Slashdot Mirror
DIY CPU Thermal Grease, Using Diamond Dust
tygerstripes writes "The dysfunctor has spotted an impressive project over on InventGeek.com; an innovative chap has developed his own thermal compound for improved CPU cooling, using diamond dust — the best available material for thermal conduction — as the key ingredient. In spite of the quick-&-dirty DIY nature of the project, the gains in cooling performance are remarkable, especially considering the material cost was only $33. Given the price many enthusiasts will pay for a top-end cooler, it's easy to imagine this product coming to market quite soon."
...a practical use for your wife's jewelry!
I feel super cool with angel dust, I don't think I'll switch.
I'm god, but it's a bit of a drag really...
I admit, I don't know jack about jewelry, but ain't diamonds the kind of carbon that's supposedly expensive?
We used to have a Bill of Rights. Now, with the rights gone, all we have left is the bill.
I'm gonna throw out a spoiler: In a test situation, at full load, the best temperature under Arctic Silver was 57C, while this diamond dust compound achieved 38C. The nearly 20 degree difference is huge, and would definitely make a difference in overclocking. I'm hoping the price can come down when produced in industrial quantities, because it'll be enormously worthwhile.
That is an impressive gain. I wish they would have tested it against other thermal compounds though, I hate arctic silver it is to conductive and I'm to twitchy.
... diamonds! Like heatskinks, processors, memory ... probably cost prohibitive though.
in the do-it-yourself flashlight-making community.
Example:
http://dmcleish.com/CPF/L1-Diamond/index.html
Fifteen years later, my brain still makes Saved by the Bell connections. Help me.
0 = 1 + e^(Alt something)
Chagrined SHAD0W is chagrined!
In the do-it-yourself LED flashlight-making community. Example: http://dmcleish.com/CPF/L1-Diamond/index.html
Ahh yes, my diamond dust. Now where did I put that? I'm pretty sure I keep it with my scrap gold somewhere....
This guy's the limit!
Finally... I can take my square-headed girlfriend to the next level.
This technology is not new. Diamonds have been used as heat sinks and thermal conductors for processors in sattelites since the 70s (natural diamonds in big sizes, so expensive). Since about 1992 there have been succesful efforts to sinter diamond dust (waste material from the polishing process of gemstones, and now increasingly synthethic diamonds, both are not expensive) for use in thermal conductors.
Lighten up, Francis.
Look up Gemesis and Apollo. Technology has progressed to the point that we can synthesize diamonds. I don't mean make cubic zirconium, I mean real diamond, made in a lab. It is still expensive in relation to a lot of materials, but it is cheaper than mined diamonds, and getting cheaper.
B&W use it for their tweeters in their high end speakers, as an example.
Are diamonds now a nerd's best friend?
"The tree of liberty must be refreshed from time to time with the blood of patriots and tyrants." ~Thomas Jefferson
That guide was posted a while ago...
Here is a company that sells it premixed and has been around for several years.
http://www.innovationcooling.com/
An idea that was recently (2 days ago) discussed on Hack a Day was using diamond stropping paste
http://hackaday.com/2009/08/03/diamond-thermal-paste/
Here is a reliable review of Diamond thermal paste.
http://hardwarelogic.com/news/137/ARTICLE/2752/2008-03-03.html
As far as just hoping on Ebay and buying your own, good luck. Finding a distributor for your dust and suspension will take you some time.
I think the most important thing people should consider and hasn't been mentioned so far is that the equation for heat transport is linear. Changing the adulterant in the suspension will be more noticeable as the temp increases. IMHO for most people it's a total waste of materials, for some people it's a decent alternative, and for a very small number of people it's a good idea to spend the time locating materials to make a batch.
...cue up some Jeff Beck
XKCD:Xeric Knowledge Comically Dispen
$33? You can buy a tube of IC Diamond for only $7. DIY isn't too interesting when it costs so much more than off-the-shelf...
-- *My* journal is more interesting than *yours*...
