HardOCP Thermal Paste Test

[akugami]

HardOCP Thermal Paste Test

I'm surprised at how well Arctic Silver 5 and Arctic Silver Ceramique holds up even after all these years. I personally use AS Ceramique because I've been using it for a long while and there's zero chance of it shorting out anything. A large tube of Ceramique is also cheap and lasts pretty much forever.

 

[Tweakin]

Yep...I've been using Ceramique for years now...it's a little "thick" to put on, but it has great qualities...

 

[Rubycon]

Their results are dead on with Shin-Etsu. That stuff gives the best load temps AND the closest between core temps out of all the greases.

 

[cusideabelincoln]

Yeh, but we're talking about a difference of 1 C among almost all of the pastes. So any of these would probably be much better than the standard stuff on stock heatsinks, although I wish they would have thrown one sample of it in the test.

 

[BladeVenom]

After looking at the results, I guess I'll have to stop using cheese as a thermal interface material.

 

[akugami]

Originally posted by: cusideabelincoln
Yeh, but we're talking about a difference of 1 C among almost all of the pastes. So any of these would probably be much better than the standard stuff on stock heatsinks, although I wish they would have thrown one sample of it in the test.

I agree, they should have used some stock paste as test. Although you at least can see a more current testing of how different thermal pastes perform relative to one another. Considering the costs vs performance benefits, I'll just stick to Ceramique.

Originally posted by: BladeVenom
After looking at the results, I guess I'll have to stop using cheese as a thermal interface material.

http://www.dansdata.com/goop.htm

You can switch to toothpaste instead. It performs better than cheese.

 

[eelw]

Why no IC Diamond?

 

[Zap]

I really wish they had included some really "cheap" stuff. Well, actually they sorta did with that 2oz tub, but I mean more of the blister pack included with cheap heatsinks.

Seeing the close spread and the differences between different application methods should tell us something... that it doesn't matter what paste you use as long as you use some and that IT IS APPLIED PROPERLY.

It is curious that out of the three Shin Etsu pastes, one that is supposed to be good got nearly last and the one that is supposed to be not as good got second. I'm wondering if they got the tubes mixed up, or if Shin Etsu got them mixed up. The reason is that Shin Etsu states the thermal properties and if you reverse those two results, you have a match with Shin Etsu's published numbers.

[shilling] Speaking of Shin Etsu paste, BFG uses it on higher end cards. Mostly pre-applied at the factory on the fansink, but the water cooled cards get a nice thin layer of it here at our HQ. Yes, a thin layer as opposed to a blob in the middle, just like how HardOCP showed this thicker stuff to perform better.

So, for all you noobs out there who think you are being uber by re-doing your water-cooled BFG card with Arctic Silver 5... you just pwned yourself.

Also of note is that BFG uses Ceramique by Arctic Silver for the RAM.

See this review that got 39ºC core load temperatures on a GTX 285 using Shin Etsu. [/shilling]

 

[TidusZ]

Originally posted by: Zap
I really wish they had included some really "cheap" stuff. Well, actually they sorta did with that 2oz tub, but I mean more of the blister pack included with cheap heatsinks.

Seeing the close spread and the differences between different application methods should tell us something... that it doesn't matter what paste you use as long as you use some and that IT IS APPLIED PROPERLY.

It is curious that out of the three Shin Etsu pastes, one that is supposed to be good got nearly last and the one that is supposed to be not as good got second. I'm wondering if they got the tubes mixed up, or if Shin Etsu got them mixed up. The reason is that Shin Etsu states the thermal properties and if you reverse those two results, you have a match with Shin Etsu's published numbers.

[shilling] Speaking of Shin Etsu paste, BFG uses it on higher end cards. Mostly pre-applied at the factory on the fansink, but the water cooled cards get a nice thin layer of it here at our HQ. Yes, a thin layer as opposed to a blob in the middle, just like how HardOCP showed this thicker stuff to perform better.

So, for all you noobs out there who think you are being uber by re-doing your water-cooled BFG card with Arctic Silver 5... you just pwned yourself.

Also of note is that BFG uses Ceramique by Arctic Silver for the RAM.

