What was the experience for de-lidded IB or Haswell using "metal pads?"

[BonzaiDuck]

I have an idea about a method for de-lidded enhancement. It may be crazy, maybe not. First, though, I gotta question . . .

Too lazy to search back through IDontCare's ground-breaking thread, I can't remember if anyone tried Indigo Xtreme or a similar Coollaboratory metal pad, instead of the Liquid Ultra.

Anyone have firsthand experiences to relate about the "metal pad" TIM for de-lidding? Observations? Thoughts?

 

[thilanliyan]

Isn't Liquid Ultra pretty much the same thing?

 

[BonzaiDuck]

Isn't Liquid Ultra pretty much the same thing?

Yes it is. An Indium-based TIM, but the liquid is liquid, and the pads are solid when you install them. I think they could be cut to shape of the processor.

This is central to my idea. I had a much longer potential thread-OP to post, saved last night in a Notepad file.

The underside of the IHS is hollow except for the processor where it meets the IHS. Some heat would be dissipated into the square PCB of the CPU. If the vertical edges of the processor are marginal additional area from which heat could be conducted, they aren't used.

The point of it: The second-best TIM for its heat-conductivity or low resistance and total lack of electrical conductivity is diamond paste. There are well-tutored suspicions that pressing the nano-diamond particles onto the top surface of the silicon die -- damages a thin film, and allows for copper migration from the IHS to the processor -- causing it to short out, degrade or die. This apparently happened to IDontCare's processor around the time he was testing IC Diamond.

But suppose you used a custom-cut piece of metal pad to mate the die to the IHS surface, and then packed the remaining surface of the IHS with diamond paste? It could get messy. The idea -- the diamond paste would further conduct heat from the processor and dissipate it to the IHS. It would also serve as a mild adhesive to keep the IHS in place while you install it.

But the trick would be to mostly avoid getting any diamond paste between the processor and the IHS, so that the metal pad provides all the contact.

Somehow, it seems that Liquid Ultra or something like it would be a mess when this "hybrid TIM" idea would already be messy.

 

[Xpage]

Very interesting, however I have used IC Diamond before and it tends to be very thick, even boiling it for 10 minutes then applying it was a bit rough to get a thin film of it over a processor.

If you want to fill the IHS, why not, melt the Coollaboratory metal pad, and put it in the IHS and put that over the die. Basically, fill the gap between die and surroundings to IHS with the Coollaboratory metal pad. This kid of makes me want to do it now...

 

[Tristor]

If you're custom cutting the metal pad to fit the die, why would you need to use IC diamond to fill in the remaining gaps? You could use almost any non-conductive paste, for instance Noctua NT-H1 since you aren't relying on it for primary thermal transfer and it's not as high of a risk of possibly causing damage (especially when you consider Haswell with the VRMs or whatever on the PCB)?

Although I'd argue that you wouldn't see any significant difference in temperatures filling in the gap vs not, but that the ease of installation would be a valid argument for using the metal pad anyway if you can get it cut to fit. Personally the Liquid Ultra went well both times I used it but that was only because I was super patient and careful. Despite that, I still spent quite a bit of post application time doing clean-up and inspection before feeling safe using the CPU. It's just kind of scary. I kind of wish I'd bought the metal pad and tried it, honestly.

 

[Yuriman]

Filling in the hollow space will give almost no net result and might even hinder you if get some between the die and your metal pad. When reinstalling the IHS after delidding, the IHS slides when you clamp the retention bracket down so this seems fairly likely to happen.

I would be concerned about using any kind of metal TIM on Haswell because it has surface-mounted components which could be shorted out. Consider that when using a liquid paste, the excess generally runs down the sides of the die anyway.

 

[DigDog]

oh boy ... i'm using de' liquid pro nau, i'm screwed.

OR, one could apply it in a really thin layer, as showed on the instructional video.

