What is your / the best favorite thermal paste?

[BonzaiDuck]

One thing I have learned and come to respect about Bonzai Duck is that he does his research and the Duckman know what he is talking about.

My own take on thermal pastes all kinds I might add was that when I first started computing 15 + years ago the manufacturers were using a low grade thermal paste and if you wanted drastically better temps you had to purchase a high quality thermal paste!

Fast forward to 2014..that just is not the case anymore. Sure you can order a "high " quality thermal paste but why? To see a difference between 2-4 or even 2-6 degrees.....

Unless you are a power user and into extreme overclocking...I don`t mean a mild over clock, I mean extreme then IMO your wasting your money and your time.

Mind you I still have quite a large collection of CPU coolers!!

Peace!!

So far, I took off my sock and await a possible swim into water-cooling -- maybe even on the "exotic" side. But it brings complexities, and adds one dimension of maintenance. Fact is, I dread "first time" at anything, because you don't want to screw the pooch the first time.

There will be a release of a new Noctua cooler -- possibly dubbed the "D15" -- sometime in April or thereafter. IT will have the same number of heatpipes as the NH-U14S, but twice the fins, giving illusion of a "double U14S." The base will not be just solid copper, but a composite of copper and diamond. The heat-conductivity of the base alone will increase by 25% over previous models. This may mean the cooler will at least equal CLC/AiO water-coolers like the Nepton 280L or Corsair H100i in measuring temperature reduction at controlled ambient and test-bed thermal wattage. It may even surpass these current CLC models.

So the curse of air-cooling is the incremental nature of the improvements: lap the heatsink base and IHS (but water aficionados can also do it) -- maybe 5C or more; get best or second-best thermal paste -- worth somewhere between 2C and 5C depending on what you started with; choose the fans and "fan integration" to draw the maximum CFM through the cooler fins without adding to a prevailing noise-level. And the really "extreme" air-cooling lunatic might put together foam art-board ducts that make the heat transfer to moving air even more efficient.

The air-cooler requires less investment in time for installation over water, but these extra incremental improvements take more time, or cost a few bucks more like an $8 tube of IC Diamond for some limited number of applications. You either get three or four applications from the small tube, or purchase the large 24K tube for something like $20. But once it's been applied, it can even be "rejuvenated" and re-used, barring a need to remove it from the cooler -- and you might even try that with all its likely tedium. A "rejuvenation" option would suggest keeping tubes of "other" silicon or oil-based TIMs to "freshen" the nano-diamond sludge.

There is also maintenance with the air-cooler. You need to manage the build-up of kruft in the cooler fins -- on a regular basis.

Per my contact with IC Diamond's tech-rep and the "bare-die myth:" Given the specialized nature of the product, they want to expand their customer base. To maintain that customer base, they would attempt to avoid spawning unforeseen costs to the customer. They also want to expand the customer base by exploiting all the application opportunities a customer might have. So they're either going to tell you that the paste will damage a bare-die if such is the case, or they're going to deny it if the damage argument is a myth.

Post-Script: I started overclocking my system when I built it two years ago, and found the settings that had the highest core temperature reading at 84C. I good chunk of improvement came just for replacing the two Noctua fans with a single Akasa Viper. But getting the nickel-plate off my IHS was probably worth between 2C and 5C, and doing the same for the nickel-plate on my NH_D14 might double that. For TIM, I could go for metal pads -- another item or cost, or Coollaboratory "liquid" metal, or the diamond paste. The first one requires you to blast your heatpipe cooler or processor with a hair-dryer to melt the pad and make it set and bond to the copper. The second one can have you chasing around the motherboard for mercury-like metal balls for some types of liquid metal. You lose a degree or two with the diamond, but then you can re-use the sludge.

Now, at 4.6 and 4.7 Ghz settings with loaded VCORE at between 1.32 and 1.35V, my maximum core average is somewhere between 68 and 72, depending on the OC speed. I think I'm good for another 2C with a new fan I purchased, but I have to prove it.

That leaves some headroom to shoot for 4.8Ghz, but the real point of it: you are voltage-limited by water-cooling and heat-limited by air-cooling. As far as I'm concerned, I'm already at my chosen "voltage limit." Any decrease in voltage from further reducing temperature is likely to be very small.

 

[Tweakin]

...So far, I took off my sock and await a possible swim into water-cooling -- maybe even on the "exotic" side. But it brings complexities, and adds one dimension of maintenance...

