Gettin greasy ...

[Konrad]

Thermal grease, thermal paste, thermal gel, heat sink goop, heat transfer compound, thermal interface material - whatever you prefer to call it, the names and substances differ but the function and purpose is always the same.

I suspect that most people here at TBCS are well aware of the importance of cooling their PCs and the critical role thermal grease plays.

An introductory summary of thermal grease basics:

Thermal grease (by any name) is placed between very hot components and a heatsink; its purpose is to maximize efficient heat transfer between the part and the heatsink. Without it, even the mightiest heat sink will become inefficient (the gap would be filled with air, which is a terrible thermal conductor), leading to the part overheating and failing.

Failure on PC parts failure is often expensive, quick, dramatic, catastrophic, and permanent. A lot of electrical power flows through a lot of very dense circuitry and generates a lot of waste heat, if this heat isn't removed from chip's tiny surface area, it will zorch in seconds - literally exploding into a little blast crater filled with fused silicon slag. This explosion can sometimes blast a hole through the socket and smoke the entire motherboard.

Don't believe the hype? Check some videos here (http://www.youtube.com/watch?v=mXh_Nfom96o&feature=related), here (http://www.youtube.com/watch?v=xG0sGugsv28&feature=related), and here (http://www.youtube.com/watch?v=BSGcnRanYMM). Note that Intel or AMD, ATI or nVidia, no brand is immune and no manufacturer should be trusted - all complex large chips can and do get this hot (CPUs, GPUs, and "Northbridge" MCHs in particular). Overclocking will run them even hotter.

Almost every cooling system that exists involves some kind of heatsink or mechanism that takes the heat from one place and gets rid of it somewhere else. Virtually every one of these requires direct contact between the heatsink and the hot part and thus uses some form of thermal interface material.

Now, to the point ...

As I understand it, most of thermal grease products use some kind of sticky, greasy binder in which microscopic particles of some other substance (the efficient thermal conductor) are suspended. Countless formulations exist.

The binder is usually some kind of silicon gel, but manufacturers often cut corners by diluting with petroleums, plastics, and mineral oils. (This practice usually reduces final efficiency a little while making the manufacturing process a lot easier and cheaper. The consumer might cry foul, but remember, the alternative is paying substantially more and getting fewer choices on products that aren't significantly better.)

The suspended thermal conductor is typically
- ceramic oxide (cheapest and by far the most common)
- metallic oxide (higher cost and efficiency; typically aluminum, copper, or silver, sometimes gold)
- carbon (extreme cost and efficiency; can be graphite, diamond, fibers or filaments)

Wikipedia's Thermal Paste article (http://en.wikipedia.org/wiki/Thermal_grease) generally agrees with what I've learned, but it also includes mention of liquid metals (gallium, mercury, cadmium etc) while I personally consider them to be more of an exotic category of their own; they tend to be so expensive and toxic that I doubt many consumers will ever see one.

The chemistry of the thermal paste (especially the type and concentration of the main ingredient (http://hyperphysics.phy-astr.gsu.edu/Hbase/tables/thrcn.html#c1)) fundamentally determines how well it works. Brand doesn't matter, price doesn't matter.

Everyone already knows (or should know!) that thermal grease is absolutely required and any kind of thermal grease is infinitely better than none at all. To put it into perspective, even crappy cheap alumina-based thermal grease will transfer over 8000 times more heat than the same volume of air.

Avid overclockers and gamers automatically recognize certain premium brands (like Arctic Silver 5 (http://www.arcticsilver.com/as5.htm)). Most authorities agree that silver-based pastes are superior, though the measured temp improvements are very small (and sometimes even disputable, since the measured ranges often fall with the error margin built into most PC temp monitoring systems). Although I agree that Arctic Silver is a fine product, I feel that brand isn't important and any silver-based thermal paste should perform nearly identically. Of course, not all brands are created equal (http://www.overclockers.com/silver-thermal-pastes-buyers-beware/), so (like the linky says) buyer beware.

Most (if not all) OEMs and vendors ship their products with cheap bulk-discount thermal grease; it's generally good enough to keep the computer or computer part running *at rated specifications* until the return-period or warranty expires. After that, it's the consumer's problem, not the vendor's. I routinely clean this cheap junk off and apply a better (silver-based) grease whenever I obtain a new part, just to be sure. As a bonus, I can verify the actual part numbers on the exposed chip and install a thermal probe while I'm at it. Part numbers never lie (well, it's easier to determine if they're lying, while it's impossible to get the real truth from clever software.)

[Konrad]

Finally, my questions.

As informative/educational as all the above may be, it's all just "background" for my questions ...

Browsing around, I couldn't find many "carbon-based" thermal paste products. But I found Panasonic Pyrolitic Graphite (http://www.panasonic.com/industrial/electronic-components/protection/heat-sinking-material.aspx), after bumping into a little chat between a pair of EE guys (here (http://laserdiodetest.net/showthread.php?p=12)) where one says
Bare part 0.70 C/W
Thermal Grease 0.33 C/W
Pyrolitic Graphite 0.04 C/W

0.04 C/W sure caught my attention, since the "bare part" and "thermal grease" values are remarkably similar to what I'd expect on a CPU. Note that (if my math is correct) Arctic Silver 5 thermal grease would rate about 0.31 C/W.

My question ... if I were to grind this pyrolitic graphite stuff into fine, fine, fine powder and mix it into a silicone (or vaseline, or whatever) gel base ... would it work? No doubt I wouldn't get 0.04 C/W, but really anything under 0.31 C/W would be an improvement, no?

Another interesting idea ... what about using silver solder pastes as a thermal interface? Obviously that's not their intended application, but hey, they are available in a variety of % contents and such which are much more tightly regulated than thermal paste products. They're obviously fully conductive. They might take a lot longer to "cook off" before requiring re-application. As a bonus, they tend to be significantly cheaper, since a little $25 jar of the stuff would be good for hundreds and hundreds of thermal applications (maybe less if you use it for soldering).

