Bare-die testing: A delidded 3770k, an H100, and 9 different TIMs

[Idontcare]

After seeing some talk about this, I'm rather tempted to at least delid and apply better TIM. Although, my only question is which TIM? Everyone seems to suggest CLU, but I've already got Prolimatech PK-3 lying around. IDC, your tests showed that CLU was comparable to NT-H1, and other benchmarks show NT-H1 as being comparable to PK-3. So, it might just be worthwhile to use PK-3 instead of spending even more on CLU.

I'm not sure which removal technique (razor vs. vise) that I'd use. I've seen quite a few comments about how easy and painless delidding is with the vise method on a newer (Ivy Bridge or Haswell) processor.

What I learned with my tests is that there is a very real risk of using too little (yes, too little) TIM on the bare silicon die and this results in worse thermal performance versus just adding a little more TIM and letting the excess squeeze out under the mounting pressure.

It very much is a goldilocks type situation in which having either too little or too much TIM can result in equally poor temperatures, what you really want to strive for is the middle result where everything is just right.

 

[scgt1]

I didn't go through the whole post to see if others have tried a different tim then what is listed but has anyone tried Gelid? I know if was great stuff on my 920. I have since ran out and need to pick up some new tim but would like to get the best stuff that isn't hell to clean up.

 

[Aikouka]

What I learned with my tests is that there is a very real risk of using too little (yes, too little) TIM on the bare silicon die and this results in worse thermal performance versus just adding a little more TIM and letting the excess squeeze out under the mounting pressure.

It very much is a goldilocks type situation in which having either too little or too much TIM can result in equally poor temperatures, what you really want to strive for is the middle result where everything is just right.

My original concern with too much dealt with everyone recommending Liquid Ultra since it's conductive. However, I'm actually curious whether or not my water block (Koolance CPU-380I) can simply mount directly to the die. I'd have to look, but I'm pretty sure that the block isn't supported by the nuts that hold the four screw posts in place, which means that it should be able to lower the amount to make up for the lack of an IHS. If I mount it like that, I'd rather just use PK-3, which is non-conductive.

 

[Idontcare]

My original concern with too much dealt with everyone recommending Liquid Ultra since it's conductive. However, I'm actually curious whether or not my water block (Koolance CPU-380I) can simply mount directly to the die. I'd have to look, but I'm pretty sure that the block isn't supported by the nuts that hold the four screw posts in place, which means that it should be able to lower the amount to make up for the lack of an IHS. If I mount it like that, I'd rather just use PK-3, which is non-conductive.

Yeah you don't want sloppy amounts of Liquid Ultra anywhere in your system, be it on the die itself or on an IHS in a traditional TIM application setting.

That said, CLU itself really doesn't like to wet surfaces all that well, meaning it would rather ball up and stay where it is versus spreading out and flowing like water.

So in practice I suspect it would be rather difficult to get enough CLU on your die or IHS such that it actually "ran" down the side to where it could cause electrical shorting problems.

But being prudent is in order, use what you need and don't get sloppy about it, good advice from you that I hope others take to heart.

 

[SavageThrash]

I generally don't like to bump old posts but this review/guide was fantastic. I'm toying with the idea of naked cooling a 4770k with a h100i. This guide gave me a great idea of how to do it. I was looking around for a kit or something to buy for mounting but you seemed to have solved this one already.

I may need some calipers just to verify nothing has changed with haswell dimensions. Have you been using this setup since your review? Any negative effects of the h100 direct die setup apart from TIM staining? Some people seem to think the vibrations from the pump may not be good for the die.

Going to use AS5 myself as it's what I have laying around if/when I man up enough to do this.

Thanks a bunch, simply a great read!

 

[crashtech]

In my experience, AS5 has proven to be inferior to NT-H1 in performance AND ease of use. Also, it seems like the AS5 degrades more over time, but I have not really used the NT-H1 long enough to say for sure. If you do use AS5, reapply it at least once a year.

 

[stinger608]

Awesome threat Idontcare!!! I put my 5 star vote in :thumbsup::thumbsup::thumbsup:

 

[bononos]

From a previous post of mine in a different thread:

I have not had any issues with CLU needing to be re-applied. The liquid ultra itself does not contain any volatile components that could evaporate and dry out.

