Dear forum members,
We are delighted to inform you that our forums are back online. Every single topic and post are now on their places and everything works just as before, only better. Welcome back!


Discussion on Article:
Thermal Interfaces Roundup: Retail Products

Started by: 3Dkiller | Date 03/08/11 05:05:22 AM
Comments: 39 | Last Comment:  11/28/13 01:05:35 PM

Expand all threads | Collapse all threads


Good review just what i was looking for thanks now i can make my choice.
0 0 [Posted by: 3Dkiller  | Date: 03/08/11 05:05:22 AM]
- collapse thread

0 0 [Posted by: Jordan  | Date: 03/08/11 06:39:44 AM]

Not convinced with the article's assertion that you should spread the compound thinly over relying on the pressure from the cooler.

Trying to spread it out uniformly across the processor manually, generally leads to using more compound than is required which promotes lateral heat spreading instead of just direct to the cooler. It also increases the chance of introducing bubbles of air into the system.

That said, I don't have concrete proof of this. It'd be quite nice if someone with the time and resources proved whether I was correct using some kind of solid methodology - hint hint.
0 0 [Posted by: hahnchen  | Date: 03/08/11 06:50:07 AM]
- collapse thread

I agree. Every other expert I've read so far and even many manufacturers will advise against spreading the paste.

It's weird the reviewer recommends a practice almost everyone else is against to.
0 0 [Posted by: Aleve Sicofante  | Date: 03/10/11 08:52:13 AM]
Manufacturers recommend to apply "drops in the center" for regular users - it should be easy and convenient. Any professional will tell you that (especially with modern dense greases) it is always necessary to distribute a uniorm thin layer of the compound over the entire surface. Under the pressure of a cooler it may not spread evenly and fill all empty space. Judging by you are saying, I am not surprised that you may not have known that.
0 0 [Posted by: Jordan  | Date: 03/12/11 02:49:26 AM]
You need to actually prove this instead of just of the ad hominem attacks.

You already have your thin layer results, shouldn't take that much time to settle it.
0 0 [Posted by: hahnchen  | Date: 03/13/11 07:25:39 AM]
I don't need to prove anything to anyone. Of course, I studied difference ways of applying thermal interfaces of various density before testing, and I concluded that the approach I use in the curent article as well as in the previous ones is indeed the most correct one. Those who think the opposite should try putting a drop of such product, as Arctic MX-3 or MX-4, for example, in the center of the CPU heat-spreader and then check out the imprint
0 0 [Posted by: Jordan  | Date: 03/13/11 09:15:09 AM]
It is perfectly visible that only liquid thermopastes are more or less normally distributed at drawing by a drop or strip method.

Much more densely, therefore them it is necessary to put all modern thermopastes here so:

By the way, some manufacturers (for example ZEROtherm) recommend such method.
0 0 [Posted by: Jordan  | Date: 03/21/11 04:31:35 AM]
did you actually WATCH the top video that you linked (@1:55)? This demonstrates pretty conclusively that the spread method traps air. It may not directly say that this is bad, but it is I guess also worth pointing out (that a couple commenters to that video missed) that the acrylic thing he was putting down was meant to simulate a heatsink. He wasn't advocating spreading it with that, removing it, then attaching your heatsink. Basically the video guy's point was a) spread method is awful b) make sure you use enough compound and c) don't use too much compound. Multiple dots, lines, globs, smiley faces...whatever - they don't matter.

Also, look at the comments to video 2. Look at how the video has received more thumbs down than up (not that the masses are always right, but if youtube viewers think something is awful, typically it is). Note the second-highest rated comment. Also, WTF is the point of making a pretty shape if you're just going to spread it everywhere manually anyway?

You challenge us to look at your pictures and ask us to let you know if we see anything wrong with the coverage you have. The point is we can't. You've already removed the heatsink and things never seperate evenly. We can't see where the air bubbles were. You point out that with the dot method, the corners are sometimes not covered - this is a good point and likely leads to sub-optimal heat transfer. However, the chip manufacturers as well as the thermal compound makers will tell you that MOST of the heat comes out the center region (because that's where most of the processing occurs). Additionally, many heatsinks have a round base that is slightly smaller than the square chip anyway, so it doesn't matter if those corners are uncovered, in fact it will help A BIT if they are uncovered (heat transfer from the HS to air vs HS to TIM to air). Of course, if you had a square base or circumscribe the chip with the base of the heatsink, it'd be better to go out all the way to the edges, but that's where the line or cross methods come in handy.