Just so you know, I don't run any "exotic" dyes or blends in my rig, just good ol distilled water and biocide and I have not had to perform any maintenance other then chasing the dust bunnies away in the past 3 years on any of my rigs.

My water buffalo actually was self contained for 5 years before I took her apart, and upon inspection I could find no reason why it wouldn't have gone for another 5 years...

I actually view water as a low maintenance item that brings silence to the experience...just in case you want to take your other sock off!

 

[BonzaiDuck]

Just so you know, I don't run any "exotic" dyes or blends in my rig, just good ol distilled water and biocide and I have not had to perform any maintenance other then chasing the dust bunnies away in the past 3 years on any of my rigs.

My water buffalo actually was self contained for 5 years before I took her apart, and upon inspection I could find no reason why it wouldn't have gone for another 5 years...

I actually view water as a low maintenance item that brings silence to the experience...just in case you want to take your other sock off!

That wasn't what I meant by exotic. I've been thinking about building a bong cooler to integrate into a simpler initial setup. That brings more maintenance problems, though. The water must be filtered; the reservoir periodically replenished.

I understood that a regular six-month maintenance regimen was recommended for water-cooling. Maybe it involved draining and bleeding the system -- I can't remember.

There will be a point where air-coolers can't be further improved, although thermal requirements may continue to drop. I think water-cooling kits are keeping up with the improvements in air-coolers to maintain their advantage.

On the overclocking issue, reduction of temperature and heat means less signal noise to make an OC setting unstable. How this would matter as much in a 45C to 60C range, I can't say, except that it would mostly matter for fairly extreme clock settings. But you will still raise your voltage at those lower temperatures to achieve a higher clock.

I'm only guessing, but I thought [among statements and examples around the forums] I had seen people touting 45C load temperatures or thereabouts for good water-cooling. So if another advance in heatpipe cooling can reduce my own load temperatures another 10C, that puts me at probably 60C for my existing clock settings. But at this point -- on this SB-K system -- I'm probably not going to update to a water-cooling kit for an extra 100 or 200Mhz. If I did, I'd still need to push voltage to 1.4V or a little more. Next year, if I push forward with a plan to build a Haswell-E system, then I'm going to spend the money on the WC parts.

But either way -- you're going to benefit from the same incremental tweaks you might make with air-cooling: better TIM, thermally controlled PWM fans-on-a-string. If your margin is good enough -- say a load value of 50C tops -- you may not want to bother, but looking at the cost of IC Diamond for instance, why wouldn't you spend $8 after shelling out as much as $200 for good water-cooling?

The reason to do it after spending less than a Franklin on an air cooler: you'd want to do everything possible to close the gap between air and water even if the latter is impossible to match. If you aren't an "enthusiast," then you wouldn't have a motive to spend the $8. But the difference between cheaper TIM and the diamond is maybe enough to buy two tacos or get the 5-for-$5 Wienerschnitzel chili-dogs with onions.

My homemade tacos and bombero sauce may last about 15 minutes. Diamond . . . . lasts forever . . . :biggrin:

 

[Aikouka]

I use Prolimatech PK-3 and Noctua NT-H1. They probably don't provide nearly as much thermal conductance compared to their cost, but if you consider the cost per application, it's only a dollar or two for somewhat better thermals.

 

[It's Not Lupus]

arctic cooling mx2 and mx4

 

[BonzaiDuck]

I use Prolimatech PK-3 and Noctua NT-H1. They probably don't provide nearly as much thermal conductance compared to their cost, but if you consider the cost per application, it's only a dollar or two for somewhat better thermals.

Well, the diamond paste doesn't pump out like other thermal greases. And as I said, you can use the old paste again and again if you can salvage it from the copper surfaces -- just add a drop or two of the Prolima .. Noctua. Just seems to me that splitting hairs over the $8 price-tag doesn't make a lot of sense: it's an "investment" that doesn't need further attention, so one application should last the life of the computer.

The biggest gripe centers on the thickness -- making for tedium in application.

People are going to select what they prefer, what costs less, what applies easier. You just can't beat the nano-diamond or the liquid metal for their thermal conductivity, though. And -- sure -- $8 for maybe four applications. I think I used a tube of it for two applications to my CPU (could've saved the sludge from the first application and re-used it), and one application to my GPU's bare die and heatpipe cooler. And I remember the GPU die was fairly large.

But this isn't much different than people touting their fan preferences. They can't all be absolutely great products, and reviews depend on test-beds and measurement approaches. Ultimately, good data is "good data," and would eventually bear out as meaningful across product comparisons.

