Thermal Paste Comparison, Part Two: 39 Products Get Tested

[Calculatron]

No need for pressure. The packing and separation process can occur even inside a tube resting on a table. It will simply take a few years instead of a few weeks or months - colloidal suspension does not last forever.

As for evaporation, silicon is the base of many greases and does not evaporate. When you pull a HSF off a CPU that got pasted with a silicon-based thermal grease after a few months, you will usually see a ring of "wet" transparent stuff around the clay-like paste patch. That's the silicon oil that got pressed out of the thermal interface over time.

The process will happen regardless of pressure but more pressure does make it happen faster.

"pump out" ? - http://www.nordsonefd.com/images/comp300.jpg

 

[maestintaolius]

All the hype aside, could Tom's include an actual industrial "Silicone Heat Transfer Compound", such as the one by MG Chemicals :http://www.mgchemicals.com/products/greases-and-lubricants/thermal-greases/silicone-heat-transfer-compound-860/, any of the future comparisons. If this stuff is good to use in mass industrial applications, how can it not be any better then let's say AS5.

Plus, AS5 can short out your mobo if some of it gets on the PCB and it costs 10 times as much as the industrial stuff and I don't think it is 10 times better.

A few things:

1) Have you looked at the technical specifications of that particular paste ? It says, the thermal conductivity is only 0.773-0.657 W/mK, which is about a order of magnitude LESS than tested products (AS5 at 9 W/mK and MX4 at 8.5 W/mK).

2) The compound is only made from zinc oxide and a binder oil. This is no different from the cheap Zalman compounds (the generic white paste).

3) These pastes are meant for low power density applications, I have 2 tubs, and they are piss poor on computers (high power density applications), but just fine for what they are made for - cooling very low powered circuits like voltage regulators (not for computer use, but analog use), opamps, low powered transistors (like for amplifiers). This is the stuff you find inside your stereo, between the transistors/chip amps and the heatsink, where the chip is dissipating about 50 watts of heat maximally (and not prolonged) over a surface area 10 times larger than a CPU-heatsink contact.

4) The AS5 isn't conductive. It doesn't contain un-oxidized silver. It is slightly capacitive, more so than other compounds. But it isn't going to burn out a computer by shorting it, it is not a conductor. That said, I much prefer Ceramique 2 (the original Ceramique was terrible) or MX-4.

In the case of grease application TIMs, bulk thermal conductivity is a meaningless number, it's mostly just marketing, I could run down to the lab and make you a grease that was 20 W/m-K in about 20 minutes but it'd completely fail you as a thermal grease. For greases, the more important number/chart is the thermal resistance vs pressure. The best greases available will actually have fairly low bulk thermal conductivity as the goal is to have a low viscosity with a small filler size, allowing the compound to go to thin bond line. The best way to accomplish that is to have low filler levels with very small particles, as small particles have high surface area:volume ratios you can't load them too heavily or the viscosity shoots through the roof and it takes too much pressure to reach bond line.

That said, ZnO fillers are poor, typical high performance greases will use ATH, ALN, BN, alumina, aluminum or liquid metals. The ZnO filled greases are usually just lubricating greases that the company also happens to dual purpose sell as a thermal grease.

I didn't see any data on their website about the thermal resistance or conductivity vs. pressure. It seems like you may be suggesting that these thermal greases are non-Newtonian fluids ? If the viscosity sharply changes as a function of load, I mean. My understanding was that these were very much Newtonian fluids with a well defined viscosity, because of the non-polymerized state of the internals.

Conductivity is a bulk property that is not pressure dependent. However, the overall thermal resistance in application is going to be pressure dependent as it will be effected by the final bondline thickness and the interfacial resistances. It doesn't surprise me that it's not readily available as there's no official standard, usually a D5470 capable machine is used in a pressure control mode and the resulting interfacial resistance is recorded. We will typically run a pressure vs thermal resistance spectrum starting at 10 psi and ending at 500 psi.

