First time delidder with multitude of questions

[WithAlligators]

Hey guys,

First time poster...

About to put my 4790K under water in an Ncase M1. Would like the best thermal transfer possible so am going to delid. I've read a ton, to the point of being confused. Everyone says CLU is the best for die to IHS heat transfer, but I have a question about that. It seems like there is a gap between the die and the IHS that is filled with Intel's TIM. Is CLU able to fill that gap being that it is a liquid? Also, I may wish to transport my PC around occasionally. Do I need to worry about vibration shaking the liquid down in the IHS and slopping it around? Should I coat the mini capacitors and transistors and stuff on under the IHS? Has anyone found anything better than nail polish, etc? What's the gray epoxy they use to seal the die?

Now that people have some years on delidded Ivys and such, is there any indication that the gap between the die and the IHS is a purposefully engineered thing? Does it serve any purpose, and if not, why is it there?

I also have a 3770K in a different rig. All the same concerns apply, right? Or is there some real difference?

Cheers

 

[DrMrLordX]

Yes, CLU can fill that gap. You've just got to use an abnormally large amount of it IF you're going to do a relid. Since you're watercooling, you may wish to consider leaving off the lid and doing direct-die cooling.

If you want to ask about transportation issues, you may wish to talk to the folks in Cases & Cooling. There are some watercooling veterans in there that should be able to address any questions you have about moving a watercooled rig. Most of your concerns about TIM movement under the IHS would be alleviated if you simply removed the IHS.

 

[DownTheSky]

In case you want a 2nd opinion. I agree with mrlord, direct-die cooling is the best option, just be careful when fitting the block. Too much pressure may break your die.

 

[WithAlligators]

Sorry, I should have provided more details. I understand bare die cooling is best, especially with a light waterblock, unfortunately I am an Apogee Drive II, so it is quite heavy. I kind of don't want to deal with the hassle of getting the shims right and applying pressure evenly across the CPU PCB and die and such. Maybe in the future, but not now. I wish one could get that MSI die guard protector separately. Problem solved.

Thanks for the suggestion though. I probably will at some point do just that, but not yet.

 

[GreenChile]

Yes, CLU can fill that gap. You've just got to use an abnormally large amount of it IF you're going to do a relid. Since you're watercooling, you may wish to consider leaving off the lid and doing direct-die cooling.

That is not an accurate statement. Once the IHS sealant has been removed, the IHS will rest directly on the die. There is no gap.

You only need a very thin coating of CLU applied to both the die and underside of the IHS to make a very good thermal connection.

 

[DrMrLordX]

Have you actually done a delid/relid with CLU? Read Idontcare's epic 3770k delid thread and find out his experiences.

Or just ask me: being a long-time CLU users, I thought I could get away with a thin CLU application between the die and my IHS on my A10-7700k. Boy was I wrong! I had to slather it on thick to get it to work. Once I did, it worked beautifully.

 

[GreenChile]

Yes I delided my 3770K nearly 2 years ago and it's been cranking along at 4.7GHz ever since. I applied the thinnest amount of CLU I could manage and it worked great. The IHS was resting directly on the die so no gap fill was needed.

This is inline with IDC's observations as well.

 

[Techhog]

Personally, I don't think that you should be delidding for the sake of delidding, especially on Devil's Canyon. You should play around without delidding first to see if you really need it, since you're seriously risking your CPU and voiding the warranty completely for a drop in temps by a few degrees.

 

[DrMrLordX]

Yes I delided my 3770K nearly 2 years ago and it's been cranking along at 4.7GHz ever since. I applied the thinnest amount of CLU I could manage and it worked great. The IHS was resting directly on the die so no gap fill was needed.

This is inline with IDC's observations as well.

Uh no it isn't. People were knocking Idontcare for using too much CLU on the die. Personally I had to use a ton of it. I had two different prior applications on the die that were incredibly thin, and neither one worked with a darn. It overheated like mad. Between the IHS and HSF, a thin application worked just fine like always.

 

[ehume]

Personally, I don't think that you should be delidding for the sake of delidding, especially on Devil's Canyon. You should play around without delidding first to see if you really need it, since you're seriously risking your CPU and voiding the warranty completely for a drop in temps by a few degrees.

This. I have both the i7 4770k and the i7 4790k. The latter is a dream. If you had a 4770k, you'd probably see a substantial improvement in heat transfer if you delidded; i7 4790k, not so much.