 

[Herkulese]

All I can say -- and I probably have the bar charts to prove it -- I did tests with specific overclock settings for a Q6600 rig in '07 using AS5 first, and then the IC Diamond. The latter definitely made the difference of up to 5C lower temperatures. I'm sure there are comparison reviews extant over the last several years with as many as 20+ thermal compounds. Some of those would include the diamond in the mix, others wouldn't. But I've known that you can simply apply transitivity logic to deductions from two different reviews of overlapping comparisons: If A>B, and B>C, then A>C and so on.

It all depends on whether the few C degrees in lower temperatures is important to you. It was to me, but I could get up to another 2C improvement over the diamond with the liquid metal formulations. After I saw pictures of how the liquid metal behaves like mercury -- how it can drop off the IHS during application and roll around on the motherboard in little silver balls -- I decided it wasn't worth it.

Meanwhile, on the "direct-die" application front. I have drafted a letter to Innovation Cooling to get their direct take on the issue. I had used the diamond paste on my GPU die several years ago, with no apparent problem developing. So I want to hear what they think about this.

I looked on Amazon, and they have two versions of the IC Diamond paste.
24 Carat of micronized diamond per tube - $15 a tube
7 Carat of micronized diamond per tube - $7 a tube.

I assume the tubes to be the same size, but cant say for sure.
Do you know if there is a marked difference based on the possible difference in the consentration of the micronized diamond dust?

 

[BonzaiDuck]

I looked on Amazon, and they have two versions of the IC Diamond paste.
24 Carat of diamond past per tube at $15 a tube
7 Carat of diamond past per tuve at $7 a tube.

I assume the tubes to be the same size, but cant say for sure.
Do you know if there is a marked difference?

No, no. . . . the difference in "carats" is in the amount of diamond and the amount of paste you get. The micronized (synthetic) diamond particles are the same for both, the loading specification is identical -- you just get more. I'm thinking the $15 tube is roughly more than three times the amount. Either way, you'll pay ~$5 or so in shipping charges -- the burden we bear for buying specialized and exotic products that can't be had running down to Best Buy [but on that, figure $4.50/gallon, heh-heh. Guess you're going to pay the oil and gas men either way . . . ]

 

[Herkulese]

No, no. . . . the difference in "carats" is in the amount of diamond and the amount of paste you get. The micronized (synthetic) diamond particles are the same for both, the loading specification is identical -- you just get more. I'm thinking the $15 tube is roughly more than three times the amount. Either way, you'll pay ~$5 or so in shipping charges -- the burden we bear for buying specialized and exotic products that can't be had running down to Best Buy [but on that, figure $4.50/gallon, heh-heh. Guess you're going to pay the oil and gas men either way . . . ]

Thanks for the extreemly prompt clarification.

As far as the price goes, it is well worth it, given all the data that you have provided.
I will order the larger tube and have some left for future use. I have a good heat pipe cooler, and plenty of well placed air movement, so this stuff should just give me some good insurance.

Thanks again for providing all the information on this stuff.

 

[Tweakin]

One thing I have learned and come to respect about Bonzai Duck is that he does his research and the Duckman know what he is talking about.

My own take on thermal pastes all kinds I might add was that when I first started computing 15 + years ago the manufacturers were using a low grade thermal paste and if you wanted drastically better temps you had to purchase a high quality thermal paste!

Fast forward to 2014..that just is not the case anymore. Sure you can order a "high " quality thermal paste but why? To see a difference between 2-4 or even 2-6 degrees.....

Unless you are a power user and into extreme overclocking...I don`t mean a mild over clock, I mean extreme then IMO your wasting your money and your time.

Mind you I still have quite a large collection of CPU coolers!!

Peace!!

+1

 

[BonzaiDuck]

Thanks for the extreemly prompt clarification.

As far as the price goes, it is well worth it, given all the data that you have provided.
I will order the larger tube and have some left for future use. I have a good heat pipe cooler, and plenty of well placed air movement, so this stuff should just give me some good insurance.

Thanks again for providing all the information on this stuff.

Tweakin and Jedi have submitted their cost-benefit-analysis, but for the chump change it could give you anywhere from 2 to 4C improvement in temperatures over Arctic Silver 5. I measured it -- figure for thermal wattage exceeding 110W. It could be more -- likely not less -- I don't remember the precise figure.

If you don't mind technically voiding your warranty on the processor (with little or no risk of damage), you can lap the IHS surface to eliminate the nickel-plate and make it flat. Could be worth another 5C, depending on thermal wattage. If there's nickel plate on your heatsink base, you can lap that off and maybe double that.

