Applying Coollaboratory liquid ultra tim

[Laststop311]

When applying this TIM do you paint the tim on the heat spreader only or do you paint it on the heat spreader and the base of the heatsink/waterblock as well?

 

[BonzaiDuck]

I'm embarrassed to say this -- everybody knows I'm OCD about TIM, lapping, ducting -- anything but water-cooling. Truth is -- I haven't tried CLU myself. I bought my processor binned and CLU-relidded. I've got a tube of it. I wonder if there's any sort of shelf-life expectation.

I use micronized diamond -- generally -- and IC Diamond as long as I can just buy it. With the nano-diamond paste, there's no risk posed by conductivity. It's nothing but an insulator. The excess beads up when surplus is squeezed out between the IHS and HSF/waterblock-base.

With something like CLU, I'd follow the directions to the letter. There only has to be a film of the stuff consistently covering the IHS. It wouldn't matter how thin it is, as long as it thoroughly covers the IHS even under pressure. And it is going to make a bond with bare copper, maybe less with the stock nickel-plate on the processor. Since I've been in the habit of lapping all my processor IHS's to bare copper, that becomes more of a drawback for me if I ever want to remove the heatsink.

 

[Laststop311]

It supposed to be ok with nickel plated copper, I really don't plan on having to ever change the cpu. The generational performance increase is barely even existant and by the time the performance increases is worth it i will need a new mobo anyways. So I wouldn't mind if it permanently bonded anyways.

 

[BonzaiDuck]

You just need to be careful because it's conductive. But you just need a few atoms-worth of gallium (or whatever it is -- indium or gallium . . ) between the surfaces. I've begun to toy with the idea of using it on my 2700K system. That system typically idles between 35 and 38C. Under severe load, it will go up to ~75C. But it's only worth about 3 or 4C degrees improvement.

Ordinarily I would go after such a grain of Centigrade rice, because every improvement is additive. IF I covered all the other bases, though, the diamond is fine.

 

[Laststop311]

Well Im not going to go as far as delidding and applying it right on the die. I will just thinly paint the IHS and that + a d15 should be enough to run 5ghz kaby lake comfortably. I don't think ill put so much that it drips off the IHS so nothing should short.

 

[BonzaiDuck]

Well Im not going to go as far as delidding and applying it right on the die. I will just thinly paint the IHS and that + a d15 should be enough to run 5ghz kaby lake comfortably. I don't think ill put so much that it drips off the IHS so nothing should short.

Understood. Silicon Lottery has begun offering delidded 7700K processors -- the delidding with CLU costing $50. I'd have to look more closely at the thermal results for the 7700K right out of the retail box. We had another forum member start a thread recently to describe his 7700K delidding experience. If you can get to 5Ghz with peak package temperature maxing out at 80C -- or less! -- you might be "good to go."

 

[Laststop311]

Yea as long as i can hit 5ghz and max out in the 70's I'm happy with that. Thats crazy 680 dollars for a 5.2ghz 7700k. For 700 I might as well get 6 core 6850k with 40x pci-e. Plus those are soldered on and can probably OC to like 4.8ghz no prob. Serious retards spending that on a 7700k for an extra 100-200 mhz

 

[BonzaiDuck]

Yea as long as i can hit 5ghz and max out in the 70's I'm happy with that. Thats crazy 680 dollars for a 5.2ghz 7700k. For 700 I might as well get 6 core 6850k with 40x pci-e. Plus those are soldered on and can probably OC to like 4.8ghz no prob. Serious retards spending that on a 7700k for an extra 100-200 mhz

It was quite a deliberate choice for me with the i7-6700K. I think I paid $80 over retail for a binned chip at 4.8Ghz, and $50 for the CLU-relid. I don't think of the relidding as part of the CPU purchase as much as it was part of my cooling solution. The $80 was simply insurance that I didn't get dealt a bad hand as luck of the draw.