I'm also not suggesting they're nonnewtonian, I'm flat out saying it. The silicones used in TIC greases are newtonian and not typically crosslinked, but the filler level is so high that they rapidly lose their newtonian behavior, particularly once you get to bondline. Highly filled materials have very large wall slip issues (which may give the appearance of newtonian behavior in capillary or parallel plate rheometers and makes for nice data sheet values) and once your gap becomes on the order of 10x the largest filler particle size, particle-particle interaction comes into play causing a further deviation from newtonian behavior. Below a gap of 5x your max filler size, particle-particle becomes a dominating contributor to rheological behavior and typically pressures become extreme at 2-3x which ultimately limit your minimum bondline thickness.

Usually, when you see a viscosity quoted on a data sheet is a value taken off a brookfield using a particular spindle at a particular RPM (T-F @ 20 RPM is pretty popular), its should never be assumed that the visocosity is newtownian.

 

[maestintaolius]

It seems like you may be suggesting that these thermal greases are non-Newtonian fluids ?

If you compress thermal grease/paste to the point where its solid contents that were in colloidal suspension in the grease/oil/whatever prior to application start packing into a solid mass, it isn't a fluid anymore.

You don't even have to go that far, once you start getting over 10 volume percent filler newtonian models start to break down depending on filler geometry.

 

[maestintaolius]

No need for pressure. The packing and separation process can occur even inside a tube resting on a table. It will simply take a few years instead of a few weeks or months - colloidal suspension does not last forever.

As for evaporation, silicon is the base of many greases and does not evaporate. When you pull a HSF off a CPU that got pasted with a silicon-based thermal grease after a few months, you will usually see a ring of "wet" transparent stuff around the clay-like paste patch. That's the silicon oil that got pressed out of the thermal interface over time.

The process will happen regardless of pressure but more pressure does make it happen faster.

"pump out" ? - http://www.nordsonefd.com/images/comp300.jpg

Actually, its not correct to say they don't evaporate, depending on the type of silicone used there can be a substantial volatile portion. Vapor phase migration of silicones is a major concern for a lot of customers, particularly ones with optical connections and hard drive manufacturers. One of the ways to improve performance of a grease is to increase filler loading, but that increases viscosity, which usually prompts using a lower molecular weight silicone with a lower base viscosity to make the grease usable. I've seen greases that used silicones that were 8% volatile by mass, any silicones below D20 or L20 can, and will, undergo vapor phase migration. Silicones can be stripped of these lower molecular weights using distillation methods, but that makes them much more expensive and they're not available to retail consumers.

Liquid phase migration takes many forms, settling/separation and bleed are two dominate forms. Bleed is typically what you'll see when you dispense some grease on a surface and you can come back a day later and see a hazy halo surrounding the bead. Separation and settling can occur based more on filler loading, geometry and types of filler used. Settling happens a lot in greases because bondline is what dominates grease performance which means small fillers, which means low filling ratios (to hit a target usable viscosity) which means there will be settling. If you have a lot of fillers (like you do in gels and pastes) the fillers will interact with each other and slow or stop settling. Also, the lower the visocity of the resin, the more likely separation can occur.

Bleed, well, bleed happens with everything, pads, gels, greases, all of them bleed to various degrees and for various reasons. Bleed has been something I've been working for about 7 years now and could end up writing a book concerning it.

Pump-out is a kind of phenomena that occurs commonly with thin bondline applications and can be affected by liquid and vapor phase migrations. Generally, what happens is the TIM starts to separate (but separation isn't necessary for pumpout to happen), creating regions of higher and lower viscosities, then the system turns on and the gap changes shape due to thermal expansion, then, as a result, the TIM moves and some parts move more than others due to the different viscosities, then the system turns off and things change again. If the TIM was perfectly elastic, it'd return to its start point, but it's not, there's a viscoelastic behavior that basically prevents that from happening. So, what you end up having happen is things with low viscosity (like separated silicones) move more than others and gradually migrate out of the gap, causing a gradual decrease in thermal performance or a 'drying out' of the material. Movement can be from the center outwards, or from thinner gap regions to thicker gap regions. Sometimes to mitigate this people will make greases that harden in the gap (either by curing over time or vapor phase migration) to prevent pumpout, but it does create the risk of increased stresses on solder joints (which are less tolerant to stresses now that everything is lead free and keep getting smaller and smaller) or, the ultimate risk, loss of surface contact, which can result in device failure due to failure of the thermal solution.