Haswell's problem was the HORRIBLE tim, once the tim is replaced with something non-horrible, you really don't need to overdo it. Just put some Coolab pro, some noctua, or even some crap artic silver and be done with it.

Ofc this is invalid if you are trying to break the OC limit on LN2, in which case, disregard that.

 

[john3850]

Adding any extra plate between the IHS reads like a bad idea do to the expansion and contraction from the heat.
If I delided I would only use the IHS if I were using a heavy large HSF.
With a WB I would do a direct to die and use 4 tension springs.

 

[crashtech]

The only way to achieve this reliably would be to drill two holes opposite each other in the IHS and inject TIM once clamping force is achieved between the die and IHS. This would be the only way to minimize impingement of the secondary TIM into the primary indium TIM.

 

[BonzaiDuck]

The only way to achieve this reliably would be to drill two holes opposite each other in the IHS and inject TIM once clamping force is achieved between the die and IHS. This would be the only way to minimize impingement of the secondary TIM into the primary indium TIM.

YUP!! I also had thought of that: drill two holes and "inject." And just because the whole idea is "messy."

Adding any extra plate between the IHS reads like a bad idea do to the expansion and contraction from the heat.
If I delided I would only use the IHS if I were using a heavy large HSF.
With a WB I would do a direct to die and use 4 tension springs.

I think you misunderstood. I wouldn't be adding anything between IHS and die except the "metal-pad" or Indigo Xtreme -- assuming I didn't try "direct-to-die" with, say, a Corsair H100i. The idea instead was to fill the remaining space under the IHS.

Filling in the hollow space will give almost no net result and might even hinder you if get some between the die and your metal pad. When reinstalling the IHS after delidding, the IHS slides when you clamp the retention bracket down so this seems fairly likely to happen.

I would be concerned about using any kind of metal TIM on Haswell because it has surface-mounted components which could be shorted out. Consider that when using a liquid paste, the excess generally runs down the sides of the die anyway.

That's really the crux of it -- little or no net result. The only thing that might happen get some ICD between die and metal pad: the metal pad would absorb the nano-diamond particles -- it partially melts, I think. This was a problem with the metal-pad: just getting the HSF removed from the IHS had been a problem.

But packing the IHS interior with diamond particles would have no electrical drawbacks: it is entirely non-conductive. If the abrasive effects posed a risk in using ICD over CLU, there would likely be none in this secondary application.

I think the most likely speculation over this was Yuriman's -- no significant effect. The additional ICD would need to be under pressure to get any additional heat transfer.

OK -- then I had another idea. If you need an IHS at all -- probably with a HSF-heatpipes with the IHS-CPU under pressure -- you get yourself a silver dollar, mold it into an identical IHS and replace the copper one.

I was also thinking about replacing my HSF->IHS TIM (ICD) with Liquid Pro or INdigo Xtreme -- just for my i7-2600K. The most optimistic expectation would by a maximum 4C drop in temperatures -- but I think they would be more likely 2C+ -- per comparison reviews I'd read for ICD and Indigo.

All of these things -- beginning with the choice of TIM for HSF->IHS -- are marginal solutions. It now looks more and more like an AiO or CLC water loop is also marginal, from the comparison reviews.

The best you can do with Ivy Bridge is de-lid and replace with CLU, Indigo or some similar Indium-based product -- with some risk. If you do it for Haswell, there might be more risk.

I've been trying to "certify" my i7-2600K for 4.7Ghz. Ran several tests at different voltages using LinX at 20 iterations each. Thought I'd found the "sweet spot." Then ran LinX set for 50 iterations -- Affinity under hyperthreading set to cores 0,2,4 and 6. It got through 23 just fine before I fell asleep, and I woke to see the BSOD with the x101 STOP error. It appears the best I can do is bang up the Xtra Voltage for Turbo setting by 24mV above the 4.6 Ghz setting.

With the new processors, the most effective solutions would be chilled water or phase change. Who's going to do that? If only the commercial phase-change device were smaller than a shoe box.