Maybe some silver solder paste, powdered graphite, and a dash of WD-40 mixed together with a mortar and pestle?

Any thoughts?

[SXRguyinMA]

the first vid is fake, and the second two link to the same vid lol. other than that nice article

[x88x]

The exploding CPU video is a fake. The Tom's Hardware video is an accurate depiction of what will happen when a CPU overheats. The reason the AMD CPUs got so much hotter is because they were not made with integrated heatspreader.

As for your question, no matter what material you use, one of the key things to keep in mind is that you need to make sure that it will never leave the area directly between the heatspreader and heatsink. This would be my concern if a material (like silver solder) with a low melting point were used. If such a material were to leak onto your MBB, Bad Things(TM) will happen. The reason why silicon grease is used as a medium is not only to ease application of the material, but also to keep it in place during use. If you are going to make your own paste, I would use powdered diamond (http://www.dkhardware.com/product-13959-c0m1410-1-4-micron-100-000-mesh-diamond-compound.html) instead of graphite. It's really not very expensive.

I think Kayin has done some work in this area; he'll probably be around at some point.

[Oneslowz28]

Arctic Silver 5 is about the best TIM readily available and cheap. (you can buy it from newegg or Radio Shack for about $6) It takes 200 hours to fully cure, but you will see awesome results from it right from the start. But its not just about the TIM you use, you also have to make sure both surfaces are flat, smooth as possible, and very very clean. Here at TBCS we use Arctic Silver Arcticlean to clean both the CPU and the heatsink before we do any testing.

There are commercially available diamond based TIMs but they have a very low Diamond to medium ratio. Kayin and myself have discusses making some home made diamond powder based TIM. If you look at other high end TIMs the particle size of the base component is very uniform. There will be a mix of specific sized particles as to remove as much of the gap between particles as possible. One of the commercial products we looked at used 3 different sizes. IIRC it was something like 40% 20micron, 10% 10 micron and 50% 5 micron, and it used a silicone dioxide based medium to hold it all together.

We found a few people who had made it at home before, who were seeing awesome temps, but not near the potential diamond based TIM has. I did some research and found that the key factor in making your own TIM is to place your finished product in a vacuum chamber for X hours. This will remove all the trapped air in your TIM. This is where things got complicated for us. I have a vacuum pump capable of about 90kPa which was enough for what we want to do. The hard part has been finding a suitable cheap vacuum chamber. I tried making one out of iron pipe, pipe dope and end caps. I was only able to take it down to about 45kPa. There was still air leaking in from somewhere. So until I can find the time and $$ to build a proper vacuum chamber the project is on hold.

[x88x]

If you look at other high end TIMs the particle size of the base component is very uniform. There will be a mix of specific sized particles as to remove as much of the gap between particles as possible. One of the commercial products we looked at used 3 different sizes. IIRC it was something like 40% 20micron, 10% 10 micron and 50% 5 micron, and it used a silicone dioxide based medium to hold it all together.

Wouldn't it be better to have uniform, very small particles? For example, the stuff I linked is 1/4 micron, so the gap would be smaller with that than any mix of 20/10/5/etc micron stuff, right? Or is it that the larger particles transmit heat better since a larger portion of the TIM is the base component?

[Mach]

The hard part has been finding a suitable cheap vacuum chamber. I tried making one out of iron pipe, pipe dope and end caps. I was only able to take it down to about 45kPa. There was still air leaking in from somewhere. So until I can find the time and $$ to build a proper vacuum chamber the project is on hold.

Hmm, every considered using an investing table? I've got the domes around here somewhere. They're good for 27-29" of mercury.

http://www.ishor.com/VacuumInvestMach.php


Wouldn't it be better to have uniform, very small particles? For example, the stuff I linked is 1/4 micron, so the gap would be smaller with that than any mix of 20/10/5/etc micron stuff, right? Or is it that the larger particles transmit heat better since a larger portion of the TIM is the base component?

Heh, the first time my job knowledge ever applied to modding. Picture a box of marbles that are all the same size. Assume that each layer is packed as efficiently as possible on the layer below it. That's called rhombic packing or close packing (http://en.wikipedia.org/wiki/Close-packing). Now pour water in the box. The volume of water is equal to 27% (theoretically) of the total cube but for real life applications like TIM the empty space volume is at best 36% (likely more). That means that over 1/3 of the TIM that you spread is binder if the particle size is uniform.

If you do the experiment over but this time use alternating layers of small and large marbles, you could theoretically get the water volume down to 14% with end to end or cubic packing. In real life, you'll never get to 14% but you get the idea.

The particles in the TIM transmit the heat so the more particle volume that you have in a layer the better heat transfer. The paste is just the binder for the particles and does not transfer heat that well. Mixing particle size means more particle volume in the layer between the heat spreader and cpu.

[mDust]

Wouldn't it be better to have uniform, very small particles? For example, the stuff I linked is 1/4 micron, so the gap would be smaller with that than any mix of 20/10/5/etc micron stuff, right? Or is it that the larger particles transmit heat better since a larger portion of the TIM is the base component?

Perhaps having the larger particles reduces the overall thermal resistance by reducing the number of gaps of filler material? So instead of heat energy transferring to a small conductive particle, through some base material, to a small conductive particle, through more base material, to a small conductive particle, etc...it would transfer from a small particle, through some base material, to a large conductive particle, through some base material, to another large conductive particle...
The smaller particles would still be needed to fill the empty 'corners'.