 

[Maxx_Power]

Ahh okay, it is a structural, pore-space related packing process. I just wasn't aware that the pastes were under sufficient pressure to actually undergo this process, I assumed that the pastes became harder over time purely due to evaporation, leaving the filler materials behind.

No need for pressure. The packing and separation process can occur even inside a tube resting on a table. It will simply take a few years instead of a few weeks or months - colloidal suspension does not last forever.

As for evaporation, silicon is the base of many greases and does not evaporate. When you pull a HSF off a CPU that got pasted with a silicon-based thermal grease after a few months, you will usually see a ring of "wet" transparent stuff around the clay-like paste patch. That's the silicon oil that got pressed out of the thermal interface over time.

The process will happen regardless of pressure but more pressure does make it happen faster.

That is a coherent piece. I think this makes a lot of logical sense. I do observe that silicone rim (only on some compounds I used to use), but not recently, so I must have forgotten about it. Now that you mention it, the pressing action is actually enough to drive the particles out of suspension. Thanks for chipping in! Much appreciated.

 

[Maxx_Power]

No need for pressure. The packing and separation process can occur even inside a tube resting on a table. It will simply take a few years instead of a few weeks or months - colloidal suspension does not last forever.

As for evaporation, silicon is the base of many greases and does not evaporate. When you pull a HSF off a CPU that got pasted with a silicon-based thermal grease after a few months, you will usually see a ring of "wet" transparent stuff around the clay-like paste patch. That's the silicon oil that got pressed out of the thermal interface over time.

The process will happen regardless of pressure but more pressure does make it happen faster.

"pump out" ? - http://www.nordsonefd.com/images/comp300.jpg

Actually, its not correct to say they don't evaporate, depending on the type of silicone used there can be a substantial volatile portion. Vapor phase migration of silicones is a major concern for a lot of customers, particularly ones with optical connections and hard drive manufacturers. One of the ways to improve performance of a grease is to increase filler loading, but that increases viscosity, which usually prompts using a lower molecular weight silicone with a lower base viscosity to make the grease usable. I've seen greases that used silicones that were 8% volatile by mass, any silicones below D20 or L20 can, and will, undergo vapor phase migration. Silicones can be stripped of these lower molecular weights using distillation methods, but that makes them much more expensive and they're not available to retail consumers.

Liquid phase migration takes many forms, settling/separation and bleed are two dominate forms. Bleed is typically what you'll see when you dispense some grease on a surface and you can come back a day later and see a hazy halo surrounding the bead. Separation and settling can occur based more on filler loading, geometry and types of filler used. Settling happens a lot in greases because bondline is what dominates grease performance which means small fillers, which means low filling ratios (to hit a target usable viscosity) which means there will be settling. If you have a lot of fillers (like you do in gels and pastes) the fillers will interact with each other and slow or stop settling. Also, the lower the visocity of the resin, the more likely separation can occur.

Bleed, well, bleed happens with everything, pads, gels, greases, all of them bleed to various degrees and for various reasons. Bleed has been something I've been working for about 7 years now and could end up writing a book concerning it.

Pump-out is a kind of phenomena that occurs commonly with thin bondline applications and can be affected by liquid and vapor phase migrations. Generally, what happens is the TIM starts to separate (but separation isn't necessary for pumpout to happen), creating regions of higher and lower viscosities, then the system turns on and the gap changes shape due to thermal expansion, then, as a result, the TIM moves and some parts move more than others due to the different viscosities, then the system turns off and things change again. If the TIM was perfectly elastic, it'd return to its start point, but it's not, there's a viscoelastic behavior that basically prevents that from happening. So, what you end up having happen is things with low viscosity (like separated silicones) move more than others and gradually migrate out of the gap, causing a gradual decrease in thermal performance or a 'drying out' of the material. Movement can be from the center outwards, or from thinner gap regions to thicker gap regions. Sometimes to mitigate this people will make greases that harden in the gap (either by curing over time or vapor phase migration) to prevent pumpout, but it does create the risk of increased stresses on solder joints (which are less tolerant to stresses now that everything is lead free and keep getting smaller and smaller) or, the ultimate risk, loss of surface contact, which can result in device failure due to failure of the thermal solution.

