Liquid Metal Thermal Paste

[Cooper]

Hey guys ! Can I use this compound on CPU with HS on ? Or it`ll get messed up ? I have an Opty 146 here.

[Angry_Steel]

Coolermaster NanoFusion is better, but not as sticky, takes some effort to apply it. But the good thing is it won't short close traces on a motherboard like AS5 would. I don't think it's out yet... or is it?

So if you dont even know if its out yet, how can you say its better? Ive seen like one "review" for the stuff that was pretty much good for laughs is all.

As far as ruining something with AS5, I really dont see how unless you put the stuff on like peanut butter, or start dropping it all over the place.

[Ray_GTI-R]

After 5 months I've finally gotten around to trying Liquid Metal (LM) on a CPU.

The bottom line is that comparing LM to my "reference" paste - ShinEtsu Microsi (SM) there's no real difference in temperatures. I'm using just the BIOS, for "pure" temperature readings. Rather than stress the CPU to get variations in temperatures I took a different route ... to vary the speed of the "case" fans. So 100% fans give a blast of cool air while a minimal fan speed just keeps the rest of the components cool. This results in a spread of 5C, purely for comparison, to mimic a hotter CPU.

Initally in all the tests SM was 2 degrees C cooler than LM. Not satisfied, I took everything apart and checked. I'd only applied LM to the cooler, the Intel IHS merely had tiny spots of LM. So, I cleaned the IHS (with Brasso & isopropyl alcohol) and applied LM to the IHS as I should have done first time. With both the heatsink and the IHS coated, this time LM performed better but was still very disappointing ... LM was still 1 degree C warmer than SM.

I also ran some windows benchmarks and they simply confirmed what the BIOS had already told me.

Conclusion:-
Maybe this is an extreme compound for extreme situations but my "average" setup just doesn't need it. Given that LM is electrically very conductive (= dangerous if not used carefully), hard to apply/remove properly, must never be used on aluminium & shows no improvement over SM ... I'm sticking with SM. I still have LM on my test rig and have no problem with that but next time I put it together I'll use SM.

HTH, Ray

Equipment used;-
Modified HSPC Tech Station V2, large (twin 120mm fans)
Intel 2.4C standard
Lapped Intel Extreme Edition HSF
Asrock P4COMBO
(other stuff which isn't relevant)

----------------------------------------------------

For the record, below are some snippets I found very extremely when deciding what to do ... my comments are in square [ ... ] brackets:-

Geeky 102-06-2004, 03:30 AM
Well, here's my take on the whole thermal compound debate.

First, most of the reviews that I've seen of various thermal compounds are totally worthless. They use CPUs as heat sources, which immediately invalidates their results. It is impossible to get a consistent heat output of known value from a CPU, at least with the tools available to virtually all enthusiasts. Intel or AMD can do it in their labs, but they have tools that we do not.

So, the only valid reviews of heatsinks, thermal compounds, etc., are those that use a fixed heat load of a known value. The most common way of doing this is to use a wirewound resistor that is attached to a heatspreader that simulates a CPU die.

Dan's Data (www.dansdata.com) and Frostytech (www.frostytech.com) are the only two places I know of off the top of my head that use this method. The only one of the two that has done testing on thermal compounds, to the best of my knowledge, is Dan's Data.

Dan of Dan's Data published "http://www.dansdata.com/goop.htm" an article on March 13, 2002, on this very topic.

His results are consistent with what I would expect, and what I've seen in my own personal experiences.

Arctic Silver in particular, and various other high performance compounds, may in fact outperform generic silicone-based thermal grease, but the difference is negligible at best.

I've used Arctic Silver's products for years, and I still do, but I do not think that their thermal compounds are the be-all, end-all solution to heat problems, which people (not here, mind you) sometimes seem to think.

The net effect of switching from generic white silicon-based grease to Arctic Silver might be 1-3*C, if you have a decently powerful HSF. (I would expect a larger decrease from a less powerful HSF, though) The reason why I use Arctic Silver's Ceramique in preference to "normal" thermal compound is that it's not much more expensive ($6 for 22g), and the performance may in fact be a bit better. But the price premium for AS5 is just too high to justify the negligible (if any) decrease in temperature, IMO.