And like I said -- you can re-use diamond paste just by adding a drop or two of any other thermal grease.

 

[Herkulese]

Tweakin and Jedi have submitted their cost-benefit-analysis, but for the chump change it could give you anywhere from 2 to 4C improvement in temperatures over Arctic Silver 5. I measured it -- figure for thermal wattage exceeding 110W. It could be more -- likely not less -- I don't remember the precise figure.

If you don't mind technically voiding your warranty on the processor (with little or no risk of damage), you can lap the IHS surface to eliminate the nickel-plate and make it flat. Could be worth another 5C, depending on thermal wattage. If there's nickel plate on your heatsink base, you can lap that off and maybe double that.

And like I said -- you can re-use diamond paste just by adding a drop or two of any other thermal grease.

I ordered the larger tube of Diamond Paste. It was $14 and change with free regular shipping. As far as I am concerned, the price is in the noise. No expense at all.
Thanks again for the info.

Roger

 

[BonzaiDuck]

I ordered the larger tube of Diamond Paste. It was $14 and change with free regular shipping. As far as I am concerned, the price is in the noise. No expense at all.
Thanks again for the info.

Roger

I can also tell you that it's possible to mix your own, but watch out for the expense. Abrasive powders of the same micron-specification for industrial use can be had as either natural or synthetic diamond. Stuff is probably expensive either way -- I have five grams of powder and could've used the money to buy software, toward a graphics card or a dozen Mexican dinners.

There are these little round plastic cases for contact lenses, or something similar in size. You cut a hole in the top center so you can fit it with a paddle with crank-handle which you bend out of a large paperclip. then put the powder and sparing amount of a cheap silicone or grease-based TIM in there, and crank away. [And make sure not to breath any of that stuff if it becomes airborne, also. I think it can cause scarring of lung-tissue. But at least it's not your asbestos and mesothelioma . . ]

 

[Herkulese]

Mind you I still have quite a large collection of CPU coolers!!

Are you saying that you have several coolers for sale?

 

[BonzaiDuck]

JediYoda may have a bigger inventory of surplus or discard coolers than I have. I have perhaps one -- the OCZ Vendetta which might be good for someone, but I'd have to order the appropriate bracket-kit for whatever CPU socket-design. Or -- I could tell the buyer to order it. I also have a CM 212+ (including all sockets through 1155 plus the kit for 2011) which I intend to keep. But I doubt you would want a ThermalRight Ultima 90, or a ThermalRight XP-120 -- unless you thought it suited your HTPC build. The mounting kit for those only includes socket LGA_775, for which I'd made special order sometime way back when . . .

 

[MongGrel]

Noctua NT-H1 myself.

*shrug*

 

[BonzaiDuck]

Noctua NT-H1 myself.

*shrug*

All of those formulations pretty much do what they're supposed to do, but with varying degrees of effectiveness, and the benchmark tests pretty much bear it out.

Now that you bring it up, I find it ironic that the NT-H1 and many other formulations aren't made with micronized diamond, but now Noctua is planning to release a new cooler with a heatsink-base made of a copper-diamond composite, for which they boast a thermal conductivity greater by 25% than that for solid copper.

Here's a comparison review of the NT-H1 against AS5 and several others:

http://www.overclockersclub.com/reviews/noctua_nt_h1/5.htm

My choice -- the micronized diamond paste -- is of course missing from the comparison. But other reviews and tests, including my own tests for a testbed TDP of > 110W, shows a 2 to 4C temperature reduction over AS5. The liquid-metal formulations best the diamond paste by about 1 to 2C.

Some of my colleagues here find my obsession with incremental improvements to be amusing. I might think it amusing myself, but each grain of rice that you find adds up to a mouthful or even a whole bowl.

 

[Herkulese]

My choice -- the micronized diamond paste -- is of course missing from the comparison. But other reviews and tests, including my own tests for a testbed TDP of > 110W, shows a 2 to 4C temperature reduction over AS5. The liquid-metal formulations best the diamond paste by about 1 to 2C.

Some of my colleagues here find my obsession with incremental improvements to be amusing. I might think it amusing myself, but each grain of rice that you find adds up to a mouthful or even a whole bowl.

Very well said.
Just like, ever jurney begins with the first step.

BTW, my tube of IC Diamond past just arrived over the weekend. It took awhile to get here, but I am not yer ready to build, so no problem.