But -- $700 -- not sure what you mean. Today's price at Silly-Lots for a 7700K binned for 4.9Ghz is about $350, and the additional re-lidding service makes it $400.

It was either that, or spend north of $200 extra on water cooling. I can't be retarded if I can run my system at 4.7 Ghz under LinX stress-testing with a peak package temperature of maybe 75C and IBT testing at 70C.

Look at it another way, as a cooling solution. Cooler plus CLU-relidding totals $120. An EKWB Predator 240 -- when it wasn't out of stock a couple months ago -- was around $200. I come away with a 7C edge over the Predator for $80 less. The $80 I paid for the testing and binning.

I'd also be interested to know why Silly Lots hasn't binned any 7700K's for 5.0Ghz. 4.9 is the best they can offer at the moment. Somebody could send an e-mail, and they'd likely answer that question.

 

[StefanR5R]

For 700 I might as well get 6 core 6850k with 40x pci-e. Plus those are soldered on and can probably OC to like 4.8ghz no prob.

Broadwell-E overclocks badly; it appears to be an architectural handicap. With ambient cooling, i.e. anything apart from chilled water or LN2, do not expect better than 4.3 GHz from Broadwell-E, and even that may prove difficult to get stable.

 

[Laststop311]

its 700 dollars for a 5.2ghz binned 7700k. I think like 680 than+ 50 to delid so 730 total. I would rather buy a 6850k thats soldered on and have a 6 core and 40 pci-e lanes than pay for that one. People gotta be stupid to do that.

It;s like 65% success rate on 5.0 ghz according to silicon lottery's stats. I'd just get a retail unit from microcenter on a deal for like 300-330 and take my chance they are pretty good. Tho the delid cost of 50 dollars isn't bad. They say it improves temps by a solid 10C or more by repasting it with CLU. Maybe getting like a 5.1ghz binned one would make the 5.0ghz run at a lower voltage so delidding wont be needed. Spend the extra 50 on a higher binned cpu and just CLU it to a d15 and be able to turn down the voltage a bit and just go for 5.0ghz. Only like 23% of chips binned to 5.1ghz and only 4% binned to 5.2ghz

 

[StefanR5R]

Interesting. Thanks for the binning statistics.

 

[BonzaiDuck]

its 700 dollars for a 5.2ghz binned 7700k. I think like 680 than+ 50 to delid so 730 total. I would rather buy a 6850k thats soldered on and have a 6 core and 40 pci-e lanes than pay for that one. People gotta be stupid to do that.

It;s like 65% success rate on 5.0 ghz according to silicon lottery's stats. I'd just get a retail unit from microcenter on a deal for like 300-330 and take my chance they are pretty good. Tho the delid cost of 50 dollars isn't bad. They say it improves temps by a solid 10C or more by repasting it with CLU. Maybe getting like a 5.1ghz binned one would make the 5.0ghz run at a lower voltage so delidding wont be needed. Spend the extra 50 on a higher binned cpu and just CLU it to a d15 and be able to turn down the voltage a bit and just go for 5.0ghz. Only like 23% of chips binned to 5.1ghz and only 4% binned to 5.2ghz

Yeah, for that price, it isn't worth it to buy a binned/tested 5.0Ghz chip, and the statistics suggest you'd be better off buying the retail box and either delidding yourself or sending it to Silly Lots for the re-lid/CLU service.

As for the Silly Lots binning, the only thing they guarantee is that the processor will actually run at [whatever Ghz -- 4.9, 5.0] at a specific voltage, with the encouragement that it "could be less."

My 6700K binned for 4.8 goes up the multiplier ladder to 4.7 showing the voltages an engineering study proved you could expect at the various speeds. If I wanted to run it at 4.8, I'd have to set the voltage to around 1.424V.

The delidding and CLU relid for the nominal $50 allows you more flexibility in cooling strategies if you plan to overclock into the higher expectational range.