 

[hawtdawg]

The gap between the core and the IHS has nothing to do with the TIM. Even the high temps of IB and Haswell barely have anything to do with the TIM. The issue is the black glue that holds the IHS on. that's what causes the gap.

I put a very thin layer of CLU on my core, and my max temps dropped by over 20C and gave me enough thermal headroom to gain another 200mhz.

make absolutely sure that you get all of the black glue off.

 

[Idontcare]

Uh no it isn't. People were knocking Idontcare for using too much CLU on the die. Personally I had to use a ton of it. I had two different prior applications on the die that were incredibly thin, and neither one worked with a darn. It overheated like mad. Between the IHS and HSF, a thin application worked just fine like always.

Damn, good find by you :thumbsup: I had forgotten I even wrote about my experiences, glad to see I took the time to document it. Really helped jog my own memory!

 

[Soulkeeper]

I used black rtv to glue the IHS back to the package with my AMD apu's
it has the same color/look as what they originally used
I just put it on right before the install so the pressure of the heatsink holds everything down while it dries.
you only need the smallest bit
this made them look like they were never delidded

 

[Soulkeeper]

I saw some reports that CLU causes corrosion after time, have any of you seen this ?

 

[hawtdawg]

I saw some reports that CLU causes corrosion after time, have any of you seen this ?

I delidded 2 years ago and have seen no change in temps during that time.

 

[DrMrLordX]

Damn, good find by you :thumbsup: I had forgotten I even wrote about my experiences, glad to see I took the time to document it. Really helped jog my own memory!

The Internet never forgets. Well, okay, sometimes bitrot makes it forget, but fortunately the AT forums don't have that problem. Yet!

I saw some reports that CLU causes corrosion after time, have any of you seen this ?

Corrosion of what? Copper? Silicon?

 

[Soulkeeper]

The Internet never forgets. Well, okay, sometimes bitrot makes it forget, but fortunately the AT forums don't have that problem. Yet!



Corrosion of what? Copper? Silicon?

of the heatsink/ihs metal
http://forums.anandtech.com/showthread.php?t=2362347

edit:
this youtube video claims "that rumors squashed"
https://www.youtube.com/watch?v=5tQx1LYDKsk
I guess bare aluminum is a no, nickel is ok, and copper is probably ok.

 

[GreenChile]

Uh no it isn't. People were knocking Idontcare for using too much CLU on the die. Personally I had to use a ton of it. I had two different prior applications on the die that were incredibly thin, and neither one worked with a darn. It overheated like mad. Between the IHS and HSF, a thin application worked just fine like always.

Yeah I wasn't referring to his application of CLU. I was referring to his observation that the IHS rests directly on the die once the glue/sealant has been removed.

If there is no gap remaining between the IHS and die then there is no need for large amounts of CLU to "fill the gap". YMMV.

Edit:
I just realized you seem to be making your argument about your experiences with an AMD CPU. How is this relevant to delidding Intel CPUs?

 

[DigDog]

nobody here saying anything against delidding?

OP, can you give us an idea of volt/temp/speed of your CPU first? You might not need to delid - or at least, not see much improvement, for the risk.

 

[DrMrLordX]

of the heatsink/ihs metal
http://forums.anandtech.com/showthread.php?t=2362347

edit:
this youtube video claims "that rumors squashed"
https://www.youtube.com/watch?v=5tQx1LYDKsk
I guess bare aluminum is a no, nickel is ok, and copper is probably ok.

Coollaboratory has made it clear in all the application instructions and other literature related to the CLU and CLP products that the gallium they use will lead to the rapid oxidation of aluminum. It has no such effect on nickel or copper. It will "alter" copper and change its appearance and texture (and possibly other properties), but it doesn't lead to mass corrosion or other unwanted effects.

Yeah I wasn't referring to his application of CLU. I was referring to his observation that the IHS rests directly on the die once the glue/sealant has been removed.

If there is no gap remaining between the IHS and die then there is no need for large amounts of CLU to "fill the gap". YMMV.

Right, but the point is that the post I linked shows how the observation of the IHS resting directly on the die had no bearing on the amount of CLU he had to use for proper heat transfer between the die and the IHS. He goes on to speculate as to exactly why.

Edit:
I just realized you seem to be making your argument about your experiences with an AMD CPU. How is this relevant to delidding Intel CPUs?