See this review that got 39ºC core load temperatures on a GTX 285 using Shin Etsu. [/shilling]

I replaced the TIM on my BFG Gtx 295 with OCZ freeze and load temps dropped ~10c. The TIM on my gpu's was caked on really thick, certainly not the thin layer on those water cooled cards.

 

[MrPickins]

Speaking of application, CM utilizes the most ingenious method for applying its paste. It includes a thin reusable sticker with a square cut out of its center. By placing the sicker over the base of your heat sink you can apply a bit of paste and spread it by using the included plastic card. In this way you get a perfect square of thin paste spread across the base of your heat sink. Think of how some heat sinks come with paste already applied and you get the basic idea. Very simple, very effective.

Genius.

I'm gonna have to try improvising something to do that. It would make ceramique much easier to apply evenly.

 

[Stoneburner]

Originally posted by: BladeVenom
After looking at the results, I guess I'll have to stop using cheese as a thermal interface material.

Are you attacking cheese?

 

[DrMrLordX]

Originally posted by: eelw
Why no IC Diamond?

Good question. The stuff's pretty cheap too, and readily-available.

Originally posted by: Zap

It is curious that out of the three Shin Etsu pastes, one that is supposed to be good got nearly last and the one that is supposed to be not as good got second. I'm wondering if they got the tubes mixed up, or if Shin Etsu got them mixed up. The reason is that Shin Etsu states the thermal properties and if you reverse those two results, you have a match with Shin Etsu's published numbers.

I find their results to be a bit weird as well. G751 has been reputed to be the best of their mainstream pastes for rough, unpolished surfaces, but the 7783D was supposed to be the best for lapped surfaces. 7762 is supposed to be an older product that is inferior to 7783D, or so had been lead to believe anyway.

Without knowing anything about the state of their Ultra-120 or the IHS on the QX9650 used in their test, it's impossible to know how rough either surface was.

 

[BonzaiDuck]

Originally posted by: eelw
Why no IC Diamond?

Can't help but chime in here, as I have . . . again . . . and again. . . . and again . . . on this issue. But especially, the quantitative, graphed results are revealing for what I want to say.

This year, I'll start drawing Social Security, so I'm old enough to have "senior memory events." But I recall several of these comparison reviews over the last two or three years.

IC Diamond's bench-test results were first touted by Joe Citarella at OverClockers.com, in January (17th), 2007. The product didn't appear on the market until around April of that year. Citarella had ended his review with a disclaimer, saying that he was a principal in a firm pursuing "innovative cooling" [ :laugh: ] solutions. This was responsible disclosure, and his test results utilized a $25,000 calorimeter, showing considerable performance gain over other pastes -- including Arctic Silver 5.

Later, Innovative Cooling's web-site published another evaluation under a test-bench configuration for a processor with approximately 100W TDP or effective load thermal power. The test-results showed between a 3C and 4C improvement with IC Diamond under those conditions.

At the time IC's own evaluation (probably Joe's) appeared, I was running my own tests on a C2D E6600 over-clocked to between 3.2 and 3.3+ Ghz. I was using CoreTemp configured to sample the core sensors every 8 seconds and save the results to a file. My tests at that time loaded the processor with PRIME95 or Orthos -- both cores, with the small-FFTs test for a minimum of one hour -- some test runs for several hours. The results were then imported to an Excel spreadsheet. I computed averages, minimums, maximums and standard errors, developing bar-graph frequency counts across the integer-values covering the range between Max and Min.

Tests were run against the AS5 paste, my own concoction of silicon-grease mixed with micronized diamond powder, and IC Diamond.

My tests were completely consistent with the results shown by Joe at OverClockers and IC Diamond on their web-page.

So if I see the results in the OP's linked review pages, I can pretty much say that IC Diamond will match anything that tops AS5 -- S.E. X23, S.E. G751, MX-2 -- whatever it is that shows only a 1 to 2C improvement in load temperatures.

We can speculate as to why IC Diamond hardly ever appears in these reviews, such as those I've seen and the one cited here. It may not be much different than Maximum PC Magazine showing comparison reviews for coolers like the Zalman 9500, various Tt models, Arctic Cooling, etc. -- to exclude the ThermalRight TRUE, then post a 1-page color advertisement for their "Kick-ass 9" winner just opposite the final page of the comparison review. Maybe it's something else, but I stand by my results with IC Diamond in comparison to AS5.