 

[DrMrLordX]

I have applied CLU many times to both copper and nickel surfaces. DO NOT USE CLU ON ANY HSF/BASE THAT HAS EXPOSED ALUMINUM ELEMENTS. The same technique works in either case:

1). Get a cheap dollar-store paint brush set. Select a single brush and use some sharp scissors to cut the bristles short. As short as you can get, really. You don't want there to be any horizontal play. When you are finished with the brush, throw it away. You won't want to re-use it later.

2). Install the CPU in the socket.

3). Put a tiny - TINY - amount of CLU on the center of the IHS (or die if you're going naked) and start spreading it out with vertical or horizontal brush strokes. Be concise in your movements. Start from the inside and brush out. If you have a tube of CLU that you've already used, you may find residue clinging to the end cap from where the tube has leaked in various places. It's okay (and often wise) to scavenge CLU with the brush from those locations, preferably before there is any CLU on the brush, though doing it with a stained brush is okay too.

4). Repeat your brush strokes until you can no longer coat any more of the IHS/die surface. Place smaller-than-what-you-used-originally amounts of CLU near the edges of the coated area, and use brush strokes along the edges of the coated area to start spreading it. Then progressively move closer to the IHS/die.

If you're having problems reducing the size of the application, one option is to "dip" the bristles of the brush directly into the aperature of the syringe; alternatively, you can repeat scavenging efforts around the cap or around the end of the syringe itself.

5). Repeat 4 until the entire IHS/die is coated.

6). Take a coffee filter, turn it upside-down so that it forms a dome that can stand on its own, and put it over the CPU/socket.

Repeat 3-5 for the base of the HSF/water block. Then mount the HSF/block fairly quickly after removing the coffee filter "dome".

As an aside, you will find that it takes longer to get CLU to spread on a nickel surface, but you can spread it a bit further per unit mass initially applied since it doesn't cling to the surface as readily as it will a copper surface. It loves copper. Oh yes it does . . .

 

[BonzaiDuck]

I have applied CLU many times to both copper and nickel surfaces. DO NOT USE CLU ON ANY HSF/BASE THAT HAS EXPOSED ALUMINUM ELEMENTS. The same technique works in either case:

1). Get a cheap dollar-store paint brush set. Select a single brush and use some sharp scissors to cut the bristles short. As short as you can get, really. You don't want there to be any horizontal play. When you are finished with the brush, throw it away. You won't want to re-use it later.

2). Install the CPU in the socket.

3). Put a tiny - TINY - amount of CLU on the center of the IHS (or die if you're going naked) and start spreading it out with vertical or horizontal brush strokes. Be concise in your movements. Start from the inside and brush out. If you have a tube of CLU that you've already used, you may find residue clinging to the end cap from where the tube has leaked in various places. It's okay (and often wise) to scavenge CLU with the brush from those locations, preferably before there is any CLU on the brush, though doing it with a stained brush is okay too.

4). Repeat your brush strokes until you can no longer coat any more of the IHS/die surface. Place smaller-than-what-you-used-originally amounts of CLU near the edges of the coated area, and use brush strokes along the edges of the coated area to start spreading it. Then progressively move closer to the IHS/die.

If you're having problems reducing the size of the application, one option is to "dip" the bristles of the brush directly into the aperature of the syringe; alternatively, you can repeat scavenging efforts around the cap or around the end of the syringe itself.

5). Repeat 4 until the entire IHS/die is coated.

6). Take a coffee filter, turn it upside-down so that it forms a dome that can stand on its own, and put it over the CPU/socket.

Repeat 3-5 for the base of the HSF/water block. Then mount the HSF/block fairly quickly after removing the coffee filter "dome".

As an aside, you will find that it takes longer to get CLU to spread on a nickel surface, but you can spread it a bit further per unit mass initially applied since it doesn't cling to the surface as readily as it will a copper surface. It loves copper. Oh yes it does . . .

