Last time I bought some thermal paste I bought the following:

http://www.innovationcooling.com/ Innovation Cooling Diamond 7 carat

It dropped my temps by about 2*C compared with Artic Silver 5.

That was about 6 months ago, so that diamond thermal compound isnt anything new!

It was new to me... and thank you for the link. We've ignored the raging debates about TIM for a long time as the variances and margins of error in testing seemed to be at least as big as that claimed by various makes... but perhaps these diamond-based ones are really better. We'll try them out, then post a report in a while.

So, it's the next to best material for thermal conductivity? What's the best?

To be honest, I agree mostly, when it comes to TIM then provided you buy one of the better ones I don't think it really matters all that much.

I believe there probably is a worthwhile difference between "Generic TIM 1" and say "Artic Silver 5" or one of these diamond ones. But comparing AS5 with AS3 or ceramique or diamond or whatever, you can get enough variance just with slightly different application methods!

What has a bigger impact, and is widely known by the overclocking communities, is lapping your heatsink and CPU's IHS, to provide completely flat even surfaces to mate with TIM.

Uhm... it IS diamond. The second best is silver.

Diamond has 5 times better thermal conductivity than silver and 6 times more than copper. Just imagine a diamond heatsink

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

http://benchmarkreviews.com/index.php?o ... 8&Itemid=1

This is my favorite Thermal Paste article -- it is a 33 way comparison of all types of thermal paste. It has "Innovative Cooling 7 Carat Diamond" Thermal paste in it --- not sure if it actually has diamonds or not.

Actually, it's Graphene which is at a minimum over twice as conductive as diamond. However it's a little hard to make a atom-thin layer of carbon into a paste...

I would take it in atom thick sheets.

If you were to use this sheet as a thermal pad, then your cooler and cpu mating surfaces would need to be flat on the atomic level. Not really feasible. The sheet would also probably break when installing the cooler, even though graphene is one of the strongest materials in the world.

This can be done badly though. Noticed this recently:

http://www.prolimatech.com/index.html

The manufacturer of one of SPCR's highest rated heatsinks clearly doesn't like people messing them up.

Yes I realise that, and thats exactly why I personally have never done it.

It really depends on your CPU and your heatsink, some heatsinks are more concave to match the convex shape of most IHS's.

It's really a last resort for those people that really like to overclock to silly levels.

But as I said, as long as you buy an aftermarket TIM as opposed to the sub-standard gunk that ship with most heatsinks, then you're unlikely to see much difference between the market leaders of TIM products.

quick! someone call dan from dan's data!

I read somewhere recently (link) that Fraunhofer have been touting diamond as a possible TIM component, but then they've been hawking diamond everything these last few months, right down to ball bearings, presumably in an attempt to secure further research cash. So my interest was piqued when this came up, as I'd be surprised if someone had beaten them to the punch with a product on the market.

This is Electrospell's business address. Seems it might be a bit of a garden shed operation...

Just reading all the stuff on their website, all the home made PDF flyers with random microscope images, I think it's odd that there is this sheen of it being aimed at industry, when the language and applications discussed are clearly aimed at PC enthusiasts. I think I will file this one under snake oil until someone proves me wrong.

I don't doubt there are thermal pastes out there for which some enterprising souls have found a marketable use for industrial diamond dust. I'm only guessing here, but I think there is a world of difference between some stuff that happens to contain diamond, and paste which properly harnesses the thermal properties of diamond.

How much of the thermal resistance between CPU and ambient air is typically accounted for by the TIM? If, for example it is 1%, then reducing this by a factor of three is only going to decrease your total thermal resistance by 0.7%, which is unlikely to be worth while. On the other hand, if it is (say) 10% of thermal resistance, then reducing it would be useful.

It wouldn't be hugely difficult to measure TIM thermal resistance scientifically. Here's one idea: we have various chunks of Big Fat Copper Rod (BFCR). The ends are polished flat, and we have fittings to hold two sections of rod very close to each other. The fittings have poor thermal conductivity. Two of the BFCRs have heatpipes attached. All BFCRs have a temperature probe. The basic experimental setup is to have the heatpipe BFCRs at each end of a 'stack' of BFCRs, with different TIMs at each join. (Typically we'd have just two interfaces - one for the TIM being tested, one for a standard TIM. Other configurations would be useful for callibration - figuring out how much thermal resistance is in the heatpipes and BFCRs.) Then the heatpipes would be put into water baths at two different temperatures. For example: if we have 3 BFCRs (two joins), bath temperatures 0 C and 60 C, and the middle BFCR turns out to stabilize at temperature 40 C, then the join on the hot side has half the resistance of the join on the cold side.
(The BFCRs also need to be insulated against the air, which adds complication.)

Anyhow, I don't think SPCR is up for this level of complication. Here's a simpler set up:
A rod of highish thermal resistance metal sits with base in a ice water bath, and projects some distance above. A temperature probe is inserted just below the top end. The top end is where we apply TIM to attach to a copper block, which is connected by heat pipes to the hot bath. The probe sees thermal resistance between its location and the cold bath due to the poorly conducting metal it is in. It sees resistance between its location and the hot bath due to the TIM, copper block and heatpipes. The better the TIM, the hotter the probe will be - but it will be difficult to callibrate into a degrees-per-watt measurement.

i think he means that, given diamond is the best material for thermal conductivity, and its diamond that has been turned into a thermal paste here, then penultimate is the wrong word to use. the sentence "...the penultimate material for thermal conductivity has been turned into thermal paste..." would seem to suggest that its silver that had been turned into a thermal paste, which of course was done many years ago..

now I'm no engineer (phew!) but surely any part in the chain between silicon and air (via IHS>TIM>hs base>heatpipes>fins) is going to be at equilibrium once temperatures have stabilised? i.e., presumably the TIM can only transfer heat as fast as it can be evacuated from the base of the heatsink and at a steady state (idle, full load) that will equilibrate. Otherwise you would see a steady increase in CPU temperature at a steady state, which you don't.
The only difference I can see this making is that it will marginally increase the speed at which that equilibrium is reached - so the CPU will take longer to reach steady temp at full load and less time to cool down. The only things that will reduce CPU temp at steady state is how much energy is going in (CPU heat) and how much is going out (heatsink>air transfer efficiency).