I think we need a perfectly flat (down to molecule) diamond plate with a single drop of of liquid on each side of the plate. The liquid could be anything that is highly thermal conductive and contains a high-concentration of diamond molecules or particles in solution. Through capillary-action, the diamond solution fills the nooks and crannies in the HS and IHS and then evaporates, thus providing a mostly diamond heat-super-highway. Though, the price of the HS would be many times that of any computer you could possibly build at home.

[Kayin]

Let's just say guys, I have my eyes on this.

Good research Konrad, and pyrolytic graphite is an EXCELLENT material. It's easier to work with than diamond, cheaper, but at the same time it's still a few ticks off the capability as well. It's all tradeoffs.

What you guys need to do is get something like what AS5 is suspended in WITHOUT the silver. While it will settle in matrix, you'll get uneven conduction. Also, look for something that won't be easily pumped out by hydraulic pressure (that set of fractal lines when you remove a heatsink? it's working its way out...)

I'll read up some more on that graphite. Now if you can find me graphene, I'll dethrone diamond.

BTW, to get teh pyrolytic craphite to work, use silicone oil. A heavier weight, it'll create a non-Newtonian fluid and it should be easier to work with as well.

[Konrad]

Fake vids and broken links. Damn, I shoulda known better than to trust Youtube.

I'll admit the video with the Russian guys blasting their chips apart did seem a little bit staged, but it appeared sufficiently genuine to pass my casual scrutiny. I've little doubt that the boiling water and Tom's hardware vids are authentic. Fake or no, I think my point about CPUs being killed by thermal overstress is still valid, if perhaps not as exciting.

Vacuum chambers, molecular geometries, capillary matrices ... obviously I'm just a new kid on the block when it comes to engineering a better sludge. You've obviously advanced past my medieval mortar & pestle conceptions. If any of you guys manage to develop a superior TIM then I'll be one of your first customers.

I didn't seriously expect the solder paste idea to work, at least not without a hitch - just sort of tossing the idea out there. I wonder if it would convert into a dry solid after a few weeks of "burn in", so maybe I'll keep playing with the idea a bit. I do agree that having it leak out into the mobo socket would be a Very Bad Thing.

Use silicon oil to get the pyrolitic graphite to work? An interface on the interface? Wouldn't that still introduce a weak link in the chain that can be improved upon?

[Edit]

Another new (and probably bad) idea: what about a magnetically clamped heatink? Iron is an inferior thermal conductor, true, but some of the AlNiCo alloys would make fairly decent heatsink materials. Admittedly the CPU block is probably copper faced, but it can always be chemically plated with something else. (Powered electroplating isn't strictly required for most metals, it just makes the reaction faster.) I'll admit I have reservations about how stable n-billion circuit gates would be when in immediate proximity of an intense magnetic field. Again, just throwing the idea out there.

[x88x]

Another new (and probably bad) idea: what about a magnetically clamped heatink? Iron is an inferior thermal conductor, true, but some of the AlNiCo alloys would make fairly decent heatsink materials. Admittedly the CPU block is probably copper faced, but it can always be chemically plated with something else. (Powered electroplating isn't strictly required for most metals, it just makes the reaction faster.) I'll admit I have reservations about how stable n-billion circuit gates would be when in immediate proximity of an intense magnetic field. Again, just throwing the idea out there.

Maybe I'm missing something here, but this would just be to clamp the heatsink to the CPU, yes? In that case you would run into the same problem that TIM was created to solve all over again. Ideally TIM would never be used, and both the heatspreader on the CPU and the bottom of the heatsink would be perfectly atomically flat with absolutely no space inbetween them. Because this isn't feasible, we try and fill the inevitable empty space with the most conductive stuff we can get. Using a different method of attaching the heatsink wouldn't solve that problem.

[Oneslowz28]

Mach, that is the exact reason we came up with the 20, 10, 5 idea. If you can get the binder % down to around 15% or even less then the conductivity rises exponentially. The truth is though, that once you get to a certain point, you will only see temp differences in the fraction of a degree range. There are companies that spend hundreds of thousands of dollars on TIM R&D. They have lab grade equipment, lab grade materials, and scientist with Graduate and PHD level chemical education working to squeeze out that last few tenths of a degree. Sure we can make an awesome TIM but at what point does it become more cost effective to just buy a $10 tube of the stuff and call it a day? There is also a very good reason you need a liquid binder. With out it you will never get your TIM to lay even or "flow out" The key to a good TIM application is getting it to evenly cover the CPU and HSF base with the right amount. Too little and your temps suck, too much and your temps suck. This is the exact reason I spread the TIM on the HSF with a fresh razor blade. It allows me to uniformly cover the base and control the thickness of the application. Some people like to put a dab on the CPU and let it level itself. This is bad practice IMO, but that's an argument for another night.

I am still going to make some diamond TIM just because its fun and making things (http://themakersworkbench.com) is what I live for.

Mach, I thought about using a painters pressure pot or modifying a pressure cooker. If I can pull 80-85 kPa I will be happy. Do you think either would be sufficient?

[Konrad]

Spreading with a razor blade? That's magnificent, I've never thought of it! A tech-buddy of mine demonstrated "proper" application of the TIM as shown on page 120 in this old IBM service manual (http://download.lenovo.com/ibmdl/pub/pc/pccbbs/thinkcentre_pdf/19r1295.pdf) and I've stuck with it since.

I also know about honing and lapping the undersurface of the heatsink to a nice shiny bright and perfectly smooth flat finish to help insure best contact; apparently that's standard fare these days for anyone who's serious.

Can you impart any other worldly tricks, advice, wisdom, or linkies that might be helpful?

[Kayin]

http://www.thebestcasescenario.com/frontpage/?q=node/360

Read all the pieces, come back, and let's add a third mind to our TIM development group.

And if you wanna know why I'm not employed somewhere, I'm handicapped. I do this for the sheer love of it. So I can do whatever I want, including invent stuff if I feel like it.

We need to make up a batch of this stuff, CJ. I have an idea...