Here's a question . . . Doctor . . . LordX . . . The problem with CLU and possibly even Indigo Xtreme is that the gallium or indium or whatever bonds with the copper, less so with the nickel. for that reason, I've abjured using it, because it means difficulty and greater risk just removing the HSF from the IHS.

So I'd been toying with this idea of using something I bought before I started buying IC Diamond: micronized diamond powder of a certain average diameter. (And just glad that I don't think I ever inhaled any of that stuff, because it CAN become airborne.)

I was thinking if it were possible to get an even distribution of the abrasive particulate in the CLU, that it would weaken any amalgamation or bonding between copper and CLU and copper. All three. That is, if some force were needed to break the bond between HSF and IHS, it would take less to do it.

What do you think about that?

 

[DrMrLordX]

Depends on the particle size. I would think it would increase effective bondline thickness if you mixed in that stuff, though I could be wrong. One of the attractions of using CLU is the incredibly thin bondline thickness possible when using it. Copper -> copper bonding is pretty hard to break once the CLU sinks into both surfaces, but hey, at least you know you have good contact.

If there's a nickel surface involved, it doesn't really stick that hard.

 

[superstition]

A few pointers:

1) Q-tips are dusty. You definitely don't want to use standard Q-tips with liquid metal TIM. The ones that come with Liquid Pro aren't dusty but they can shed big threads.

2) Liquid Pro partially dried out on my EK waterblock before I even had the chance to install it. That's how fast the bonding can occur. When you apply it have everything ready so you can immediately install it. As mentioned, nickle is more resistant.

3) You'll have to lap the parts if you take apart the bond so be ready with 2000 grit or better paper, a piece of glass, and a willingness to lap.

4) Whatever brush you use to apply the stuff shouldn't shed or have any sort of material coating the bristles that will come off. Sometimes brushes are coated/sprayed and need to be cleaned prior to use.

5) Liquid Pro likes to form little beads and they roll around all over.

6) Don't use the syringe anywhere near your motherboard. I've had Liquid Pro shoot 3 feet or more through the air coming out of the syringe with a gentle press. My suggestion is to let it "explode" into a fresh plastic garbage bad so you can use the brush from the material that ends up on that plastic surface. Don't try to have it come out of the syringe onto the chip or you could end up with a big mess.

7) Worrying about thickness of application is less important, in terms of heat bottlenecking, than with polymer TIM. Liquid metal TIM has a vastly better thermal conductivity so it's not as big of an issue if it's thicker. This one reason why applying it to both the spreader and the block isn't a big no-no. Just don't have so much excess that it will drop beads onto your board. Consistent contact is more important than layer thinness with liquid metal TIM.

 

[BonzaiDuck]

Depends on the particle size. I would think it would increase effective bondline thickness if you mixed in that stuff, though I could be wrong. One of the attractions of using CLU is the incredibly thin bondline thickness possible when using it. Copper -> copper bonding is pretty hard to break once the CLU sinks into both surfaces, but hey, at least you know you have good contact.

If there's a nickel surface involved, it doesn't really stick that hard.

Well, I can see issues to surmount. You'd want the particles distributed evenly, so you might incline to mixing it with the CLU in a small dish -- smaller the better. How thick should it be? If you look at IC Diamond the way it's made is something short of freshly-laid sidewalk cement -- it is heavily loaded with diamond. But the purpose of this only would be weakening the amalgamation or bonding.

Superstition doesn't think bondline thickness is an issue. The diamond wouldn't much degrade the TR or inverse conductivity of the CLU, and the particles would be immersed in it. But then you get into possibility of creating some Frankenstein of little balls of the s***. They make another item in pads like Indigo without the need to heat up before it spreads.

On either processor I have, I'd want to think of not having to remove the cooler again, but you never can tell.