Then again, I'm no engineer

mattthemuppet --

You're right, but timing in cooling, as in life, is sometimes everything. Uh where do I go after that throwaway quip?!

If the TIM represents a 10% impedance in the heat transfer of the whole system, then improving it to 0% would actually lower temp in real conditions, where the thermal load does not stay steady, but instead bounces up and down depending on what the user is asking of the computer.

Better TIM simply means that it's less of an impedance or bottleneck. It doesn't means something else isn't even more of a bottleneck and hence you might not see the benefit. For example, with some heatsinks, the slow quiet fan we prefer is probably the biggest bottleneck to performance. With say a Thermalright HR01 and any 120mm fan at 500rpm, surely the difference in TIM will tend to be minimized because the airflow is the bigger bottleneck. With a 125W CPU instead of an 65W CPU, the difference in TIM would show up bigger again.

I think you can safely say that using better TIM ensures that your cooler does the best it can under whatever conditions you give it. The degree of difference may be negligible in may "silent PC" scenarios, however.

I’m a bit sceptic about using diamond dust in a TIM, at least when it comes to the expectations proclaimed.

Diamond is a very good thermal conductor (2.5 to 12.5 times that of silver, depending on exact composition) because of its crystalline properties where heat is easily transferred between the atoms. If you turn it to dust and try to transfer heat through the dust there will be lots of non crystalline contact surfaces where the thermal resistance is considerably higher.
Mixing diamond dust with a liquid to produce TIM will result in a mix where the thermal and mechanical properties of the liquid will be a, if not the, deciding factor in its overall performance.
The claims that this TIM is many times better than “conventional TIMâ€

Results of the IC diamond paste, average 3.38deg improvement to AS5 based on 125 users.

http://www.ocforums.com/showpost.php?p= ... tcount=433

I've used the stuff, improved my temps by a a couple of degrees. As said by mikec it makes a bigger difference with high power cpus or gpus, hence some of the >10deg improvements.

Cheese is no slotch either:
http://enthusiast.hardocp.com/article.h ... VzaWFzdA==

It would be interesting to compare this VS ocz freeze, arctic silver 5 etc (that ocforums.com post is not very trustworthy).

By definition things in stead state don't change, so most of what you just said is meaningless.

The TIM does effect the "steady state" CPU temperature. A higher resistance TIM increases the temperature difference required between the CPU and the heatsink to transfer the same amount of heat energy. This means a hotter CPU and a cooler heatsink. Also since the heatsink isn't the only heatsink you are also increasing the temperature of everything else connected to the CPU (i.e., the motherboard.)

Resistances like TIMs don't effect transient response much. Capacitance effects transient response, which for CPU heatsinks looks something like using water for a coolant or lots of copper mass close to the CPU. Neither will help steady state at all, but both can hold more transient heat.

I don't see why not. its the compiled results of users from a number of forums. Having a large number of results reduces potential for error. if i give you an OCAU post for example is it suddenly trustworthy? eg.
http://forums.overclockers.com.au/showt ... p?t=714640

If there's major flaws in methodology that make the results not useful (see OCAU link for details) sure point them out, but i don't think its helpful to dismiss a test purely on association.

MikeC, I'm disappointed to see you write this. Better heat transfer certainly affects steady-state operation, not just transient heating. Take the "no steady-state difference" conclusion to the next level, and you'll believe that the heatsink doesn't matter once temperatures stabilize. Hmmm... Something may be wrong with that idea.

QuietOC's later post is a more accurate description of thermal resistance.

No one mentions the stuff called "Liquid Metal"! First, it sounds cool. And I've read that it does a very good job as a TIM. I suspect that its benefits are not really about its conductivity (though that must be important too), but about its ability to make good atomic-scale contact with the surfaces you are trying to connect.

There were some downsides, at least to the original product I read about 6 months ago. It had a tendency to harden over time and eventually bond your cpu to your heatsink. Not so cool. And I seem to recall that some of the original formulations were highly toxic (I'm thinking mercury). But I wasn't planning on drinking the stuff. The links I can find now all claim to be non-toxic, so maybe that has changed (or maybe my original information was incorrect; net is full of nonsense and rumour mongers like myself).

I think there are a couple of competing products, but this is the first one I read about:
http://www.coollaboratory.com/en/product_pro.shtml

Note that I haven't tried it, so I'm not really recommending it. I have tried the diamond stuff (results were inconclusive) and lapping my heatsink (results were very impressive).

I didn't say better heat transfer does not affect steady-state operation, simply mentioned the more typical/realistic condition for a heatsink: The load varies, it's not steady at maximum load as in stress testing (although it's often more of less steady, as during most operations that call for extended typing like this). I agree that QuietOC's later post is a more accurate description of thermal resistance.

Fair enough. My apologies.