The thing is, the ideal device would only need to keep the processor under load between room-ambient and 5C above room-ambient for regular use under a really good overclock setting.

 

[BonzaiDuck]

Another thought. Someone told me that you could use Ethanol instead of water in a liquid cooling loop -- a more effective coolant.

And if it leaked? "Oh my God! My God! It's such a tragedy! People running everywhere! Then entire computer is in flames!"

 

[Lil Frier]

IDK, isn't having any aquatic leak in a PC basically a tragedy anyway? Just do it, and worst-case scenario, you can make Sandy Bridge s'mores.

 

[BonzaiDuck]

IDK, isn't having any aquatic leak in a PC basically a tragedy anyway? Just do it, and worst-case scenario, you can make Sandy Bridge s'mores.

That's pretty good, Frier!

Conceptually speaking, I'm at a crossroads just because the Sandy Bridge is 2.5 years old. My Quicken files and wallet tell me "You're not at a crossroads yet." Then someone posted another thread asking people for their Cinebench 15 scores, and I did that . . . too . . . It was "nice to see" what the Sandy can do compared to a 4770K.

If I wanted to do water-cooling, I wouldn't be satisfied without "chilled" water cooling. So, looking at the last two Intel generations, I'm trying to figure how I might make THAT better. I'd feel compelled to De-Lid. I'd take the risks with Liquid Ultra.

MAY-BEEE . . . I'd dump the heatpipe cooler approach and get an H00i or something similar. That would make "bare-die" application feasible.

 

[Headfoot]

Just put some Coolab pro, some noctua, or even some crap artic silver and be done with it.

Artic Silver 5 isn't crap. http://archive.benchmarkreviews.com...k=view&id=150&Itemid=62&limit=1&limitstart=12

I don't know when it became trendy to act like AS5 was junk but it's really not. It's not the highest end Liquid Metal stuff but it ranks very highly and is inexpensive

 

[BonzaiDuck]

Artic Silver 5 isn't crap. http://archive.benchmarkreviews.com...k=view&id=150&Itemid=62&limit=1&limitstart=12

I don't know when it became trendy to act like AS5 was junk but it's really not. It's not the highest end Liquid Metal stuff but it ranks very highly and is inexpensive

I would agree, but I started with AS5 before I switched to IC Diamond, and the latter was worth the trouble. The Liquid Metal or similar Indium-based solutions are worth about 1C degree improvement over the diamond paste.

 

[crashtech]

The results shown in that link don't match my own personal experience, but then again most of the higher rated materials are all within a degree of each other. It may be that some TIMs work better than others depending on individual methods of application and type of surface to which they are applied.

 

[BonzaiDuck]

The results shown in that link don't match my own personal experience, but then again most of the higher rated materials are all within a degree of each other. It may be that some TIMs work better than others depending on individual methods of application and type of surface to which they is applied.

Well -- with the diamond paste and a heatpipe cooler under pressure -- you can afford to get sloppy. The paste is a bit more expensive than your various greases, but the excess just oozes from the edge of the IHS and hardens into a rubbery bead. Of course, I don't do that on purpose, but for the ICD -- more is better than less to get full coverage on the IHS.

 

[DigDog]

I don't know when it became trendy to act like AS5 was junk but it's really not.

Jump on the bandwagon bro. If it gives one degree of heat over the latest model, it's bad and anyone using it should be derided, even if the latest TIM eats your heatsink, molests your daughter and sells your car down the pound.

 

[BonzaiDuck]

Jump on the bandwagon bro. If it gives one degree of heat over the latest model, it's bad and anyone using it should be derided, even if the latest TIM eats your heatsink, molests your daughter and sells your car down the pound.

Every time this topic comes up, I find myself sighing at folks who advocate grease-based TIMs -- some packaged with the heatsinks. As far as I see it, this all boils down to a simple factor: the thermal conductivity -- inversely, the thermal resistance -- of the material.