[Konrad]

Ha, that's a ton of info; important details I knew but you evidently know better, important details I'd never considered examining. I won't be the first to state your guide is a Must Read.

It led me to some pages about Intel IHS Lapping (http://www.legitreviews.com/article/402/1/) (which I've done) and Removal (http://www.legitreviews.com/article/402/2/) (which I don't think I've got the balls to try) - seriously, removing the IHS from an OC-worthy processor with a blowtorch? Do you actually do this sort of thing? I can certainly see the advantages of getting a crappy factory IHS off, modding or fixing or replacing it with something more efficient. I'm already haphazardly thinking about directly immersing an exposed running processor die to circulated Midel ... workable?

[Konrad]

And a question in a bit of the opposite vein:

Is it possible to overcool a processor? To actually bring the running part under minimum operating spec or cause some kind of thermal shock/failure between it and the surrounding electronics (motherboard socket, etc)? I'm not talking LN2 or some other exotic, but I'm asking if a point is reached where there's electrical or physical strain between a very cool part and the hotter parts running nearby.

[mDust]

...seriously, removing the IHS from an OC-worthy processor with a blowtorch? Do you actually do this sort of thing?
I don't think anyone here does this regularly, but I wouldn't be surprised if some of the 'crazies' here have done it before. :D

I'm already haphazardly thinking about directly immersing an exposed running processor die to circulated Midel ... workable?It's workable, although only if the liquid is actively cooled as well. Otherwise, the liquid will continue heating until it reaches a relatively high plateau. At that point it's no longer effectively cooling the immersed hardware. I've been planning an immersed rig for over a year now...:(...one day, if I stop making changes to it, I'll start the build.:)

Cooling components well below freezing temps puts stress on them, but you aren't going to accidentally over-cool your chip!;)

[Kayin]

I ran a direct-die cooled chip. It worked, and I set a record with it, but I'll never do it again.

And I AM one of the crazies here.

Oh, and BTW, I have torched a chip to remove an IHS. It was fun.

[mDust]

And I AM one of the crazies here.

/nod
:D

[Konrad]

Non-blowtorch alternatives for IHS removal? I really suspect even momentary exposure to that sort of heat will degrade the part.

Do iCore processors have any embedded firmware? (I'm asking because mechanical cutting-drilling methods might generate ESD.)

[mDust]

Non-blowtorch alternatives for IHS removal? I really suspect even momentary exposure to that sort of heat will degrade the part.

Do iCore processors have any embedded firmware? (I'm asking because mechanical cutting-drilling methods might generate ESD.)

I think the blowtorch is necessary to melt solder under the IHS. The solder is what holds the IHS firmly to the chips below.

I don't think any processors have embedded firmware...though I could be wrong. If you ground the IHS it shouldn't be a problem. I think you better be certain where those cores are before you start drilling/dremeling into it!;)

[Oneslowz28]

Note that removing the IHS will cause most current Air Coolers to not properly clamp to the CPU any more. This is because of the space the IHS fills between the die and the cooler. The result will be less than the recommended clamping pressure.

[Oneslowz28]

Oh and you can crush/crack the die also.

[Konrad]

The best info I've found (so far) is here (http://www.xtremesystems.org/forums/showthread.php?t=222040) and here (http://www.overclock.net/intel-cpus/305443-ihs-removals-how-do-should-i.html).

Some people use razors to work the corners/edges off a bit.
Some recommend working on the CPU while it's warm (fresh out of the socket after stressloading) while others recommend keeping it cool (fresh out of the fridge). A few guys use high-temp torches while others use bic lighters.
Some use a mill (or even dremels or just sheet abrasives) to precision cut or grind metal away.
Some even lap to exposed silicon to mirror shine after the IHS is off.

As many people bust their i7s as succeed.
No doubt they each have different measures of skill and patience, but I think it's fair to assume that nobody wants to just throw hundreds of dollars away for the simple hell of it. For me, for now, regardless of the approach, the risk just appears to be too high to gain that potential 3-7C temp advantage.

[Kayin]

No, and no.

Also, no degradation of the part.

The degradation comes from heat because the channels in the chip will swell, actually choking electrons, causing the channels themselves to degrade. Hysteresis, causing cascade failure.

[Konrad]

By degradation I mean something that is common to all electronic parts, from humble resistors to mighty processors. No part lasts forever, they all degrade over time, some parts even while being unpowered. Not to exaggerate, of course, a decent CPU should last many years, I still have ancient Z80 and 650x parts that work well after nearly a quarter of a century. But the parts do degrade over time, they age, and thermal stresses are the surest known way to accelerate this aging. Not at all the same thing as a part failing.

So a degraded part will still work fine. It just won't have the exact value it should, it will drift, it will run a wee bit hotter or slower or less efficiently, it might even still fall within normal tolerances for a "new" part of it's kind. Degradation can be subtle and sinister and impossible to detect aside from long-term performance comparison or logging with identical parts.

I worry about an i7 because it is a very delicate, complex, dense *expensive* part. Even minor degradation can result in some particular minor and seldom-used circuit breaking or a thirst for 0.005 more volts to run stable or a loss of 100MHz ... all meaning it will not run as fast or stable as before ...

I suspect you probably already know all this, but just wanted to be clear.

Can you be confident that a few seconds of blowtorching won't degrade the part? (I would like to be before trying this!)

I know it's easy for a noob with a blowtorch to make mistakes - but I'm willing to overlook that for the moment.

I know the CPU has a generous heat threshold, after all the process of just gluing the damned IHS in place with solder was probably ~230C anyways. And it's rated to run as high as ~65C during "normal" operation, perhaps 90C before catastrophic failure. But how much aftermarket heat blasting can the part sustain before degrading?

[Kayin]

So long as you're not torching the green substrate or the LGA lands, enough blowtorch for a while.