I see what you are saying. I know that most of the data sheets I have come across states some amount of percentage based evaporation over an amount of time.

As for the Newtonian aspect of the thermal grease, I had no idea the contact zone depth is approaching within an order of magnitude of the average size of the particles suspended in the mixture. I would have guessed the fluid becomes highly non-Newtonian if I had suspected this; it is certainly not an aspect for someone not in the field to be able to easily accomplish by measurement. Although I suppose one could calculate, by measuring average applied drop sizes to estimate volume of paste used, then calculate the final surface area after press-fitting the surfaces, then estimate the thickness by dividing the volume by an average diameter of spread... I now have a better idea why the filler size and quality is an important aspect.

Long story short, thank you very much for chipping in, given that you seem to be in this field.

 

[InvalidError]

Although I suppose one could calculate, by measuring average applied drop sizes to estimate volume of paste used, then calculate the final surface area after press-fitting the surfaces, then estimate the thickness by dividing the volume by an average diameter of spread...

Or simply use an automatic measurement microscope: spread a test sample on a glass plate, put the plate on the microscope and let the microscope count particles along with statistics about their size.

 

[wdmfiber]

LOL, newegg is out of Gelid GC. Mass removal of heat sinks to wipe off the poor performing AS5?
http://www.newegg.com/Product/Product.aspx?Item=N82E16835426020

AS5 on sale... save 25%
http://www.newegg.com/Product/Product.aspx?Item=N82E16835100007

 

[patlaw]

Where did you find Noctua NT-H1 for $6?

 

[Maxx_Power]

Although I suppose one could calculate, by measuring average applied drop sizes to estimate volume of paste used, then calculate the final surface area after press-fitting the surfaces, then estimate the thickness by dividing the volume by an average diameter of spread...

Or simply use an automatic measurement microscope: spread a test sample on a glass plate, put the plate on the microscope and let the microscope count particles along with statistics about their size.

That's a great idea. I actually do have access to some measuring microscopes, I think I'll try that on some samples and take some pictures!

Thanks again!

 

[wdmfiber]

Where did you find Noctua NT-H1 for $6?

It's $6 on newegg.ca, 4 bucks more on the USA newegg.com
Prices at stores change weekly, sometimes daily. If the $4 bucks is killing you; I can help (PM me, I'll mail you some singles).

 

[Christopher Shaffer]

Why would you use thermal *adhesive* in a situation where the heatsink has something else (like a bolted mount) to keep it in place?

It makes no sense to test this on a GPU with a cooler held down by screws.

It would make a lot more sense to test it on RAM or other ICs where you can't physically retain the cooler, mostly because this is the type of situation a real user is going to apply this type of solution.

Even Arctic Silver claims the performance of this adhesive vs. standard AS5 is the same.

 

[Christopher Shaffer]

Also... I'm surprised you didn't try mayonnaise. Hardware Secrets' paste round-up landed mayo right in the middle on cooling ability. Margarine scored slightly worse than your toothpaste test.
I'm still waiting for the results of peanut butter vs. hemorrhoid cream.

 

[youcanDUit]

what are mounting pressures? i didn't understand that part too well.

 

[InvalidError]

what are mounting pressures? i didn't understand that part too well.

It is the amount of force the HSF combined with its retention mechanism put on the CPU's IHS.

 

[tsaavik]

Please do dry test (no thermal paste). I think it would be interesting to see on the graphs.

 

[InvalidError]

Please do dry test (no thermal paste). I think it would be interesting to see on the graphs.

Getting numbers from a dry-run might be tougher than it sounds since the CPU might hit thermtrip during boot and shutdown.