I'm actually concerned with the apply-pressure methods (dots, lines, etc) as well, because I figure if I initially apply pressure a bit more on the left side, it'll squirt the stuff to the right, not leaving much to spread to the left. I assume the high viscosity of modern pastes corrects for this a bit, which is why manufactures (of chips, heatsinks and pastes) don't really mention it, other than to suggest that you apply diagonal pressure. In any case, a spread method is undesirable, so your best bet is learning the proper technique of applying equal pressure and application of an ideal amount of TIM with a dot/line method.
1 0 [Posted by: mp  | Date: 06/29/11 01:58:20 PM]
I also agree. Also minimizes the chances of air getting trapped. Built about 60 overclocked systems that are currently in service, and re applied thermal compound thousands of times and never spread it out!
0 0 [Posted by: csimmo  | Date: 03/11/11 11:03:08 PM]
It is a pity. You could achieve much more impressing results in overclocking.
0 0 [Posted by: Jordan  | Date: 03/12/11 02:51:24 AM]

What is your control? What is your temp without any paste?
0 0 [Posted by: rlthomas  | Date: 03/09/11 05:41:15 PM]
- collapse thread

Read article attentively. Without thermal grease the OC processor is stable under load. Nobody uses processors without thermal interface, that is plain silly.
0 0 [Posted by: Jordan  | Date: 03/10/11 12:11:43 AM]

Also, if manufacturers are stating a certain burn in period, why are you not taking the greatest common denominator? It's not more than a few hours so laziness (lack of time) should not be an answer.
0 0 [Posted by: rlthomas  | Date: 03/09/11 05:47:21 PM]
- collapse thread

Once again - read the article attentively! Each of the tested pastes was in action fo at least 2 hours total. It ws impossible totet each of them for 72 hours due to a large number of participating products.
0 0 [Posted by: Jordan  | Date: 03/10/11 12:13:28 AM]

I'm unsure, I stumbled across this site via TechReport and I'm not impressed so far. I registered to comment and hopefully you'll improve but there is a reason Anandtech and TechReport are so highly regarded.
0 0 [Posted by: rlthomas  | Date: 03/09/11 05:48:41 PM]
- collapse thread

Lie. Our article is way better, and the testing techniques are more thorough on some usage models, than at Anandtech or TechReport.
0 0 [Posted by: Jordan  | Date: 03/10/11 12:14:54 AM]
Lie??? What an elegant and professional way to reply...

I find this enough not to trust you.
0 0 [Posted by: Aleve Sicofante  | Date: 03/10/11 08:54:53 AM]
Simply try to compare techniques of testing and to think. And the trust of those who is lazy it to do us simply doesn't interest.
0 0 [Posted by: Jordan  | Date: 03/10/11 08:18:11 PM]
Wow, you are terribly arrogant. I hope you're not running XBitLabs yourself and I also hope some adult is supervising you and take appropriate measures about your complete lack of professionalism.
0 0 [Posted by: Aleve Sicofante  | Date: 03/11/11 05:05:57 AM]
Don't worry so. Probably, when you will get hand in learn to perceive adequately our materials and to penetrate into testing techniques. Otherwise, to us it is valid there is nothing to discuss.
0 0 [Posted by: Jordan  | Date: 03/11/11 08:48:52 AM]
Your English is VERY hard to follow. May I ask what's your role here at Xbitlabs? An occasional collaborator? The owner? The main editor? I just need to know if I should keep reading the site or deleting the bookmark for good, you know?
0 0 [Posted by: Aleve Sicofante  | Date: 03/19/11 10:26:23 AM]
This your business to delete a site from your bookmarks or not.
I write articles about computer 8 years completing already more and I will write and further, what claims at you here wouldn't be.
0 0 [Posted by: Jordan  | Date: 03/19/11 11:10:38 AM]

Sorry if I missed it, but how did you guys apply the TIM on the card? I've always used a plastic baggy since it's easy to get coverage right. But lately, I've been using the dot + pressure method, check for contact...adjust size of dot.