Under the IHS, they're about the same, at least as far as how the IHS makes contact with the die. I followed Sandy Bridge/Haswell delid instructions for razors (I did not trust the vice method for a CPU with pins and for other reasons), and my experiences with CLU's performance on the die mirrored Idontcare's almost exactly. I finally defeated my runaway temperature problem by using larger-than-normal quantities of the stuff on the die, just like Idontcare did. His post was the only reason why I knew to try that. That's why it's relevant.

nobody here saying anything against delidding?

Sure. It's risky, and you can kill your CPU doing it. It also voids the warranty. OP is an enthusiast after my own heart: he's apparently made the decision to delid whether or not practical circumstances require it. He's going the extra mile, just in case it might have some beneficial effect somewhere, somehow. Kudos to him if he can pull it off without murdering the CPU.

 

[DigDog]

i run a delidded cpu; but mine is a 4650k, and a pretty bad one too. the delid (coolab pro, nepton AIO) did not help much, and even though i was super-ridiculously-careful, i managed to nick the surface of the pcb. fortunately, no damage, but literally came within 1mm of disaster.

anyway, my point is that the 4790k, allegedly, doesn't have the concave HS and low quality TIM that together created the problem on the earlier haswells.

in particular, the 4790k which i just used to build a hackintosh runs at 45c@4.400, on 1.19v, on it own. i didint touch anything.

you don't delid a chip like that - you delid horrible chips which are overheating and that need extreme cooling, because of the gains. if you delid a chip that has already a "nearly perfect" thermal contact, you are risking without reason.

OP clearly stated "first time delidder". My point of view is that, if not necessary, not a good move with a brand new chip with great thermals.

(if first time delidder = LN2 pro, then go ahead, i won't stop you)

 

[Soulkeeper]

I personally wouldn't use clu, but I guess i'm just not as adventurous
I used AS5 on my relids
My main system I run bare die with as5

Whatever you do, just take your time and you'll likely be happy.
The only delid that went bad for me was cause I wasn't taking my time and sliced off a smd cap (amd has them under the IHS).
maybe take some pics and before/after temps

 

[DrMrLordX]

i run a delidded cpu; but mine is a 4650k, and a pretty bad one too. the delid (coolab pro, nepton AIO) did not help much, and even though i was super-ridiculously-careful, i managed to nick the surface of the pcb. fortunately, no damage, but literally came within 1mm of disaster.

Haswell's an interesting beast. But until you actually do the delid, or lapping, or whatever it is you're going to do that might help you get that extra 100-200 mhz or whatever it is you're trying to get, you never know how much it will or won't help.

I also nicked my CPU, about 3 times actually. It was a sloppy job overall, especially at the start. Next one will go better . . . that's what I keep telling myself.

anyway, my point is that the 4790k, allegedly, doesn't have the concave HS and low quality TIM that together created the problem on the earlier haswells.

in particular, the 4790k which i just used to build a hackintosh runs at 45c@4.400, on 1.19v, on it own. i didint touch anything.

You're correct. But you can still drop a few degrees with a delid/relid, especially if you also take the time to lap the IHS while you've got it off. Knocking off the nickel coating is good for a few degrees, usually. It doesn't mean you'll get the mythical, mostly-non-existant 5 ghz Haswell for your efforts, but it will run a little cooler.

you don't delid a chip like that - you delid horrible chips which are overheating and that need extreme cooling, because of the gains. if you delid a chip that has already a "nearly perfect" thermal contact, you are risking without reason.

Some people just delid everything.

OP clearly stated "first time delidder". My point of view is that, if not necessary, not a good move with a brand new chip with great thermals.

Somehow, he got the idea in his head that it's something he really wants to learn how to do. Might be a good skill depending on where he's going next. Custom water? LN2? We don't know. Hell he might pick up 100 mhz on his 4790k from the delid, and for some people, that might actually be worth it. You know, if it still works after he's done.

 

[eton975]

Somehow, he got the idea in his head that it's something he really wants to learn how to do. Might be a good skill depending on where he's going next. Custom water? LN2? We don't know. Hell he might pick up 100 mhz on his 4790k from the delid, and for some people, that might actually be worth it. You know, if it still works after he's done.

Well, OP shouldn't be practicing on a shiny new $300 4790K. He/she should be doing it on a Celeron G1610/G1840 or something.

 

[JM Popaleetus]

Does anybody just run the chip naked to the waterblock? It reads like everyone reapplies the IHS.