 

[youcanDUit]

what are mounting pressures? i didn't understand that part too well.

It is the amount of force the HSF combined with its retention mechanism put on the CPU's IHS.

ooohhh. thank you for being nice and answering as well.

 

[shulkman]

I'd be really interested to know how well the paste included with Corsairs H100i cooler performs in comparison to these. I went ahead and used it since it was already applied to the cooler. Everything seems to work well, but my system is so well cooled that the temps inside the case only run a couple degrees above ambient. I was actually able to set my heatsink fans to their lowest speed (500rpm, almost silent) and get barely a 1 degree F rise in temp. on a 3750K at stock speeds. I've never bothered OCing, since my computer rarely works hard enough to get to max speed on anything I do.

Per Corsair, they say what they have works best compared to anything else they have tested and wont change it.

Its hard to say for sure.

Ryan, thanks for the input. I liked IC Diamond but due to the insane amount of pressure I moved on to Zalmans current stuff as it works way better than their older STG1 and so far has kept my CPU happy. I also have a 9900Max and Zalman is well known for their mirror like finish so I doubt anything else would make a massive difference since the heatsink is already polished beyond what the majority of companies do.

Yeah, I have no complaints about the current Corsair compound, whatever it may be. I'm curious about the amount/thickness they provide with the package, since it's already applied to the heatsink. Maybe it should be less for Intel processors since they're smaller than AMD... I dunno. I haven't attempted to remove it since installation since I haven't had a need. But, after reading this article, it sets my inquiring mind off onto plenty of directions.

I would assume what they put on is sufficient for both since they do tons of internal testing. There is no way they can do one cooler for Intel only, the other for AMD etc.

I wouldn't mess with it if it keeps your temps where they need to be.

Now I just need Corsair to release a H110i as I want to throw it up top in my 500R but want the better air pressure fans they have.

And I thought Arctic silver 5 was the best after seeing all those reviews on newegg!

AS5 was the best. A long time ago. Things change and whats the best now may change later down the road.

Disappointed in Arctic Silver 5 after having been told time and again, not that long ago, it's the best stuff out there. Also needs a long burn-in time from what I recall.

It can take up to a year to cure. Which is pretty long TBH.

I've considered swapping out the fans, but I don't think I'll go for Corsair fans... I was leaning more towards the BeQuiet! fans or Noctua. I'm a little turned off by Corsair. The majority of parts in my build are Corsair components (Case, RAM, Cooler, a couple others) and I believe that they should have included their better fans with their products to begin with, particularly when it comes to the case (600T White). It just seemed a little cruddy to sell items with basic components in order to get the customer to shell out more for quality.

 

[Christopher Shaffer]

I'd still like a response to my question asking why you tested thermal ADHESIVE on a cooler that has a secure fastening method.

I'd really like to see it in real-world applications where it would actually be used, like sticking a heat sink somewhere where you CAN'T fasten it in place.

 

[FormatC]

It is simple: comparison between adhesive and thermal grease. Normally I use adhesive only for VRM or RAM heatsinks. But it was interesting to know, how performs this adhesive

You also can mix the Arctic adhesive with ceramic grease to get a compromise.

 

[Raghar]

Where are our favorites. IC Diamonds 7 - scratches your CPU clean. And Indigo Xtreme - would it be compatible with my heatsink and would a hairdryer suffice to force it to melt properly?

 

[sslayer777]

Great article, for those looking for benchmarks that might not be on here, Maximumpc also did a review on 17 pastes http://www.maximumpc.com/article/features/geek_tested_17_thermal_pastes_face?page=0,1

 

[Es0terin]

Another vote for Arctic Silver Ceramique II, Disappointed. Any specific reason you left it out?

 

[ITFT]

Can you also compare the Russian made one? It's dirt cheap and pretty decent, so I heard -) Here is the descriptionhttp://www.ulmart.ru/goods/158600#http://ru.wikipedia.org/wiki/%D0%9A%D0%9F%D0%A2-8Also could you add some cheap Chinese ones off the ebay, for the sake of comparison?