It's low temp solder, BTW.

[Konrad]

Does the IHS contact the silicon anywhere other than at the solder joint?

[mDust]

Does the IHS contact the silicon anywhere other than at the solder joint?

It shouldn't. You can think of the solder as a TIM between the die and IHS. Because it's all designed to conduct heat away from the die, it's likely just as conductive the opposite direction also. If you're thinking about doing something like this, I'd give it a go on some cheapo eBay chips or something first before 'noobing up' an i7.:D I won't be trying this to avoid noobing up my own chips.

[Konrad]

I've got piles of inferior (<2.66GHz) P4 Northwoods available ... none of 'em worth more than 5 bucks these days, so I was thinking of practicing.

I'm certainly not a master craftsman (yet) but I've got plenty of confidence in basic tools (done MCU projects for years), including flame soldering SOICs and 1206 SMTs by hand, so that doesn't scare me. Busting an i7 does.

[Edit]
Now, I'm not rushing out to grab a new i7 just yet but I am targeting the i7-980X (http://ark.intel.com/Product.aspx?id=47932&processor=i7-980X&spec-codes=SLBUZ) ... I'm waiting a bit because I want the 2nd gen version (1st gen products always overlook numerous bugs/flaws, and the i7-980X is 1st gen in many ways; # of cores, extreme QPI, cache size, lithography, etc)

To get a basic idea, I'm seeing the package size = 42.5x45.0mm. What are the dimensions for the die itself (after IHS removed)?

I'm asking because pyrolitic graphite isn't expensive in small wafers, but larger chunks get costly pretty quick. Example here (http://www.kjmagnetics.com/products.asp?cat=85&gclid=CPq7ifrbxqMCFR5ggwodPB-Hsw) - the PG1 and PG3 sizes are < $7, but the not-much-larger PG5 size costs $55; obviously I'd buy the cheap ones if they're big enough, and waste less material when grinding down to required thickness. I'm assuming this stuff (plus 1-2 layers of good TIM?) will be significantly superior to mystery solder plus aluminum/copper IHS. Can't tell yet what the surface is like, if it can be smoothed/honed or it's rough.

...

Any thoughts on some kind of machined nubs which use a Lego-inspired processor/heatsink interlock? Assuming machinist (or even Lego) tolerances for flatness/etc ... wouldn't it actually slightly increase total direct surface contact area? I suppose thermal expansion might be an issue (parts might "stick" when too hot?), but I really don't know if it'd be a factor within these temp ranges, especially if things are clamped.

[x88x]

I think the smaller graphite wafers will probably be big enough; if you want something to base a rough guess on, look back at the last CPU line that didn't have an IHS (AMD Athlon XP). Using that as a guestimate base, and taking the die size of the time into account (150+ nm, iirc), then extrapolating the increase in number of cores/etc and the shrinking die (i7's are all 40nm atm), I would guess that it's at most maybe twice the size.

As for interlocking nubs, if you had identical materials on both sides it might work out well; my concern would be if each side was a different material and one expanded more than the other from the heat. Best case the inner material would expand more and you would just have to worry about mechanical stresses on both materials. Worst case, the outer material would expand more and the contact surface would decrease the hotter your CPU got...a very bad thing, to say the least. I could be wrong, but I think graphite would expand less than copper.

[Kayin]

Far worse on Lego. Google gage blocks. You'll learn a lot.

I'm leery on pyrolytic graphite by itself. Especially with extra TIM layers. Anything but metal to metal with no air is significantly worse. Try soldering a sink on and see what happens. That should be a really good idea (no joke.)

Google Indigo Extreme. Perfect.

[Konrad]

Ah, the interlocking idea would probably work best with (milled lego-top) copper IHS and (milled lego-bottom) copper HS. Actual lego dimensions seem pretty tight and would serve as an excellent starting point, but I'm not sure what the actual blow-molded-ABD lego part tolerances are (I'd be surprised if they couldn't be bettered with CNC metal machining). Apparently copper is entirely unique in that it counterintuitively shrinks when heated instead of expanding like every other metal, don't know if that'd be an issue. I suppose incredible (beyond my ken) materials engineering could compute alloy mixtures which would expand/contract at perfectly matched rates at the point of contact across the entire expected thermal range.

It's such a cool idea that I really wanna try it out just for the hell of it; trying to machine a gap as small as my tooling will allow (±250µ" for each piece, so ±500µ" overall, maybe less with soft-metal copper). And who else would be able to use his CPU as a lego-building platform? Maybe I could even stick the heatsink onto a lego robot, heh.

[Edit]
@Kayin

I'm a CNC machinist; my gauges are pretty good, I assure you.


I'm leery on pyrolytic graphite by itself. Especially with extra TIM layers. Anything but metal to metal with no air is significantly worse.
So basically ... removing the IHS is a essentially wasted effort if I'm not going to cool the die directly?

Indigo does look impressive. Any real-world comparisons vs Arctic 5?

[Kayin]

the issue with the lego sink is getting TIM where it's supposed to go, and air gaps. It would prevent capillary pumping as a benefit, but it's not an idea I'd like to go with.

I want a pyrolytic graphite waterblock. Now if I only had a silver nanofluid to pump in that...

Edit at Konrad:

My apologies. I've studied this VERY well, as you can see. I'd be more than willing to help you with some ideas if you'd like. I've been designing but getting nowhere with machining for a while.

[Trace]

I want in on this!

[Kayin]

Konrad, I have a deal for you. You want something to test it on, I'll help you out. I have a 4870x2 sitting here. Card IS operable. You want to design a cooling system using anything you want, and TEST it, go for it. Want to make a TIM, or a block, go for it.

I'd love to have blocks for this card, which is still top of DX10, but they're phasing out blocks for. I think it would make a perfect test subject, as it's a blast furnace.