The future is in a completely different molecular organization of carbon: graphite.
Maybe it's a hard hack because diamonds are the hardest material on earth. Just a thought.
Just mention "Screetch" at a Dustin Diamond stand-up concert...
He'll break you of that notion REAL QUICK and it'll be fun for everyone else. >:-)
Since no one reads the article and seems more interested in whining about diamond prices, here are the improvements that were achieved with the new thermal paste.
Idle - Max load
42c - 57c - Arctic Silver with a fresh application
39c - 54c - Arctic Silver with 2 week cure
29c - 38c - Diamond Grease with a fresh application
These numbers are almost too good to be true. A 19c drop under load with diamond paste? With my 4GHz OC'd system, I'm happy getting 38c when idling. If this data is true, it will really be revolutionary.
produces something available for order. No reason to DIY.
http://www.heatsinkfactory.com/ic-diamond-24-carat-thermal-compound.html
_______
I just wish I could c:\format Internet
Hardest "natural" material on earth. Aggregated diamond nanorods are currently the hardest.
http://en.wikipedia.org/wiki/Diamond_nanorods
Sorry to be a bastard, but I'm a bastard.
Diamond Powder Compound Cost:
100 CTS 60k mesh industrial Diamond Powder: $50
(0.2g : Carat)
36g of Silicon Grease: $20
Final Volume: 56g @ $65 or $1.25/g
Arctic Silver 5: 12g @ $18 or $1.50/g
There are clear shenanigans in play, that or (more likely) methodological errors.
The type of paste you use has very little difference. Let's not forget this comparison which includes toothpaste.
I feel like there is an Arrested Development joke somewhere in here.
Lindsay:
There's a cream with real diamonds in it. I can actually smear diamonds on my face! And it's only $400 a tub! That's, like, what, like, a million diamonds for $400? A million bleeping diamonds!
While we're on this topic, why don't razor blade manufacturers use diamond dust/shards for the tips of the shaver blades?
I'm sure some will be thinking 'conspiracy' so joe punter has to buy more, but I'm hoping there's a more rational explanation.
Why OpalCalc is the best Windows calc
I'm sorry, but the gains he is talking about are simply unrealistic. Lets do a little math shall we?
.75 inches (it will be larger), his CPU is disapating 100 Watts(probably higher than it is), and we take the advertised number for AS5 at 8 W/m*K, and you end up with a thermal circuit that takes 13 degrees to cross.
If we take a rather thick installation of AS5 at 0.015 inches and assume the contact area is a square with sides of
He claims to have a new thermal compound which reduces the temperature by 14 degrees. Now lets take a look at some more realistic numbers... 1 sq in area, 75 watts, 0.010in thick paste, same 8W/m*K and you get a tempeture delta of 4 degrees to cross.
Furthermore, when we start looking at websites that have done reviews of thermal pastes like [url=http://hardwarelogic.com/news/137/ARTICLE/2752/3/2008-03-03.html]IC Diamond 7 Carat[/url] and they show a range of 1-2 degrees difference between AS5 and the paste it makes it hard to belive.
For a little more background, perhaps we should consider what is going on here. We have some material that is being used for thermal conduction, silver or diamonds, and to that we are have a material it is being suspended into. Thermal conductivity of silver is over 400 W/m*K and artic silver which is made from pure silver plus the suspension yields a conductivity of 8 w/m*K. The idea that exchanging that for something with a thermal conductivity of somewhere between 900 and 2000W/m*K is going to yield a paste with orders of magnitude better thermal conductivity.
So based on that, I'd like to call shens. If he made a mistake with his numbers or he faked them I don't know, all I know is the numbers he is reporting are outside the realm of reality.
Arctic Silver with a fresh application: System Max load 57c
Diamond Grease with a fresh application: System Max load 38c
The author notes that the fancy diamond paste results in LESS heat at the heat sink.