0 0 [Posted by: Jodiuh  | Date: 03/09/11 11:44:24 PM]
- collapse thread

All thermal pastes were applied to a surface in a uniform minimum layer. We have mentioned it in our review as well.
0 0 [Posted by: Jordan  | Date: 03/10/11 12:16:49 AM]

couple of reasons idigio extreme wouldn't have worked for you.

1. bottom of heat sink needs to be 1 thing (ie. a solid copper block), looks like the heatsink you used base is made of heat pipes connected (with minute gaps between)

2. sounds like you got it no where near hot enough for reflow, you are meant to fire all cores on your cpu (for a six core it says you need 12 instances of CPU burn-in program running)

3. having fans running on your heatsink (or near) would also severely limit it's ability to reflow "melt"
4. also indigio suggests using spring mounted/pressured heatsink.

as Indigio Extreme is one of the better thermal interfaces around at the moment maybe you should review it's instructions and redo it's section of your review. (anyone interested in Indigio extreme should review all documentation regarding it and your heatsink before purchasing)

(also for large air cooled heatsinks i have heard people have used heatguns to quicken heating the massive amount of metal to achieve reflow)
0 0 [Posted by: Bullus  | Date: 03/10/11 01:51:55 AM]

Concerning the Indigo Xtreme, I agree with Bullus, especially his #2 point about your not getting the cores hot enough for the reflow.

I haven't used the Indigo Extreme yet for my new build (I'm waiting on mobo to arrive), but I've read the instructions and closely watched their posted video. When I read what you did, it sounded wrong right from the start were you where talking about 90ºC and the like.

Instead of the video you reference in your article, you ought to watch THIS video: That is their 'official' video demonstration for how to install it, NOT the one to which you linked.

OTOH, I know this report was not an easy one to do. Actually it is probably one of the most difficult I have yet seen. In most of it you should be commended for your hard work, explanations, and comparisons. Thx!

Two other things. Besides the CPU's they list at their website (, I have an email from Enerdyne Solutions which states the Indigo Xtreme made for the LGA1156 chipset will work fine for the new Sandybridge LGA1155 chipset.

Crazy PC has reduced their price from $17.99 to $14.95 at
0 0 [Posted by: Tim the Enchanter  | Date: 03/12/11 01:29:14 PM]

I have to agree with #2 & #4. Even arctic silvers instructions for application says to use the "dot" or "line" method when applying the paste.
As for #4, having used both arctic silver and mx2 a lot its my experince that the burn in time specified by arctic silver is not exagerated. It takes at lot longer to burn in than at least the mx2. So im not surprise about your results, as 1 hour of burn in doesn't make any difference at all with that compound. I would expect the temps to be at least around mx2 & 3 after the correct burn in time and probs. I've also found that AS retains its heat transfer properties longer than at least mx2 meaning you dont have to reapply as often, which is why im staying with it myself.

But anyways, respect to you for doing a behemoth of a test like this, not only is it time consuming as hell but taking on testing so many tims will always result in a lot of ppl commenting on methodology and what not its kinda like religion and i dont think its possible to do to everyones satisfaction, not even if you spent a year

0 0 [Posted by: Futte  | Date: 03/16/11 05:42:40 AM]

Look, Sergey Lipilov or Jordan (if you guys worked on this project together), I really do appreciate all of your hard work. But..

If you guys insult the overclocking enthusiast community by treating all of us like idiots saying that we're so dumb for doing the dot/line pressure method, what is really happening here is that the gun is backfiring right at you as we feel that the disrespect is totally unwarranted, so we in turn disrespect your absolute conclusion that your paste-spreading method is superior to our many years of experience with experimenting with the dot-pressure method.

I know it was a lot of hard work, that you're probably feeling frustrated with all of the comments here, but it does not mean you have to blow us off and treat us like idiots to the point where you completely ignore everybody else including the LN2-pouring experts.