The "transitivity axiom" carries the weight of my argument and conclusion. "IF A > B, and B>C, THEN . . . . " [you know . . . . ]

By the way -- Joe recently posted a benchtest evaluation of the Sunbeam Core-Contact Cooler. He ALWAYS posts his measurements in terms of Thermal Resistance. And some may remember here at Anandtech that I was skeptical about FrostyTech's review of the Sunbeam against other makes. Well, Joe shows it to have a TR= 0.13 C/W, and we long ago determined that the TRUE's thermal resistance was just above 0.09 C/W.

I guess I won't waste any change on the Sunbeam.

 

[lopri]

Originally posted by: Stoneburner

Originally posted by: BladeVenom
After looking at the results, I guess I'll have to stop using cheese as a thermal interface material.

Are you attacking cheese?

You can always substitute cheese with tooth paste. Works rather well.

 

[Rubycon]

Originally posted by: lopri
You can always substitute cheese with tooth paste. Works rather well.

Toothpaste on a hamburger?

 

[BonzaiDuck]

Originally posted by: lopri

Originally posted by: Stoneburner

Originally posted by: BladeVenom
After looking at the results, I guess I'll have to stop using cheese as a thermal interface material.

Are you attacking cheese?

You can always substitute cheese with tooth paste. Works rather well.

Except for the abrasives already found in toothpaste, you could mix micronized diamond powder with it.

On the other hand, I'm not eager to grind off the enamel of my teeth . . .

 

[BladeVenom]

Originally posted by: lopri

Originally posted by: Stoneburner

Originally posted by: BladeVenom
After looking at the results, I guess I'll have to stop using cheese as a thermal interface material.

Are you attacking cheese?

You can always substitute cheese with tooth paste. Works rather well.

It may work in the short term, but toothpaste will dry out. In an emergency, I would consider some household or automotive grease, like regular silicone grease.

 

[Idontcare]

Originally posted by: BonzaiDuck

Originally posted by: eelw
Why no IC Diamond?

Can't help but chime in here, as I have . . . again . . . and again. . . . and again . . . on this issue. But especially, the quantitative, graphed results are revealing for what I want to say.

This year, I'll start drawing Social Security, so I'm old enough to have "senior memory events." But I recall several of these comparison reviews over the last two or three years.

IC Diamond's bench-test results were first touted by Joe Citarella at OverClockers.com, in January (17th), 2007. The product didn't appear on the market until around April of that year. Citarella had ended his review with a disclaimer, saying that he was a principal in a firm pursuing "innovative cooling" [ :laugh: ] solutions. This was responsible disclosure, and his test results utilized a $25,000 calorimeter, showing considerable performance gain over other pastes -- including Arctic Silver 5.

Later, Innovative Cooling's web-site published another evaluation under a test-bench configuration for a processor with approximately 100W TDP or effective load thermal power. The test-results showed between a 3C and 4C improvement with IC Diamond under those conditions.

At the time IC's own evaluation (probably Joe's) appeared, I was running my own tests on a C2D E6600 over-clocked to between 3.2 and 3.3+ Ghz. I was using CoreTemp configured to sample the core sensors every 8 seconds and save the results to a file. My tests at that time loaded the processor with PRIME95 or Orthos -- both cores, with the small-FFTs test for a minimum of one hour -- some test runs for several hours. The results were then imported to an Excel spreadsheet. I computed averages, minimums, maximums and standard errors, developing bar-graph frequency counts across the integer-values covering the range between Max and Min.

Tests were run against the AS5 paste, my own concoction of silicon-grease mixed with micronized diamond powder, and IC Diamond.

My tests were completely consistent with the results shown by Joe at OverClockers and IC Diamond on their web-page.

So if I see the results in the OP's linked review pages, I can pretty much say that IC Diamond will match anything that tops AS5 -- S.E. X23, S.E. G751, MX-2 -- whatever it is that shows only a 1 to 2C improvement in load temperatures.