Please do not kill it, it was $500 dollars. When we get something, it would be nice to have two of them. I've got her sister at home.

I've got the card to put my money where my mouth is. I'll help you figure out any idea you've got too. But you seemed to have the right skills to actually help here.

[Konrad]

lol, I'm not interested in zorching your $500 card with a fumbled IHS removal ;)

[Trace]

No, there is no IHS on a GFX card. It's straight silicon.

This is a 5870, but same idea:

http://gallery.techarena.in/data/513/medium/Radeon_HD_5870_Graphic_Card_Chipset.jpg

[Konrad]

Gotta love AMD (er ATI) for keeping it simple!

I wonder if Intel's sloppy IHS's are endemic in the industry, if so then it's obvious why AMD/ATI have eschewed them.

[Edit]
Looking more closely at that sexy pic ... are the large solder fillets on the edges of the die intentional? Do they cover up incredibly small wire leads?

[SXRguyinMA]

I think that's just excess TIM from the stock cooler :D

[Snowman]

Just want to let everyone know I came to this thread, I learned, and I now.... still hate chemistry with a passion for some reason. Love algebra hate chemistry which is probably a/b(t*y)% algebra anyway.

[x88x]

Gotta love AMD (er ATI) for keeping it simple!

I wonder if Intel's sloppy IHS's are endemic in the industry, if so then it's obvious why AMD/ATI have eschewed them.
I believe all modern CPUs use IHSs, even AMD. I think the reason they're not used on GPUs (iirc, nVidia doesn't use them either) is because the chip itself is so freakishly huge that it already has plenty of surface area to contact the heatsink, whereas with a CPU the chip itself is quite small and the IHS serves to increase the contact area with the heatsink.



[Edit]
Looking more closely at that sexy pic ... are the large solder fillets on the edges of the die intentional? Do they cover up incredibly small wire leads?

Like SXR said, it's just TIM; those chips are actually surface-mount with all the contacts on the bottom of the chip.

[Trace]

All moderon CPUs from AMD and Intel use IHS's. Not sure about Sun, IBM, etc chips, and yes, nVidia (last I checked) does not use IHS's probably due to what x88x said...Freakishly HUGE chips.

Also, extra TIM

[Kayin]

Most chip companies use an IHS, simply to cut down on returns. If it runs 10C cooler, but you get a 30% OEM return rate, nobody cares.

[mDust]

Note that removing the IHS will cause most current Air Coolers to not properly clamp to the CPU any more. This is because of the space the IHS fills between the die and the cooler. The result will be less than the recommended clamping pressure.
I was thinking about this as well, and it's easily avoided by milling the edges of the HS base to fit within the socket cover. Though cracking the die itself is a possibility, I would expect one to be extremely careful with a several hundred to thousand dollar chip.



I'm asking because pyrolitic graphite isn't expensive in small wafers, but larger chunks get costly pretty quick. ... I'm assuming this stuff (plus 1-2 layers of good TIM?) will be significantly superior to mystery solder plus aluminum/copper IHS. Can't tell yet what the surface is like, if it can be smoothed/honed or it's rough.

Any thoughts on some kind of machined nubs which use a Lego-inspired processor/heatsink interlock? Assuming machinist (or even Lego) tolerances for flatness/etc ... wouldn't it actually slightly increase total direct surface contact area? I suppose thermal expansion might be an issue (parts might "stick" when too hot?), but I really don't know if it'd be a factor within these temp ranges, especially if things are clamped.

So you're planning on risking a thousand dollar processor to replace the copper IHS with a slightly more conductive material? It's not worth it. Mill an air cooler or waterblock to fit directly on the die. The less layers the heat has to travel through, the better. As for the 'lego' design, there would be absolutely no way to effectively apply a TIM to that surface, so you'd end up with a net effect of less efficient heat transfer. If you were machining a part to mate flush with another, it would be much easier to minimize tolerances by making them completely flat rather than any other shape.

[slight thread jack]
In my own bid for a new TIM, would it be possible to ionize atomized copper powder so the particles mostly repel each other thus filling any air gaps? The powder could be suspended in a fast evaporating liquid that essentially leaves a 'solid' copper film between the parts. This would be incredibly hard to store for any amount of time, I'm sure, but it might be feasible to whip up a batch the day of the processor installation.
[/slight thread jack]

[Konrad]

Shows what I know ;)

Well, obviously a superior TIM is the way to go on GPU ...

[Konrad]

@ mDust

Your savage debunking of my wild lego-block idea sadly makes perfect sense. Simple flats are better than interlocks. I can see some arguments promoting perfectly mated curved surfaces since they'd have more surface area than flat, but I think the better perspective (assuming same metals on IHS and HS) is to remember that the only practical effect of these curves or interlocks or whatever is to increase the surface area (and air gaps) of the inefficient TIM layer within the otherwise homogenous thermal block.

Not sure how your ionized copper idea would work. How would you prevent ions from interacting with electronic circuits? Wouldn't an ionized metal be highly corrosive? If exposed directly to the die then condensation might become an issue. If not exposed directly to the die (ie, used within a sealed vapor-phase loop cooler) then you'll need something else to act as the TIM. Just my thoughts.

[mDust]

@ mDust
Not sure how your ionized copper idea would work. How would you prevent ions from interacting with electronic circuits? Wouldn't an ionized metal be highly corrosive? If exposed directly to the die then condensation might become an issue. If not exposed directly to the die (ie, used within a sealed vapor-phase loop cooler) then you'll need something else to act as the TIM. Just my thoughts.
It wouldn't come anywhere near any circuits. Potentially corrosive? Yes. Condensation? Possibly, but I doubt it. Under normal circumstances if the liquid is evaporating, it's not going to be condensing at the same time.
Ideally, the copper powder would be sandwiched between a copper IHS and a copper base-plate of the HS. As the liquid evaporates it floats away, never to be seen again. It would be like a fraction of an oz so it shouldn't take too long to evaporate away and 'cure'. Maybe a day or two of operation?