 

[DrMrLordX]

1) Q-tips are dusty. You definitely don't want to use standard Q-tips with liquid metal TIM. The ones that come with Liquid Pro aren't dusty but they can shed big threads.

OP was talking about CLU, nor CL Pro. Pro behaves differently. If I recall, the most popular method of spreading it was with the tip of a syringe.

3) You'll have to lap the parts if you take apart the bond so be ready with 2000 grit or better paper, a piece of glass, and a willingness to lap.

Since the main corruptive ingredient in both CLU and CL Pro is gallium, I'm going to have to disagree here. You don't have to lap anything, and 2000 grit silicon carbide sandpaper will barely put a dent in the gallium surface corruption anyway. I remember trying to "lap out" the greyed metal on my nh-d14 with 400-grit sandpaper and I barely put a dent in it. I gave up on that.

Anyway when removing and re-pasting a copper surface that has been corrupted by CLU (or Pro), just scrub it down with coffee filters and 91% or better Isopropyl alcohol. You'll notice the coffee filters blackening as you scrub. Keep it up until a fresh filter + alcohol doesn't turn black very much. Then you're okay.

You might have slightly-degraded performance using the corrupted surface with a non-metal TIM, but probably not, and I have had no issues remounting sinks after several uses of CLU.

4) Whatever brush you use to apply the stuff shouldn't shed or have any sort of material coating the bristles that will come off. Sometimes brushes are coated/sprayed and need to be cleaned prior to use.

The cheap ones have no coatings, at least in my experience.

5) Liquid Pro likes to form little beads and they roll around all over.

6) Don't use the syringe anywhere near your motherboard. I've had Liquid Pro shoot 3 feet or more through the air coming out of the syringe with a gentle press. My suggestion is to let it "explode" into a fresh plastic garbage bad so you can use the brush from the material that ends up on that plastic surface. Don't try to have it come out of the syringe onto the chip or you could end up with a big mess.

Yeah, that's why I don't use Pro . . . CLU behaves very well, overall. Just don't let it get on anything you like. It stains everything. Even vinyl.

7) Worrying about thickness of application is less important, in terms of heat bottlenecking, than with polymer TIM. Liquid metal TIM has a vastly better thermal conductivity so it's not as big of an issue if it's thicker. This one reason why applying it to both the spreader and the block isn't a big no-no. Just don't have so much excess that it will drop beads onto your board. Consistent contact is more important than layer thinness with liquid metal TIM.

Bondline thickness is always important. CLU only has a W/mK of . . . 32.7 I think? Anyway it's much better than standard TIM, but it isn't perfect. Ideally there wouldn't be a TIM there at all, if we could get away with it.

But the purpose of this only would be weakening the amalgamation or bonding

I would think you'd have to salt the TIM pretty thoroughly with powder to significantly interfere with metal bonding. Who knows, it might even improve the performance of the TIM if done properly. I would think you'd want to apply the powder after painting it onto the IHS surface . . . just not sure how you'd get it on there without making a terrible mess.

Might be worth experimenting with it someday though.

 

[BonzaiDuck]

OP was talking about CLU, nor CL Pro. Pro behaves differently. If I recall, the most popular method of spreading it was with the tip of a syringe.



Since the main corruptive ingredient in both CLU and CL Pro is gallium, I'm going to have to disagree here. You don't have to lap anything, and 2000 grit silicon carbide sandpaper will barely put a dent in the gallium surface corruption anyway. I remember trying to "lap out" the greyed metal on my nh-d14 with 400-grit sandpaper and I barely put a dent in it. I gave up on that.

Anyway when removing and re-pasting a copper surface that has been corrupted by CLU (or Pro), just scrub it down with coffee filters and 91% or better Isopropyl alcohol. You'll notice the coffee filters blackening as you scrub. Keep it up until a fresh filter + alcohol doesn't turn black very much. Then you're okay.