--------------------------------------------------------------------
Nevin House 02-06-2004, 08:24 PM
Arctic Silver, Inc.

Geeky1,

The drawbacks to Dan's test are that his synthetic die had a large contact area of approximately 1 square inch (625 sq. mm), the heat was spread evenly across the entire contact area and relatively low power (50 watts) was used for the test.

An actual core on a modern CPU is between about 90 sq. mm and 190 sq. mm. But unlike a synthetic test unit, the heat is not spread evenly across the core. It has been documented that modern CPUs dissipate about 90% of their thermal energy through less than 40% of their top surface area. The cache area for example, dissipates very little thermal energy yet takes up a significant amount of space. The P4's heat spreader and Athlon64's CPU cap actually do very little heat spreading; they just protect the core, keep the heatsink flush and allow higher contact pressure.

The temperature gradients that exist across an AMD or Intel CPU core are well documented and understood by the people developing state of the art cooling solutions. http://www.eetimes.com/semi/news/OEG20031006S0028

If other factors remain constant, every time you double the contact area, you cut the thermal resistance of the thermal joint in half. Conversely, if you cut the contact area in half, you double the resistance of the thermal joint.

So if two different compounds are tested on a large contact area like Dan used and the joint has thermal resistances of 0.05C/W and 0.07C/W (0.02C/W difference) depending on which compound is used, then reducing the contact area in half will increase the thermal resistances to 0.10C/W and 0.14C/W (0.04C/W difference). Cutting the contact area in half again will result in 0.20C/W and 0.28C/W (0.08C/W difference).

Ultimately, the difference that seems so insignificant at the large contact area becomes very significant at real-world contact areas. Dan uses a contact area almost 5X larger than the entire area of the average CPU core and about 12X larger than the contact area of a CPU that actually transfers the majority of the heat. There is no dispute about the accuracy of his measurements, it is just important that people to understand how his data needs to be properly scaled to be relevant to thermal compound performance on an actual CPU.

This is basic science and math and we can use math to scale Dan's test results to where they are relevant to actual CPUs.

THE FOLLOWING ARE BASED ON DAN'S ACTUAL MEASUREMENTS.

Dan measures...
50 watts through 625 sq. mm 0.02C/W Difference
Temperature difference = 1C (0.02 x 50 = 1)

Lets cut the contact area in half .

50 watts through 312 sq. mm 0.04C/W Difference
Temperature difference = 2C (0.04 x 50 = 2)

Lets cut the contact area in half again.

50 watts through 156 sq. mm 0.08C/W Difference
Temperature difference = 4C (0.08 x 50 = 4)

(This is getting close to the actual CPU contact area, but still does not account for some parts of the CPU core being hotter than other parts which would increase the difference between the compounds.)

Now you also have to adjust the differences for the actual power output of modern CPUs which is in the 70 to 105 watt range, not 50 watts.

Adjust for power and you end up at 5.6C and 8C differences at a realistic thermal transfer area of 156 sq. mm and CPU power dissipations of 70 watts and 100 watts respectively.

So ultimately Dan's measurements are accurate, it is just the incorrect interpretation of them that has made his review a favorite of people who want to believe that there is no difference in thermal interface materials.

-------------------------------------------------------------------

Anon undated
Artic Silver 5 will [eventually???] dry out, leaving the cpu unprotected. Ceramique and GC type 44 (Spectra Cool) do not "dust away" or quit working. Spiral cut brass [assumes no lapping, no ambient "working" of the paste into the heatsink/IHS] is [best] filled by:
Wax
Platinum Flux
Silicone
Zalman's paste will do it
Heated Artic Ceramique (apply hot!-mount hot!-big mess!)
GC type 44 (Spectra Cool)

[I actually like the idea of pre-heating paste on the heatsink ... you see the result and if you don't like it you can reapply.]

-------------------------------------------------------------------

Anon undated
Chip manufacturers [specify] phase change pads - they don't dry out or creep as some types of paste will. Silicone based paste does have a tendency to creep and many silver based pastes can dry out over time. There are only three types of paste to my knowledge that have proven to stand the test of time, Arctic Silver 3, GC Electronics Type 44 and Shin Etsu MicroSi.