[Konrad]

It seems a bit bizarre to me. A continuously eroding heatsink? I understand phase cooling, and I accept that some kind of liquefying copper process might be possible below the normal liquidus/melting range for copper (~ 830-1080C). Obviously you're not talking about a molten copper heatsink, just one where copper ions are continuously flowing away like a liquid (and presumably taking a lot of heat with them) ... how? Using excitation of lasers or voltage pulses to stimulate the exposed copper? Using MHD to cycle the "liquid" ions away? I'll admit I don't quite understand your concept. Not saying it can't work, just saying I don't understand it.

[Edit]
It occurs to me is that all this reduction of copper will result in another problem. Either a vacuum is created, or something else (probably air) fills the volume of the displaced copper. I would suspect that air would be uncooperative and insist on coating as much copper as it can with oxides. An oxidizing reaction would produce (a lot of) it's own heat while converting copper (both the ions and the metal base) into nasty thermally inefficient patina.

[Trace]

No. He wants to suspend the copper atoms (ions) in a liquid (water, or such) and apply it like a TIM. Then the water would evaporate, leaving just copper filling the gaps.

[Konrad]

Ah, an aqueous soldering approach. It's just an application of "electroplating" - remember that electricity isn't strictly required to make the reaction work, it just shifts the equilibrium point and makes it much faster.

It's a good idea. The obvious downside is you'd have a hell of a time ensuring even evaporation. Any trapped pockets of water will drastically reduce efficiency of the cooling interface, along with other problems. You'd also have to constantly add more (aqueous dissolved) copper into the solution at a controlled rate, as the molar concentration will keep on diminishing while it deposits. Any impurities in the water would likely turn into Slightly Bad Things. I imagine you'd want to apply heat (perhaps 100C water boiling heat) to finish the join before x-mas.

I think it's possible. It might be easier (more economical) to simply copper braze the two parts together, or machine them from a single block in the first place, then attach to processor afterwards. Probably won't get as good as just tossing the IHS out the window.

[Kayin]

all right, wow!

I've gotten all messages sent to me, I'm currently fighting a literally life-threatening infection but I've read, studied, looked some more and gone back and done more work.

Konrad's got it with pyrolytic graphite, if I read that PM correctly. Let me map out some stuff on paper, and then I'll propose an experiment. I want to make sure we get this to work right.

Guys, we're on the cusp of something here. Konrad and I have got some stuff going, and I wanna see what CJ has come up with on the diamond paste front-I think we could set some records.

But first, more pills. All messages should be sorted and replied to tomorrow. All I have to do is work on this.

[Trace]

Keep me posted too Kayin!

[Konrad]

As amazing as highly oriented pyrolytic graphite (HOPG) might be serving as a TIM, I expect it might be a slight evolutionary refinement instead of a completely revolutionary one. Knock temps down by a handful of C's at best, maybe not even that much since Arctic 5 is already pretty kickass stuff. As I've explained to Kayin, application is a bit more involved than with a "normal" TIM but still well within the abilities of serious modders and OCers. And it doesn't require any blowtorching, even though I'm sure somebody crazy will figure out a way to do it.

@ Kayin

Don't kill yourself, man - good luck and best health. K&K cooling can wait.

My PMs only discussed my findings on (HOPG) TIM. Never got around to addressing the HS.

To be honest, there's really not anything I can do to improve on decades of HS engineering. What I can do is fab parts that use only best materials, intelligent design, and finest precision. Any basic HS improvements you see will only be the result of not cutting corners.

I assume you'd want solid copper throughout, of course. Silver would serve better but is a wee bit too expensive for this project, solid carbon forms (diamond, graphite, fiber, nanostuff, etc) would be far more expensive and probably impossible for me to obtain or machine in any event.

I'd like to know if you're looking at a HS/fan block or a waterblock. Either way, I need accurate dimensional data. In the latter case, I'd need to know all about your fittings and such, plus what sorts of liquid coolants you circulate if they have corrosive properties - I might be able to select alloys, plating, or finishes which resist chemical attack without much reduction in thermal efficiency. I've never actually watercooled a PC myself, but I think I'm well-versed in the basics. Any wisdom you'd care to offer (beyond the linky already given) couldn't hurt.

[Edit]

Does anyone have any real DATA on this indigo extreme stuff? All I can find is marketing blurbs and amateur reviews, but I am intrigued. I am of course already very familiar with Arctic 5 and similar products, so anything less doesn't interest me.

[Kayin]

Konrad, I'm working on a PM as a feasibility study on some things, and I'll sort materials there. I've some ideas kicking around in my head...

http://skinneelabs.com/tim2010part1.html

There's an actual review lab's take on the deal, with proper repeatable results.

And folks, if you give me 10 bucks an application, I can make it for you too! It's low-melt silver/indium/bismuth alloy. No gallium, which I'd expect to see, but without an MSDS, I can't tell you much more. I'm still hunting one.

I'm betting that we could do the same with solder, if we wanted to risk some boards. Who's with me?

[Konrad]

I thought the solder paste idea was already sort of ruled out?

[Edit]

The Indigo Xtreme MSDS isn't published ("As Indigo Xtreme is considered a manufactured item and does not release any hazardous chemicals, it is exempt from requiring a MSDS.")

The Enerdyne Solutions site here (http://www.enerdynesolutions.com/tech_papers.html) has the Indigo2 datasheet and a variety of technical papers, application notes, etc. None of them specifically state Indigo's proprietary formula or constituents.

Indigo complies with RoHS; it therefore contains no lead, cadmium, or mercury (outside of trace impurities). The advertisments state absolutely no lead and state no gallium.

