What thermal compound?

[emueyes]

It should be noted, too, that compounds are tailored to package type. Ceramique works really well on exposed silicon - chipsets, PLX bridge chips and so on, but not so well on the larger areas of heatspreaders. If mayonnaise is OK as a compound, the difference between top end (and expensive) compounds isn't likely to be much. Certainly application technique, including surface preparation, would be more important than the compound, and possibly more useful for inexperienced people to read than parochial bickering.

I must admit, finding out whether diamond dust will scratch a mirror is a curiousity though. I know that MX2 will very slightly abrade copper - it removes enough metal to leave a deposit on a plastic application sheet, but copper is so much softer than glass that's not much of a comparison.

[Captain Ned]

Vegemite.

Ask Dan of Dan's Data.

[Firestarter]

The best value thermal compound is the white gooey stuff that comes with whatever heatsink you buy.

[just brew it!]

I must admit, finding out whether diamond dust will scratch a mirror is a curiousity though.

Diamond is much harder than glass. It'll either scratch it or polish it, depending on the particle size.

[emueyes]

I must admit, finding out whether diamond dust will scratch a mirror is a curiousity though.

Diamond is much harder than glass. It'll either scratch it or polish it, depending on the particle size.

Yep. Getting it to do either would require quite a bit of force though, so in practical terms it may be difficult to get something to apply pressure with. I used to use automotive paint cutting compound as the last stage in lapping a heatsink, but that can be used with a cloth and is probably much larger grained than diamond hs compound.

A quick check from
http://innovationcooling.com/ICDDatasheet.htm and
http://www.arcticsilver.com/as5.htm

is quite odd. The claimed particle sizes are

Arctic Silver 5 < 0.49u
Arctic Alumina < 0.39u
IC Diamond 7 < 40u

that is, the diamond particles are an order of magnitude larger than a common silver and aluminium based compound, as claimed by the manufacturers.

My ideal for hs compound application is to try to get conductive particles into the remaining low points in a surface, and to remain there, so the smaller the particle the better. I think that would be an argument against diamond particle based compounds, or at least the sample of one that I chose.

[just brew it!]

Yep. Getting it to do either would require quite a bit of force though, so in practical terms it may be difficult to get something to apply pressure with.

If the heat spreader on the CPU and/or the base of the heatsink is uneven, it's conceivable that there could be a high spot where most of the pressure from the mounting clip is concentrated. I could imagine any shifting of the HSF after it has been mated to the CPU causing some abrasion in that situation.

[Arclight]

I haven't tested many TIMs, so far i've only used some in the box Thermalright TIM and Arctic MX 4, which ultimately i recommend since it's dielectric and it says on that package that it's good for 8 years, so i assume that performance shouldn't degrade much after time.

[emueyes]

So if the 40u figure is accurate (for diamond particles) would that be considered a scratch? Let's see. 40u is the grit size of P360 sandpaper; a copper block would be very badly scratched if sanded with P360 paper, and potential pressure points between spreader and hs would almost certainly produce visible artifacts.

MX2 would do this too, aluminium oxide is almost as hard as diamond, harder than glass and very much harder than aluminum and copper.The much finer particles, though, would leave correspondingly smaller artifacts, probably acting as buffing. The quoted 0.39u size is ten times smaller than P2500 sandpaper.

I'm just doing this from public domain data and logic; I may be way off, but much more importantly, it probably doesn't make much difference at all in cooling a chip.

[ikjadoon]

While an interesting discussion about IC Diamond, Skinee's 2011-2012 results shows Indigo Xtreme/Maingear EPIC T1000 as the best TIM ever created: http://www.legitreviews.com/article/1852/1/

LR shows a larger performance increase, but they aren't as standardized as SKinee. Cons include price ($20 for 2 applications), anxiety-inducing reflow procedure, and no heatsinks with exposed heatpipes/circular bases.

[just brew it!]

While an interesting discussion about IC Diamond, Skinee's 2011-2012 results shows Indigo Xtreme/Maingear EPIC T1000 as the best TIM ever created: http://www.legitreviews.com/article/1852/1/

LR shows a larger performance increase, but they aren't as standardized as SKinee. Cons include price ($20 for 2 applications), anxiety-inducing reflow procedure, and no heatsinks with exposed heatpipes/circular bases.

That's some crazy stuff. Given the price and the somewhat scary installation procedure, it is definitely not for everyone.

[mortifiedPenguin]

wow, so many recommendations for arctic silver...there are better pastes now for the money. penguin, how can you recommend the stuff when the very website you cited (which is very good), basically tested the stuff to be rather low performance in comparison to other (and sometimes much cheaper) pastes?