Conclusion: The home-made paste is more of an insulator than the commercial stuff, as the same amount of heat is being generated, but that heat just isn't making it to the heat sink!
the objective of a thermal grease is to fill up the various gaps (microscopical) between cooler and cpu plate; dust, even the finest one, is constituted mainly of.. sharp points. Billions of small points touching two smoot surfaces.
this must be just a joke.
I was going to just mod you up, but I figured your post would soon be at 5 anyway. (if not, mod parent up!)
Yes, you're absolutely right. Lots of bad physics and a completely incorrect conclusion in that article. If there is an improvement in the thermal interface, the heatsink should get HOTTER. All of the heat goes somewhere. That's the first law of thermodynamics... conservation of energy.
You can break it down to a thermal resistance model which is functionally equivalent to an electrical resistance model. Difference in temperature is the "voltage" that drives the flow of heat (current). The heat (current) is a fixed value. The resistances are the various materials the heat has to flow through to get to the air. It can take many parallel and series paths to get out and you can build a resistance network to calculate heat flow through each "path". One way is through the IC leads, into the board, and eventually to the air... or the most direct path is through the heat spreader, through the thermal interface material, into the heatsink and eventually to the air. If the heatsink reads a lower temperature, that means less heat is flowing through the heatsink and that heat is taking a more difficult path to get out and thus the IC junction temperature is HIGHER. That's BAD.
So yes, the math doesn't add up. The thermal interface between the IC and the heatsink should be spread so thin that the thermal resistance across it is almost negligible. Thermal grease is only there generally because it's it's much higher conductivity than air and if you don't have it, all you have is air in all of the little surface imperfections of the heatsink and heat slug on the IC. Adding solid particles to the grease only serves to hold the heatsink and IC heat slug further apart, so even if it's great conductivity, it's generally the wrong thing to do.
-S
--- What parts of "shall make no law", "shall not be infringed", and "shall not be violated" don't you understand?
Check here.
What's really sad, is that this is the only reason people go to a Dustin Diamond show.
The hardest? Nanorods? I'm sure there's a penis-joke in there somewhere.
GAAH! MY PRINTER IS ON FIRE!!! PUT IT OUT! PUT IT OUT!
Only one problem I see is you still have to deal with the thermal conductivity of the heatsink.
I still don't understand how a diamond powder has such a high level.
Does this have to do with the molecule structure or something?
While diamond is a great choice for thermal grease, the heat sink used still has a copper block. with a thermal conductivity less than that of silver, it would seem that you would see little to no gain in heat conductivity as the copper is already 'saturated'.
>DIY CPU Thermal Grease, Using Diamond Dust
I can see it now:
"Mum, can I borrow your engagement ring?"
Open Source Drum Kit, LPLC deve board - mjhdesigns.com
I hate to feed the trolls, but I see no mention of "hardware hack".
If you're speaking of the tags, well, they aren't the summary are they?
Maybe you should learn the definition of "tag".
There is already grit mixed in with Arctic Silver. Large enough to feel.
What I'm wondering about is whether this stuff is any different than using readily available diamond compound from MSC for 15 bucks a 5gm syringe.
http://www1.mscdirect.com/CGI/GSDRVSM?PACACHE=000000104110495
--
BMO
Innovation Cooling already make a thermal compound product based on diamond dust.
Who is the jerk off now? Huh? Answer me that. Please answer. I am awaiting your response.
would be using isotopically pure carbon-12 diamond dust.
It works better without the C-13 mixed in.
I bought a commercial diamond-based thermal paste when I built my quadcore machine a couple years ago. I tested it by pushing a high-res scene through unbiased physical light renderer for a week. It never went above 38C running at 100% on all four cores. My older dualcore machine idles at 55C with ArcticSilver 5.
Of course, the cooling of the quadcore certainly isn't hurt by the ridiculously large all-copper CPU cooler I put on it.
I was telling my little sister (A silversmith and jeweler) about it, and she said "Well duh, diamonds conduct heat like you wouldn't believe. It's basically the only stone you don't need to heatsink while soldering a ring because it doesn't hold heat long enough to damage it."
