probably swap positions... For pastes that are far apart in this list, there is more likely a meaningful difference.

 

Props again on the testing. I don't mean to take anything away from the work done. This sort of testing is a herculean effort, and I like to see it. It's better to have some real data than just waffle about conjecture. I just keep some things in mind anytime I read about paste testing because the results different people get are almost always different, and reasons like those I listed are why - it is really, really, hard to do in an absolutely accurate fashion.

[Mr.Scott]

JunPus D9000 and JunPus DX-1 are diamond particle pastes

Excellent. Did not know that. Thanks.

[der8auer]

Matt you're posting exactly what I was thinking about this test. I thought a lot about the different aspects and error sources of this testing method.

 

To answer your question: My thermometers are calibrated on 0°C.

 

You are right that Type-K thermocouples have a higher error rate than 0,1°C but that's no problem here as I'm using delta temperatures. I thought the same like you and did some testing because of that. I used water with different temperatures and put both probes inside to check whether the temperature is changing but it was not the case. This means the error of the thermocouples is always the same during the complete test.

 

It might be that the delta temperature is not correct. It could be 1°C more or 1°C less but as all values have the same error it doesn't matter. You can just compare the results and check which paste has the lowest delta temperature.

 

I agree that the high-end pastes are so close to eachother that propably everything could change their position. Just take a look at Gelid on LN2 -50°C test. Not ranked the best like in the other tests. Nevertheless it provides some information about which pastes to use and which not like you said

 

 

edit:

 

Excellent. Did not know that. Thanks.

 

You can also check the test table:

 

I put all data in there including the ingredients if I could find them

Edited March 24, 2013 by der8auer

[Mr.Scott]

Thanks again. My bad for being too lazy to read it.

[Christian Ney]

I started long ago to test a lot of Thermal Compounds too but my methode seemed to not be as accurate as I wanted.

 

Y u no tested without themal paste at all ?

[K404]

That is AWESOME. Thanks Roman!!!!

 

I'd like to add some info for readers..... there is more to consider than just thermal conduction at a given temperature..... you will also need to consider the characteristics of the component and the benchmarks you will be running.

 

For example: G80. Hot silicon, with a coldbug and benchmarks with short periods of low-load in-between tests. The best thermal paste for them is a SLOWER one so the card does not crap out Using Romans table, a good paste for this situation can still be picked

[Yamunsa]

very nice work there fella. solid review.

[Perica_barii]

Great work, thanks for your effort to bring this to us....

[Sweet]

Very Good review Roman , I can take me one hours to read everything in English but is very instructive, really great work here :ws:

[Guest Sinebrychof]

This is probably the best thermal paste review I've read so far!

[dinos22]

Great review mate. I've always liked guys sitting down and doing tests like this for themselves to determine what is truth and what's BS. Thanks for sharing! The process of discovering is what is really interesting and rewarding. I hope it inspires others to approach their benching or testing methodology the same as you have done this test. Threads like this certainly give me motivation to work harder at everything I do :celebration:

[Xtreme Addict]

Great job

 

From my experience mostly there isn't big difference in max screen and so on, but thermal pastes differently "hold" stress on ln2. With Gelid I mostly get much higher MHz clocks in benchies which test Physx like 3DMark11 CPU test or 3DMark Firestrike.

 

I would also test EVGA Frostbite "no name silver based thermalpaste in Taipei which few many Taiwan overclockers used for years". I really like this thermal paste

[Jimba]

great review!

 

great effort and easy to understand while still having the important information.

 

so have you got any spares?

[DaCoSa OCX]

Excellent review mate!!

 

thanks for sharing!

 

regards