My own experience is that both Tuniq TX-2 and Thermalright Chillfactor II beat Arctic Silver 5, but with a twist to it (the difference became clear only if I already used AS5 on the heatsink). After the first time AS5 was applied to the heatsink and IHS, you do not even need to worry about the curing time if you reapply AS5 because the TIM is already in the grooves of the heatsink block, even after you wipe it off. With the "ideal" silver already in the grooves, it's a better idea to use the highest thermal-conductivity paste between the heatsink and the chip to do the rest of the work. The silver particles in AS5 are rather coarse and large-sized, so using another TIM to make its way further into the grooves around the large silver particles being stuck in the cracks further boosts the overall thermal conductivity. I know silver is simply the best conductor next to diamond, but AS5 isn't pure silver (80% silver by weight, which probably means around 30% silver by volume) and the rest of the material is not as good as that of MX-4 for example. AS5 wins out in this 80-way shootout benchmark:

If you used the "five-dot" (the middle dot being 1/2 the size of a BB-gun pellet, and 4 smaller dots placed around the middle dot, halfway in between the middle and the corners, about half the size of the middle dot) method without manually spreading it out onto the CPU, it would have worked wonders. Why leave out the MX-4? If you used your lapped IFX-14 (my IFX-14 is also lapped), you wouldn't have to worry about the heatsink being concave. Also, for the heatsinks with direct-contact heatpipes, it's usually recommended by the manufacturer to use the stripe method.

I wish that you could've consulted the community about Indigo Xtreme before making such a negative review about it. A reviewer needs to be as well-informed as possible before jumping to such a conclusion.
0 0 [Posted by: Bo_Fox  | Date: 03/16/11 06:33:06 PM]
- collapse thread

Have you seen thermopaste imprints on the processor or a cooler after applying it in a drop, cross or stripe? They will NEVER turn out as uniform, as in case it is applied over the entire surface. If you apply it as a drop in the center or a stripe across there will always be empty spots and the heat transfer efficiency will be lower. Could you please tell me if you have any complaints about the thermal interface imprints described in the roundup? Do you believe they are of poor quality or incorrect? Do you think the quality of thermal imprints in our case affected out overclocking results or CPU thermal modes? It is the end result that is important. And it, certainly, is better than if I apply one drop in the center or a stripe across. I don't think you convinced me. Just try it for yourself and compare the results, and you will understand what I mean.
0 0 [Posted by: Jordan  | Date: 03/17/11 12:52:34 PM]
Yes, on my heatsinks and IHS of CPU's. I've changed out CPU's maybe 30-40 times within the past couple years, trying to find a golden stepping.

The imprint is as uniform as I wanted it to be, and better because it means improved integration into the micro-cracks of the heatsink. The 5-dot method cools it 2 degrees better than the spread method for my lapped IFX-14 with Gelid GC-Extreme. I checked the ambient temperature and it was same to within one degree Fahrenheit. BTW, Gelid GC-Extreme turned out as the best one besides liquid metal:

What I do is press the heatsink as hard as possible onto the CPU (without breaking anything or bending the motherboard) and then twist/rotate it back and forth a few times. After that, the mounting has to be secure enough that the heatsink cannot move or be twisted. That's ok if you could still twist it slightly with some force, but it should not be easy.

I'm not trying to convince you or talk you down. It's just what I've seen myself (assuming that both my eyes and the numbers do not lie), and I only wish to share it, that's all. Hope it helps, dear readers.
0 0 [Posted by: Bo_Fox  | Date: 03/23/11 09:06:37 PM]
0 0 [Posted by: Jordan  | Date: 03/23/11 09:51:33 PM]

I still don't understand the expression "liquid metal". The only liquid metal I know is hydrargyrum/mercury/quick silver, which is a highly toxic product. Surely cannot be this product, so I see here mislead info from the manufacturer...
0 0 [Posted by: TAViX  | Date: 03/18/11 03:42:04 AM]
- collapse thread

I think they use alloys of gallium and indium, but of course I can't find the sources for this so... I could be totally wrong on this. I'm very sure it's not mercury though, I can't imagine that getting through regulations.