Friend: "The NIC is misconfigured..." Me: "No prob, I'll just telnet in and fix it." *Silence*
Um... it is late, I'm drunk, and this still doesn't make sense. It looks like the temperature measurement was directed at the heat sink (last picture in the article). If the thermal paste was a perfect insulator, the heat sink would be room temperature. If it (and the heat sink) was a perfect conductor, the heat sink would be the same temperature as the chip. The whole thermodynamics thing prevents the heat sink from being hotter than the heat source. So, in this case, lower temp = worse performance.
Either I'm too drunk to think straight, or everybody who agrees with the conclusion is an idiot. Maybe both.
In other DIY news, man successfully levels wobbly table using platinum ingots.
I've used a silicone based hand cream in the past, I know about a particular computer that works till this day without overheating (3 years...)
Which basically means, unless you're a professional overclocker, its a waste of money.
So you're saying the only thing that would give this kind of improvement would be to replace the copper/aluminum heat sink with one of similar size made out of pure diamond? Anyone making those yet?
What ever happened to Sudo cream (Zinc Cream for Nappy Rash) or semi solid zinc sunblock sitcks ??? Cant say I have used it on any CPU's for the last year or 2 ,, but always worked fine in the past ..
Let's assume that 95 watts have to be removed. Given a constant airflow at constant temperature, the heatsink fins will have to be at a particular temperature (Newton's law of cooling) to dissipate 95 watts to air, and this will be a constant temperature under the stated conditions - call it T1.
Now consider the interface between the heatsink and the heat source. The heatsink base is at a constant temperature T2, so that T2-T1 is the temperature difference caused by the heat removal at the fin surfaces. The die surface is at a higher temperature T3. The temperature difference T3-T2 is needed to drive the heat transfer across the interface. The closer to a perfect conductor if heat is the interface, the smaller is T3-T2.
Thus no matter what material is used in the interface, the base of the heatsink should be at the same temperature.
So, summarising:
T1 is determined by the heat flow through the heatsink, air flow and air temperature, and a rough constant for the heatsink design.
T2 is determined by the heat flow through the heatsink, and T1.
T3 is determined by the heat output of the die, the thermal performance of the interface, and T2.
No matter how good the thermal paste is, the die temperature can never go below T2. Therefore, the heatsink is still the limiting factor.
From scarped cliff or quarried stone she cries "A thousand types are gone, I care for nothing, no not one."
Looking at the values given on Arctic Silver and the linked site, Arctic Silver 5 should have twice the thermal conductivity than the diamond-based paste (4.5 W/(m*K) for diamond paste vs. >8.9 W/(m*K) for AS 5). This might be explained by the two orders of magnitude difference in the average particle size, the given values are 0.49 microns for AS 5, and 40 microns for the diamond paste (I'm interpreting à on the latter site to mean a micron).
U+F8FF
IF he's speaking of the tags?
Um, maybe you should check the "subject" line on his post:
Yeah, I think he's pretty obviously speaking of the tags.
Somehow that makes me want it more.
Instead of using diamond dust, he should use Dustin Diamond. He is so cool that I'm sure just the powder from him would warrant several degrees of cooler CPU temps.
Does anyone know where to purchase the silicone grease needed? I found the diamond powder needed. However, I can't seem to find a silicone grease with the specified chemicals. polydimethylsiloxane and polytetrafluoroethylene. I'll post my results if anyone can help me find the right grease. Ahem, and I won't include the temp of the heatsink in my results, just core temp.
dansdata http://www.dansdata.com/goop.htm water does well as well. anyways, it seems dubious to say the least.
The diamond dust is used to "lap" the contact surfaces of the CPU and cooler/heatsing to achieve better contact. The best heat transfer would perfect metal-to-metal contact. A good machinist ar lens maker could achieve this by lapping the surfaces on an optical flat...90% contact would be good enough. The lens maker would use diamond dust...the machinist would use less expensive lapping conpounds. I use extra fine water sandpaper on a formica table top.