I never claimed that AS5 was going to be the best paste ever but rather that it was a consistently solid performer across many sites I've visited and that it was a safe choice to pick. The fact that Skinnee Labs didn't find it all that great on poor and moderate contact situations doesn't really change my opinion either and I wasn't necessarily using the site as "support" of that opinion anyway. I should also note that Skinnee Labs doesn't use a zero based graphing system, meaning that the true differences between pastes are far smaller than they would appear based on the graphs, which is only about 5 degrees Celsius. If you're honestly that worried about 5 C, buy a better heat sink.

[ikjadoon]

While an interesting discussion about IC Diamond, Skinee's 2011-2012 results shows Indigo Xtreme/Maingear EPIC T1000 as the best TIM ever created: http://www.legitreviews.com/article/1852/1/

LR shows a larger performance increase, but they aren't as standardized as SKinee. Cons include price ($20 for 2 applications), anxiety-inducing reflow procedure, and no heatsinks with exposed heatpipes/circular bases.

That's some crazy stuff. Given the price and the somewhat scary installation procedure, it is definitely not for everyone.

It is a little adventurous, but "crazy" sounds a little extreme.

Least Crazy Cooling Tools
AS5-level TIM
IC Diamond TIM
Indigo Xtreme
Water-cooling
LN2
Most Crazy Cooling Tools

Not for first-time builders or those on a tight budget, but it has some pretty tangible benefits for most others (...if you look at the LR review...7C drop in OC load compared to AC's MX-4).

[Nec_V20]

I use the AC MX-2 and for cleaning off the gunk left from any prior application I use 99% Isopropanol (you can get it dirt cheap on EBAY). It removes the grunge in no time at all.

[Chrispy_]

If you can afford to buy fancy thermal compound, you can afford a better heatsink or waterblock which will make far more difference than any properly-applied layer of goop.

Thermal compound marketing is great for suckers, but it's like buying a different exhaust pipe for your 120HP car. You might get the horsepower up to 125HP, but if you want more power just get a 400HP engine in there instead.

[GrimDanfango]

The only thermal compound I've ever found to make an actual measurable difference is Coollaboratory Liquid Pro, which I've had knock a full 10 degrees off, and even more on de-lidded CPUs.
I haven't used it in years though, as that stuff is naaasty if you ever have the slightest inclination to remove the heatsink again at some point. It's practically impossible to remove completely once applied.

As far as conventional compounds go, everyone has a favourite, and it's usually the one they've tried, and did the job perfectly well.

The two things that actually make a real difference are:
- learning to apply the paste effectively - you need to apply the thinnest possible layer that still makes even contact across the whole chip/heatspreader. A thicker layer will raise temperatures considerably, as the thermal compound will always have worse thermal conductivity than the two pieces of metal you're putting it between.
- de-lidding the CPU, at least ivy/haswell CPUs, where they've stopped using solder as a thermal interface. One weak link is way better than two weak links - the "heatspreader" on modern CPUs is there for one reason and one reason only - to protect the silicon underneath from reckless heatsink installation, and thus from needless warranty claims. It doesn't serve to "spread" the "heat" any more effectively than a dirty-great heatsink mounted directly to the chip.

If you're really obsessed with cooling efficiency, these two points will make way more difference than which brand of compound you use, and if you're not obsessed, then none of it really matters, as all but the most powerful server processors are easy to cool to a perfectly adequate level, with minimal noise, just by picking a decent cooler.

[Nec_V20]

The only thermal compound I've ever found to make an actual measurable difference is Coollaboratory Liquid Pro, which I've had knock a full 10 degrees off, and even more on de-lidded CPUs.

One thing you forgot to mention is that if Liquid Pro comes in contact with any aluminium it will cause severe damage. So anything it touches must be completely aluminium free.

Gallium will turn aluminium into something with the consistency and strength of cardboard.

[Airmantharp]

The only thermal compound I've ever found to make an actual measurable difference is Coollaboratory Liquid Pro, which I've had knock a full 10 degrees off, and even more on de-lidded CPUs.

One thing you forgot to mention is that if Liquid Pro comes in contact with any aluminium it will cause severe damage. So anything it touches must be completely aluminium free.

Gallium will turn aluminium into something with the consistency and strength of cardboard.

That's a good point- many HSFs run their heatpipes through a base which is usually Al. I always preferred the 'direct' ones that just squished the pipes instead, and the pipes are usually copper.

[just brew it!]

That's a good point- many HSFs run their heatpipes through a base which is usually Al. I always preferred the 'direct' ones that just squished the pipes instead, and the pipes are usually copper.

Even the "direct contact" ones still tend to mount the pipes in an aluminum base though.

[willyolio]

if you've got steady hands and NO ALUMINUM on your heatsink: liquid metal.