I've used "Coollaboratory Liquid PRO" as sold by FrozenCPU, which being made in Germany and sold in pretty much an identical blister pack is probably just a rebrand of what was tested. I like it a lot, but it can be a pain to apply and especially to remove (you usually must re-lap the metal surfaces if you don't plan to use the liquid metal again).

One very important thing that needs to be mentioned, do NOT use these liquid metal formulations on aluminum! The alloy prevents the aluminum from forming it's normal oxide layer, so your aluminum heatsink and/or mounts will disintegrate as they literally "rust" away!
0 0 [Posted by: xrror  | Date: 04/01/11 01:32:04 AM]

It would have been nice to see a dumb control TIM in the line up as well. When I evaluate a new TIM then I have to admit that I use marmite as a control as it's consistatly the same C/W. It does make the lab smell a bit funny though.
0 0 [Posted by: dwbrown  | Date: 03/23/11 02:56:14 AM]
- collapse thread

0 0 [Posted by: Jordan  | Date: 03/23/11 07:04:50 AM]

Hi; First of all, thanks for this awsome thermal paste review!
I have been using Tuniq TX-2 in the past and it's pretty easy to apply, until I purchased TX-3, and now TX-4... which are super waste of money.

Thermaltake TG-1
Thermalright ChillFactor III
Gelid GC-Extreme

Between these three, which one them will you rate them as the easiest to apply?

0 0 [Posted by: gakkuken  | Date: 06/17/11 11:59:41 AM]

This is kind of an old article, but most products are what’s currently available in 2013.

I'm writing this (2 years later) on behalf of any readers that would have some drought’s about the methodology or techniques used in Jordan’s tests as to how YOU the reader and superb computer enthusiast should read them.

Whatever technique you use to spread thermal paste doesn't matter here!!! Hoooouuu, this was quite brutal I'm so sorry (not)!

What really counts: Consistency!

For every product tested, Jordan used the same technique, with very similar amount of product in the exact same conditions. What you smart readers get are numbers, showing you how the products performs in these conditions.

Your job: Interpret these results for what they are. Your super techniques might bring different results, yes. You know best since you probably played with as much CPUs and GPUs then our Xbitlabs friends, right? Help yourself and share your own results and techniques and help us grown smarter! Tell us what you've done and how you did it. Be useful!

Thank you very much Jordan for taking the time to test these fine products. Thanks for sharing your techniques and posting the nice pictures showing your thermal imprints and product properties. Please don’t ever get bored of doing this since I don’t have the time and money to buy and test that much stuff. I actually really need you guys.

Personally, I favor Artic Silver MX-4, Tuniq TX-/-4 and the Gelid GC-Extreme paste. Like Jordan, I like to spread a VERY thin layer of paste all across the cpu using a sharp razor blade. How do I choose what I buy? I buy the one that's the cheapest when I need some and know what; my computers purr like kittens.
0 0 [Posted by: MHudon  | Date: 09/06/13 07:24:47 PM]

One way to end the spread vs. bead/line method debate - run the same tests for a 2013/2014 heat sink compound comparison using both methods. It's a lot of work, but worth it since this is such a contentious subject.

As Mhudon pointed out, the fact that the HSC was applied consistently with all of the HSCs under test matters, despite some users preference towards using bead/line, and doesn't invalidate the benefit of this test. What some users are saying however is that you may get better thermals with bead/line.

I personally have tried both, and haven't experienced any appreciable difference in temps with either method, but it could be due to multiple factors, such as the way I use the spread technique, the amount of HSC, and the type of heatsink itself or the way the temps are measured. One thing I did notice was that you will get a thicker amount of HSC using bead/line that you would if you applied a thin layer overall. The youtube video with the acrylic block shows air pockets, but IMHO, the HSC layer was too thick in that example.

In defence of AS5: It does require burn in time with thermal cycling to properly set. It would be interesting to know how it compares when it is properly set to the others. But it is not quite an honest rating when it hasn't properly cured. This test also doesn't factor in HSC longevity.
0 0 [Posted by: Dr_b_  | Date: 11/28/13 01:05:35 PM]


Back to the Article

Add your Comment