Indigo is "patent pending" the old trick that let's a company secure their patent claim while delaying submission of an official patent application (disclosing all the proprietary details) into public domain. ("patent pending" might also mean it's something worthless that needs a bit of "official" credence, you haven't yet prepared a proper patent application, you can't afford to, or you may withdraw or change the product ... I personally doubt that's often the case with big companies like Enerdyne.)
Nonetheless, I searched through all the online patents involving Enerdyne, PCMA, and TI anyhow with no luck.

Enerdyne seems to hold a monopoly on this product; Indigo is not rebranded and sold by any other Semi-Therm registered manufacturer/vendor (listed here (http://www.semi-therm.org/history/st24exhibits.html), overviewed here (www.semi-therm.org/pdflib/ST24-Final-Program.pdf).)

Indium-containing alloys have intermetallic problems with copper. These problems might be addressable with the addition of other metallic components, or they might not.

I've just ordered the $22 Indigo Xtreme LGA1366/i7 kit from DazMode (the only vendor of this product in Canada); I will sacrifice it to the metallurgical boys and their spectrum analysis to determine exactly what the precise constituents of this alloy are (ideally so I can have cheap batches made in-shop for personal use).

I'm a little unclear if Indigo Xtreme and Indigo2 are the same product.

[Kayin]

I've got an idea that I picked up from my silver nanofluid that I'm wondering about for everyman's diamond paste.

I'm thinking about acquiring a sample of polydimethylsiloxane and suspending diamond powder in it. That stuff resists pumping more than just about anything short of a brick, and if I load it enough, I get a bouncy, high-yield TIM that ca be used as a rubber ball if I'm bored.

I think CJ's vacuum chamber experiment may have been misplaced.

Alternately, I've got the means to whip up a few things of my own. I think I have a real idea going here. The question is do I have access to what I need to make it work. More in ten.

Bugger. PDMS is a GREAT insulator-like heat-resistant reentry tile great. More searching.

[Trace]

Literally 10? I'm waiting!

[mDust]

Literally 10? I'm waiting!

We all are waiting! I logged back on just to see this...:D

[Kayin]

Graphite. As in dry graphite for film lubrication. It's almost as good as diamond, and available in GREAT quantity for dirt cheap. I'll make my own paste up this week.

[Konrad]

Well, if you're gonna go with graphite, then HOPG would seem the best choice ...

[Kayin]

Sure would.

Now what was the price on that? You are absolutely correct, pyrolytic graphite is the best, but considering dry graphite is so bloody cheap, we could always tack together something workable and test it. The stuff you were talking about seems well beyond what most people can do. This will be a lot like AS5, but better thermally, we'll have the ability to manufacture in bulk, and we can market it with little issue.

If you can get it cheap, though, I'm all for it. I can make a base for it on the fly, so if you can get it cheap I'll make up some of it and test it too.

I love having compounding materials here at the house...

[Konrad]

lol, you could always try pyrolysis-at-home with a pressure cooker. End result might not be "purest" grade, but it should be good enough and low cost. You'll know it's done when a sample can be magnetically levitated.

[Kayin]

If you'll get me the instructions, I'll try and get a cheap pressure cooker for the thrift store. This could get FUN!

[Kayin]

This has not been abandoned. If I get a pressure cooker, how much pressure?

[mDust]

If I get a pressure cooker, how much pressure?:? Geez...what are you doing now?
10-20psi I think. 15psi is the standard cooking pressure.

[Kayin]

To make pyrolytic graphite?

[Konrad]

lol, you could always try pyrolysis-at-home with a pressure cooker. End result might not be "purest" grade, but it should be good enough and low cost. You'll know it's done when a sample can be magnetically levitated.
Haha, er ... sorry to get your hopes up with my own high hopes.

I think you'd need some serious industrial pressure. Like, hydraulic compression in a blast furnace kinda stuff. Not as much as you'd need to synthesize diamonds, but not far off.

You could try pressure-cooking at home, ramp it up as hot and high as you can (safely) get it. The results would be pretty interesting.

[mDust]

You could try pressure-cooking at home, ramp it up as hot and high as you can (safely) get it. The results would be pretty interesting.

HAHA! As interesting as it would be, I would publicly advise against this... Why? *BANG! SIRENS. FAMILY CRYING AT YOUR FUNERAL.* That's why. :)

[Konrad]

... ramp it up as hot and high as you can (safely) get it.
lol, no doubt Kayin is more qualified that I am at mad science anyhow.

But a safety reminder never hurts, eh? Modding ain't as much fun when you're bleeding and burning.

[mDust]

Modding ain't as much fun when you're bleeding and burning.
Hey, don't knock it until you've tried it!:D

[Snowman]

Hey, don't knock it until you've tried it!:D

I have tried it but it's so great.. cutting your self with plexi SUCKS.

[Kayin]

Update, I do have a partially completed TIM sitting in a jar here at the house. Currently waiting on some more high-quality graphite powder to finalize it. Keep watching for more updates.

[Kayin]

Preliminary testing (will be posting graphs soon) shows that it actually works!

It's (without giving too much away) a mineral oil based TIM that's bulk loaded with aluminium powder and high-quality microfine powder graphite (no molybdenum) with some calcium oxide to keep everything in solution.

My testing is limited by the fact that my i7 shipped with a half-height cooler that's pretty much crap, and I have no other viable coolers to test with. However, it did pass a 30-minute OCCT test, which is the first step in verifying everything. Tomorrow, I'll be getting syringes to load this up and get this ready to send out to a few people for testing.

At this point, I'm not completely sure just HOW good it is, but I know that it acted in certain situations better than the cured MX-2 I had on there before. How much of that is the TIM and how much of that is the crappy, crappy cooler is yet TBD.

[Trace]

Get me a tube. I'll test it