We can speculate as to why IC Diamond hardly ever appears in these reviews, such as those I've seen and the one cited here. It may not be much different than Maximum PC Magazine showing comparison reviews for coolers like the Zalman 9500, various Tt models, Arctic Cooling, etc. -- to exclude the ThermalRight TRUE, then post a 1-page color advertisement for their "Kick-ass 9" winner just opposite the final page of the comparison review. Maybe it's something else, but I stand by my results with IC Diamond in comparison to AS5.

The "transitivity axiom" carries the weight of my argument and conclusion. "IF A > B, and B>C, THEN . . . . " [you know . . . . ]

By the way -- Joe recently posted a benchtest evaluation of the Sunbeam Core-Contact Cooler. He ALWAYS posts his measurements in terms of Thermal Resistance. And some may remember here at Anandtech that I was skeptical about FrostyTech's review of the Sunbeam against other makes. Well, Joe shows it to have a TR= 0.13 C/W, and we long ago determined that the TRUE's thermal resistance was just above 0.09 C/W.

I guess I won't waste any change on the Sunbeam.

Bonzai, I had (until now) held the mistaken impression that IC Diamond was a gimmick more than anything else and assumed that its lack of presence in the majority of TIM reviews was proof of this perception.

But your testimony to the contrary is beyond proof to me that I was completely in error with my rash judgement. I'm not old enough to be drawing SS checks but I'm not so young as to be beyond acknowledging when I'm wrong and doing an about face.

So IC Diamond is where its at. Cool. Me loves when folks around here take the time to chime in and add clarity as you did above.

Time for me to procure a tube for some hobby time fun. Any reports of sub-zero temp problems with the IC paste? (for example ceramique is favored by the vaporphase and LN2 cooling guys because it doesn't change thermal properties too much as you go below 0C, unlike AS5 which becomes increasingly thermal resistant as the temps decline below its glass transition temperature)

 

[DrMrLordX]

I will probably try out some IC Diamond 7 myself on my next build, whenever that will be. I've been happy with Shin Etsu x23-7783D and I have some on-hand that I could test against the IC Diamond 7 sometime if I ever got the gumption to do it.

 

[theAnimal]

There is a large thread about IC Diamond 7 in Distributed Computing.

 

[Idontcare]

Originally posted by: theAnimal
There is a large thread about IC Diamond 7 in Distributed Computing.

Of course, now why didn't I think of that? It only makes sense that a massive thread discussing the impact of a TIM on CPU cooling and/or overclocking is stowed away in a DC forum thread...

I guess I should be looking there more often for the latest info on GPU's and SSD's as well.

 

[theAnimal]

Originally posted by: Idontcare

Originally posted by: theAnimal
There is a large thread about IC Diamond 7 in Distributed Computing.

Of course, now why didn't I think of that? It only makes sense that a massive thread discussing the impact of a TIM on CPU cooling and/or overclocking is stowed away in a DC forum thread...

You didn't think of it. I did, which is why I posted it. And it kind of does make sense, with people running OCed PCs at 100% load 24/7.

 

[BonzaiDuck]

Originally posted by: Idontcare


Bonzai, I had (until now) held the mistaken impression that IC Diamond was a gimmick more than anything else and assumed that its lack of presence in the majority of TIM reviews was proof of this perception.

But your testimony to the contrary is beyond proof to me that I was completely in error with my rash judgement. I'm not old enough to be drawing SS checks but I'm not so young as to be beyond acknowledging when I'm wrong and doing an about face.

So IC Diamond is where its at. Cool. Me loves when folks around here take the time to chime in and add clarity as you did above.

Time for me to procure a tube for some hobby time fun. Any reports of sub-zero temp problems with the IC paste? (for example ceramique is favored by the vaporphase and LN2 cooling guys because it doesn't change thermal properties too much as you go below 0C, unlike AS5 which becomes increasingly thermal resistant as the temps decline below its glass transition temperature)

Originally posted by: DrMrLordX
I will probably try out some IC Diamond 7 myself on my next build, whenever that will be. I've been happy with Shin Etsu x23-7783D and I have some on-hand that I could test against the IC Diamond 7 sometime if I ever got the gumption to do it.

I need to go back to that HardOCP article theAnimal links in his post. I've seen a lot of hype about diamond paste.