Of the three mentioned, AMD gives an official ok to Shin Etsu for 'testing' purposes and Type 44 is an industry standard heatsink compound that is well proven in many areas of use. AS3 has probably been the most used paste other than the generic silicone and I have yet to hear of any drying or creeping problems.

--------------------------------------------------------------------

Wikipedia current
Type 44 is Ester based rather than Silicone based.
[Ester: A general term for any organic molecule produced by combining an acid with an alcohol.]

--------------------------------------------------------------------

[Finally, of the more unusual experiments, I've heard that hardcore overclockers have used thin oil like the stuff used for sewing machines. This needs constant re-applying.]

[Ray_GTI-R]

Oh dear!

Oh dear, dear, dear!

Went to swap-out the motherboard on my test rig. Heatsink would NOT budge. Tried heating, tried freezing. No joy. Tried a TINY bit extra force on the frozen heatsink and AAAAAAAAAAAAAAAAARRRRRRRRRRRRRGH! The CPU came out of the socket. Frantically checked for broken/bent pins ... all OK. Phew!

Recap & advice.
a) Liquid Metal gave slightly sub-standard temps but mechanically OK when applied to ONLY ONE SURFACE (heatsink or IHS).
b) Liquid Metal gave slightly sub-standard temps but mechanically BAD when applied to BOTH SURFACES (heatsink and IHS).

More, miscellaneous info:-

Once removed from the motherboard socket it took a LOT of force to part the CPU from the heatsink. Checking the IHS & copper core I found that the Liquid Metal had taken on the consistency of Super Glue i.e., it wasn't liquid anymore and the two surfaces had hard patches of Liquid metal. I tried Brasso but had to re-lap the heatsink - also there seemed to have been a small reaction with the copper and the once-mirror surface looked - um - "wavey". Cleaning the IHS was a nightmare, as you can imagine. All in, getting rid of the Liquid Metal from the heatsink and IHS took 10 hours with every technique, material and chemical known to man. That includes a little bit of lying down to stop getting angry! And recalling how conductive Liquid Metal is, I was mindfull to be extremely careful about residues at every stage of the cleaning process. I'm not entirely convinced that I've removed all traces of this stuff from the IHS but it is now very, very shiney.

I have no clue why Liquid Metal is OK when applied to just one surface yet works so differently when applied to both mating surfaces. It's almost like it's a different compound

Anyone?
BTW I have photographs ...

Cheers, Ray
Always Look On The Bright Side Of Life Part II - I now have a lapped IHS

[Radical_53]

Well, I never had any problems with that stuff. Even when applying it to both surfaces (I do so if I use a block that hase been mounted before as it has a scratched surface).
Temps were around 3-4°C lower compared to some normal stuff like AS5, so I think performance changes drastically if you put the stuff on some really hot chips.
Strangely, I got the best difference when using it on my laptop's Centrino (1.5GHz Dothan running at 0.908V max).I gained 10°C there... compared to Ceramique.

[berk]

I think ive asked this question before,but have you tried putting a little bit of that LM into some AS5/ceramique?

[Ray_GTI-R]

I never had any problems with that stuff.

Yes - I'd not had any problems with it until now. Like I said, I took photo's and if anyone wants a look, hop over to http://groups.yahoo.com/group/liquid-Metal and click on "Photos" in the left-hand menu. Membership approval will be necessary in order to stop pranksters & spammers!

Cheers, Ray

Update:- no one hopped so I've removed the Yahoo group.

[Radical_53]

Well, I've been using this stuff for like half a year at least. As I said, no problems With no chip and no block I put it on.

[Ray_GTI-R]

Hey! I got solid material here. It glued my heatsink to my CPU and I proved it

Plus, I know it reacted with my copper core - I had to re-lap it back to flat.

OK, maybe I did something stoopid but what ??? I did nothing radical. No sub-zero superconducting / bong / magnetic / peltier stuff. Just a stock CPU with a stock, lapped HSF. I swear!