You might have slightly-degraded performance using the corrupted surface with a non-metal TIM, but probably not, and I have had no issues remounting sinks after several uses of CLU.



The cheap ones have no coatings, at least in my experience.



Yeah, that's why I don't use Pro . . . CLU behaves very well, overall. Just don't let it get on anything you like. It stains everything. Even vinyl.



Bondline thickness is always important. CLU only has a W/mK of . . . 32.7 I think? Anyway it's much better than standard TIM, but it isn't perfect. Ideally there wouldn't be a TIM there at all, if we could get away with it.



I would think you'd have to salt the TIM pretty thoroughly with powder to significantly interfere with metal bonding. Who knows, it might even improve the performance of the TIM if done properly. I would think you'd want to apply the powder after painting it onto the IHS surface . . . just not sure how you'd get it on there without making a terrible mess.

Might be worth experimenting with it someday though.

that was my feeling -- "someday." I suppose you could prepare a salt-shaker with really fine holes. that's the only way you could do it as opposed to mixing it first, and come to think of it, some irregularity in application might improve the desired effect of weakening the bond. But it couldn't go on in lumps that would degrade the CLU effectiveness or evenness of contact.

 

[DrMrLordX]

Right, no lumps. You'd need a form-fitted sifter to dole it out like confectioner's sugar.

I took a look around and there is .25 micron diamond powder for sale. That isn't going to increase bondline thickness at all (sans clumping), so that would be the product to use.

 

[BonzaiDuck]

Right, no lumps. You'd need a form-fitted sifter to dole it out like confectioner's sugar.

I took a look around and there is .25 micron diamond powder for sale. That isn't going to increase bondline thickness at all (sans clumping), so that would be the product to use.

Basically, I'm pretty sure I ordered <= .25 micron. I might have paid the same for it either way, but I discovered at the time of purchase that my 5 grams (or whatever I bought) was natural diamond, when ICD uses synthetic.

 

[DrMrLordX]

Natural and synthetic will be the same. Actually synthetic might be a tad bit better.

 

[BonzaiDuck]

Natural and synthetic will be the same. Actually synthetic might be a tad bit better.

Not to differ, but I'm not sure it would be easily measurable difference. I was just naïve about the commodity when I ordered it from an obscure (to me) industrial abrasives supply-house online. I did try it -- it works. I just never settled on what run-of-the-mill oil-based TIM to mix it with so it's loaded up like ICD. I would think any silicone-based TIM would do -- white stuff. Somebody on another computer forum site had acquired one of those drugstore plastic lens cases for contacts -- a round one with the cap. He poked a hole in the center of the cap, and fashioned a crank-handled paddle out of a paper-clip, to mix your powder and your TIM-base.

I must still have 4 grams of the stuff, and should probably use it to save on ICD purchases.

For what it's worth, I squandered $90 when I bought it two months before ICD became available on thel market, but after reading Joe Citarella's review and calorimeter tests of diamond at Overclockers.com. I was so interested in the test results, that I didn't see he was a founder of ICD. Apparently the testing and article were accomplished before the product was released.

Stupid? Impulsive? Maybe. But that was 9 years ago, if I remember.

 

[Suspicious-Teach8788]

Any update from OP? How did you end up applying the CLU? With the brush method? I see RockitCool recommending the Q-Tip method.

Also any thoughts on relidding? Do you guys use RTV or Super Glue or whatever?

 

[gradoman]

Any update from OP? How did you end up applying the CLU? With the brush method? I see RockitCool recommending the Q-Tip method.

Also any thoughts on relidding? Do you guys use RTV or Super Glue or whatever?

Delidded, used CLU along with the included brush. To apply, I used a very tiny amount. It's crazy how far it can spread, but it covered the CPU die and then I did the underside of the IHS and relidded with the gel Loctite super glue. It's been about 2 months, I think, and things are fine.

I used this guide: https://www.youtube.com/watch?v=bwg0HRw17lY