The facts are straightforward, though. Diamond has a thermal conductivity that (I think) is about three or four times higher than Silver -- either micronized powder from natural diamond (expensive -- let me tell you my story, or how I could be $100 richer today), or synthetic diamond (same material -- much cheaper). The thermal resistance (opposite side of same coin) is correspondingly lower than that of Silver.

Ranking the materials used in heatsinks according to thermal properties just discussed, the ascending value goes from Al to Cu to Au to Ag . . . to C (diamond).

There are elements that fit in between, and I can't be sure if gold is better or worse than copper. But I think it's better, or I wouldn't have seen an experiment with a gold coin and an aluminum heatsink performed early in this decade (and posted at OverClockers.com.) Nickel is adjacent to Copper on the Galvanic table, which makes it good for plating copper for avoidance of corrosion. But Nickel's thermal properties suck by comparison. Thus we take advantage of grinding off ThermalRight's nickel-plate and the nickel-plate on the CPU's IHS.

By the way -- forget about the coin tricks. I destroyed a $50 collector's edition silver dollar by turning it into a $15 silver-slug. By adding a second thermal interface between the IHS surface and the heatpipe-cooler base, you increase the overall thermal resistance of the configuration. That's why some people were grinding down their IHS to expose the chip surface underneath.

The difference between silver and diamond is vastly greater than the difference between copper and silver.

But back to the hype: I've seen people post boasts and claims of double-digit temperature reductions, and it's just not scientifically sound.

A $7 tube of IC Diamond gives you about four applications-worth of material. I use it on my GPUs as well as the CPU. Per application, that's probably a lot more expensive than AS5, but let's face it: for that price and for any given computer-building-project, it's cheap and probably worth it. You can also get a $20 tube of the stuff.

It doesn't degrade like AS5 and other thermal pastes. You might even re-use it, although you may have to moisten the old material with a small dab of silicon grease.

 

[pm]

I tend to use IC7 on my computer at home. Based on my own measurements, I knocked something like 3C (I have the number on my notebook at home) off of it vs. AS5 under load. I will say that whatever it's thermal abilities, my tube of IC7 is ridiculously over the top hard to apply. People say to heat the tube before applying, but that didn't help when I did it on my Core i7. The problem that I have is that it sticks to everything else better than it sticks to the CPU so whatever you spread it out with, you end up picking up a lot of it. One time I had it all nicely spread and then I touched it again and the whole thing came off as an huge flat sheet. It's so painfully annoying, and 3C is such a minor difference (I've never had a CPU fail on me at home ever), that I often wonder if it's worth the pain.

I hope there's nothing wrong in mentioning this, but at work in the labs at work we all use AS5.

 

[BonzaiDuck]

Originally posted by: pm
I tend to use IC7 on my computer at home. Based on my own measurements, I knocked something like 3C (I have the number on my notebook at home) off of it vs. AS5 under load. I will say that whatever it's thermal abilities, my tube of IC7 is ridiculously over the top hard to apply. People say to heat the tube before applying, but that didn't help when I did it on my Core i7. The problem that I have is that it sticks to everything else better than it sticks to the CPU so whatever you spread it out with, you end up picking up a lot of it. One time I had it all nicely spread and then I touched it again and the whole thing came off as an huge flat sheet. It's so painfully annoying, and 3C is such a minor difference (I've never had a CPU fail on me at home ever), that I often wonder if it's worth the pain.

I hope there's nothing wrong in mentioning this, but at work in the labs at work we all use AS5.

That may seem reasonable -- if there is a premium on time servicing a larger number of computers.

But really, I don't find it a lot of trouble with my method. I use an old razor-blade as a spreader, keeping a Swiss-army knife handy to scrape the razor to re-apply the surplus. I spread the stuff on both the heatsink-base and the IHS. Sometimes, it doesn't seem to adhere too well, but putting a tiny dab with the knife-tip on those spots where it hasn't adhered usually does the trick. IC says that the medium dries slightly, and I think they recommend waiting ten minutes before mating the heatsink and IHS.

If you apply too much of it, it just squeezes out the edges of the two mated surfaces, forming a bead -- like too much silicon-sealant around the edge of your bathtub. It's totally harmless; non-conductive and non-capacitative. So for those two properties, at least one-up on AS5.