Silver particles are better than copper, but they're conductive. Even so, AS5 doesn't offer a lot of hazard. You just don't want pieces of the stuff on your motherboard, or particles from the pieces.

Once I started using diamond paste, I tended to cringe at the thought of using a liquid metal formulation for the very risk that diamond eliminates. But -- carefully applied -- the Indium-based TIMs seem to work well, or one or two C degrees better than diamond. The problem with the Coollaboratory product: if you're not careful, you can have little shiny balls of the stuff rolling around on your motherboard like mercury. You just need to be careful.

 

[tolis626]

Jump on the bandwagon bro. If it gives one degree of heat over the latest model, it's bad and anyone using it should be derided, even if the latest TIM eats your heatsink, molests your daughter and sells your car down the pound.

I don't have a daugher.I also don't have my own car.But damn,if it eats my heatsink,it's going out of the window.

Actually I never had any problems with AS5.In fact I always suggested it.Never tried the Diamond though.May give it a go on my next build.

 

[BonzaiDuck]

I don't have a daugher.I also don't have my own car.But damn,if it eats my heatsink,it's going out of the window.

Actually I never had any problems with AS5.In fact I always suggested it.Never tried the Diamond though.May give it a go on my next build.

Use the traditional rectangular, one-edged razor blade as a spreader. It's about the size of the IHS, so a little care spreading a bead of ICD from the IHS center won't pose any risk of damaging something with the razor-blade.

Small, straight-cut pieces of expired credit cards also work.

 

[DigDog]

My thoughts on TIMs is that too often they are too similar to be considered different - a 30-TIM test where the spread is 4c means both that the installation could have caused the change, and that they are, from a practical standpoint, identical.

some people might consider 2~4c a massive drop, but most end users dont need that thight of an efficiency; of course there are cases where replacing a tim does in fact cause a real change, and they are not "identical", like my haswell chip, where changing from the intel cement to CLP gave me 10~12c difference.

In those cases, and if you are buying a new tube of TIM, sure, go with the better model .. no reason to grab the one which gives you 4c more, when you can have the one with 4c less.

But when i read that, to get those 4c less (or 2c less when you could have a different paste), you are beginning to face problems like

"it might short out your mobo and kill your cpu"
"cooking time of 200h with a special BIOS setting"
"will eat the metal of your IHS and anything else it touches, like Alien"
"costs $50 for 1.5g"

then i start to look back at the charts and find something which will give me reasonable temps for less, and for less effort, and less worries, and so on.
Like AS, which is now considered "crap" by just about everyone, despite it still being a very reasonable performer with easy application and zero issues.

So my advice is, if you are contemplating a TIM mod, and you have serious potential issues .. don't.
Go with a safer TIM, i seriously doubt you will look at your PC and think "oh boy, i sure do miss the extra 3c i would have gotten by risking the use of **INSERT TIM NAME HERE***"

 

[BonzaiDuck]

My thoughts on TIMs is that too often they are too similar to be considered different - a 30-TIM test where the spread is 4c means both that the installation could have caused the change, and that they are, from a practical standpoint, identical.

some people might consider 2~4c a massive drop, but most end users dont need that thight of an efficiency; of course there are cases where replacing a tim does in fact cause a real change, and they are not "identical", like my haswell chip, where changing from the intel cement to CLP gave me 10~12c difference.

In those cases, and if you are buying a new tube of TIM, sure, go with the better model .. no reason to grab the one which gives you 4c more, when you can have the one with 4c less.

But when i read that, to get those 4c less (or 2c less when you could have a different paste), you are beginning to face problems like

"it might short out your mobo and kill your cpu"
"cooking time of 200h with a special BIOS setting"
"will eat the metal of your IHS and anything else it touches, like Alien"
"costs $50 for 1.5g"

then i start to look back at the charts and find something which will give me reasonable temps for less, and for less effort, and less worries, and so on.
Like AS, which is now considered "crap" by just about everyone, despite it still being a very reasonable performer with easy application and zero issues.