I suspect Liquid Metal has reacted very badly with something that I used in the initial lapping/cleaning process. I used Brasso for the final polish, lighter fluid as the initial cleaner and finally, as ever, isopropyl alcohol to remove any residues. I know Liquid Metal hates cleaning fluids like petrol & acetone because it just sits there, does nothing and goes all moody. BUT Brasso shifts it really quickly although http://www.public-health.uiowa.edu/b...%20solvent.pdf says that the main ingredient (Stoddart Solvent) is a "... general cleaner and degreaser."

TIA, Ray
FWIW Brasso is said to contain Sodium Hydroxide as well as Petroleum Distillate. It is supplied to the UK Ministery of Defence ("won two world wars") & The Queen. Does that make it OK for Liquid Metal Not much more info here ... http://www.setonresourcecenter.com/m...4/wcd0045b.htm & here http://pages.slc.edu/~aschultz/chemi...H%20BRASSO.PDF

[Cooper]

I know it reacted with my copper core

Well never experienced something like that with 3 mounts. Maybe these cleaning stuff surely left some and made such reactions

[Radical_53]

@ray: Well, as Cooper said I think you did something wrong there... I've been using this stuff for quite a while on a couple of chips, numerous mounts without any problems.

[Ray_GTI-R]

OK, thanks peeps.

I'll get back to the supplier and see if I can track down what I did wrong. If I get any feedback I'll update this thread.

Cheers, Ray

[Ray_GTI-R]

1st try at researching this problem ... found this at http://www.coollaboratory.com/manuals.shtml ;-

"Im Zeitraum von etwa 48 Stunden nach dem Auftrag (abhängig von der Dicke des Auftrages) verfestigt sich das aufgetragene Coollaboratory Liquid Pro".

Which, via Babel, translates into ...

"In the period of approximately 48 hours after the [?] (dependent on the thickness of the [?]) the laid on Coollaboratory liquid pro solidifies itself."

Hmmm ... my 1st attempt with this stuff I KNOW it was on for less than 48 hours and it was still liquid. My 2nd attempt I KNOW was on for more than 48 hours - 2 months more like (check my posts ... 03-16-2006 - 04-13-2006).

I just disposed of the remaining material, responsibly.

HTH, Ray

[Radical_53]

Why didn't you use this one? http://www.coollaboratory.com/Anleitung_ENG_NEU.pdf

The manual there only says something about what the stuff will do one a heatsink. There it gets solid, and there you'll have to use a metal polish to get it off.
But it doesn't bond everything together...

[Fairydust]

Coollaboratory announced a Metal Thermal Interface Pad today, available end of October. Seems to be a different alloy which melts above ambient and can also be use with aluminum. Will order a sample and test it, once it becomes available.

[Radical_53]

Yes, sounds interesting. Just like the "Ultra" version they announced.

[Bad213Boy]

I experienced a little setback, involving a cat, pee and my testsystem. Obviously the Gallium had a aggravating effect on the poor little kitten and after first scratching the rig, decided to mark it's new hunting ground. I was only taking care of the cat during the summer vacation of a friend, she is gone now, but left me with lots of scratched furniture and bad smelling carpet. Guess that's why I consider myself a dog person.

rofl, lololol

[Ray_GTI-R]

With the greatest respect ... I tried that NEW, SHOCK, HORROR page via Altavista's Babel translation - and it failed 3 times with an error decoding the text into English. Anyway, with my limited German - that's simply ONE review. Ask users what they think when the product solidifies, causes the CPU to be inadvertently ripped out of the motherboard and needs to be scraped off - Oh that's me. And I said that already.

Anyway - for the record ... I will never use any product from Coolabratory (or whatever they call themselves) ever again. Consider, if you will:- why produce a new heatpad product when you've already got the greatest thing ever invented? Hmmmm ..... I wonder if this thread has been hijacked by a couple of salesmen?

[Radical_53]

The pad seems to be for people that might fear to spill something of the stuff. The pad should be nearly the same stuff, just solid on normal temperatures.

Btw., how did you get the stuff to be solid again? Tried it on a chiller or something like that?