So my advice is, if you are contemplating a TIM mod, and you have serious potential issues .. don't.
Go with a safer TIM, i seriously doubt you will look at your PC and think "oh boy, i sure do miss the extra 3c i would have gotten by risking the use of **INSERT TIM NAME HERE***"

Well, the ICD is a degree or so short of the Indium, doesn't eat away the metal, has no cooking time and won't cause a short. You just have to be dexterous enough to spread a pea around the IHS with a razor blade. It might seem awkward at first to some, but for me it's become a 2-minute routine. Well -- with the heatsink-base -- 4 minutes then.

With the same BIOS settings on my SB, the 4C drop seems to give me less variation or smaller range of difference between each iteration's GFLOPS in LinX. Maybe I'll find another couple C here, another there. It adds up. It might add up to a net voltage reduction of some amount. that was shown by IDontCare's de-lidding experiments.

 

[DigDog]

I dont have experience with ICD myself, but .. i read it is evil ? that it will damage your die and allow for migration of copper atoms from the heatspreader into the die, to be exact, massively accelerating the decay or the processor. Or thats what i have been told.
Tbh application has never been a problem for me, even CLP was easy .. i guess years of painting Citadel miniatures (the warhammmer 40k stuff, mmk?) has given me that kind of manual skill.

 

[BonzaiDuck]

I dont have experience with ICD myself, but .. i read it is evil ? that it will damage your die and allow for migration of copper atoms from the heatspreader into the die, to be exact, massively accelerating the decay or the processor. Or thats what i have been told.
Tbh application has never been a problem for me, even CLP was easy .. i guess years of painting Citadel miniatures (the warhammmer 40k stuff, mmk?) has given me that kind of manual skill.

That's exactly what somebody said . . . about what somebody said, per what might have happened to their Ivy Bridge which had been through a dozen mountings and dismountings in various tests. I don't discount it entirely; but nobody confirms it entirely.

There is an occasional poster here at Anand named ICD7, who is likely a tech-rep for Innovative Cooling. Haven't seen him post recently, but there were other myths about ICD which he helped to bury with detailed experimental proof.

Then, the other day, I found another post at another OC site's forum from an IC tech-rep -- it could've been ICD7, but he used a different handle. I think he was discussing application of ICD for bare-die applications, and I think he was declaring it "safe."

People didn't like ICD because it is thick -- needs a razor blade to spread. If you didn't anticipate some damage to the processor per scarring the film on the die and causing copper migration, then you'd spread it on the IHS and not the die if you were planning that approach. If it were to be a bare-die configuration, you'd apply it to the heatsink base or waterblock, and then mount the latter on the die.

I think I drafted my last post because someone was discussing the drawbacks of metallic pastes. Metallic pastes also include Arctic Silver 5, so that's nothing new.

I'm leaning toward a Coollaboratory Liquid Pro or Ultra for my next build, because I'm also leaning toward either a delidding or delidding-plus-bare-die approach. But I have an annual computer budget, and we have too many computers, so I'm taking my g**d**n time to think about it.

Under the budget, I ordered a new set of DDR3 RAM and a Samsung 840 512GB drive today. I almost added the new CoolerMaster Nepton 280 to my cart, but I'd have to mod my case to use it now, or order a new case.

I'm still celebrating for lowering my temperatures 4C degrees on my 2600K while raising the clock to 4.7. The consistent GFLOPS for the clock setting under LinX averaged 131, with a range of 1.2 GFLOPS, and I'd guess that the standard error is a mere fraction. My load voltage seems to be either 1.344V or 1.352V, depending on the test. And this SB core is 30 months old.

So I continue to plan, connive, dream -- and occasionally take advantage of my special Newegg account . . .