But being a liquid metal, it is prone to oxidizing and becoming a metal oxide (which won't be a liquid anymore, and won't have good thermal conductivity either).
.........

But aren't many thermal pastes made out of zinc or aluminum oxide?

 

[DigDog]

holy necro of resurrected threads!! five months dead!

 

[crashtech]

In my opinion, this thread is more than exceptional enough to warrant being brought to the top from time to time!

 

[ehume]

I agree. It should be a sticky. At least until Intel goes back to soldering their i7 and i5 quads again.

 

[blastingcap]

I also agree.

 

[SavageThrash]

In my experience, AS5 has proven to be inferior to NT-H1 in performance AND ease of use. Also, it seems like the AS5 degrades more over time, but I have not really used the NT-H1 long enough to say for sure. If you do use AS5, reapply it at least once a year.

Yeah I've noticed this with AS5 but when it's what you've got it's what you've got. To many tubes of it floating around to justify a new purchase. I'll put the money towards some better ram

And yes I really don't like to bump threads this old but it really is a wealth of information and I would love to see a revised haswell guide if it's needed/the dimensions have changed. I've been looking for a guide like this for a couple days. I didn't expect to find it on Anand but now that I have I guess it means I need to spend more time here.

Thanks again IDC. I've been using a mobile setup as my primary gaming rig for a while (direct die cooling a $1000 retail chip with 2 physically separate cores) and not doing the same with a single die haswell that I want to overclock just seems silly

 

[MongGrel]

I'm ordering some NT-H1 myself just for general usage, I hadn't bought any paste in awhile and thought I had some MX-2 left over and didn't when I bought this L5639 so I'm using cheap stuff at the moment.

I'm kind of a Noctua fanboi to begin with, most of my case fans and the two on the old TRUECopper on this old rig are Noctua.

I'm glad they finally became mainstream, years back people scoffed at em a bit, and yeah is a lot of good info this thread.

*salute*

 

[ali1988]

Am about to overclock a delidded 4770k. This is fantastic; thanks so much for doing all the laborious testing for benefit of OCers everywhere.

 

[Jollyriffic]

I plan on doing this same thing to my msi mpower and i7-3770k.
only thing not stated is what screws you used since the normal h100 standoffs don't work with the build.
could someone list what ones i should be looking for at lowes?

 

[Mr_Potato]

Hello IDC and rest of the forum! I'm glad I've come across this thread, it has answered some questions that I've had. Thank you for a very extensive tests and really sorry to hear about your CPU dying. I'm new to OCing as well as deliding so hopefully I will be able to get some answers from people who may have more experience than me. One question that I already have is - in terms of deliding - are all the CPUs the same? I've used vice and hammer method to "trial" one of the CPUs i found at work. it's an intel pentium 4 3GHz (i know i know - very old), however, the top lid has come off very easily. I have used a plastic card to remove the silicone from the board and something called technical acetone (used to clean the lenses on laser cutting machines) Not sure if it was a good idea. Anyway, not knowing if it's gonna work, i plugged it back in while holding a fan pressed to it and it bootted without problems. I'm little bit more confident of doing the process on my chip - i5 4670K. any suggestions so far? If there's anyone ocing same chip on a asrock z87 extreme 4 board - please message me directly as i have few questions
Once again, thank you for your hard work IDC and glad to be a part of community
All the best

 

[bloodindark]

Hello. I did some testing myself , seems like all is good for 2-3 days than the temps rise 7-10 degrees (so the test IDC did doesn't help me/us much , cause we ain't even talking about months of good temperature ,we talking days...so it isn't worth it). I really need your help, specially those who have been through my situation.the cpu is i53570k @4,4ghz @ 1.145v)
i had a hyper 212 ,with it the temperature reached 73-76 (AS5 paste). I decided to get a silver arrow ib's extreme,with it I have not seen a big difference, which annoyed me terribly 70-72 degrees . I dielidded the cpu, initially i applied AS5 between the die and IHS and same between IHS and HS. temperature reached 65 degrees, which was gorgeous, but after 2-3 days temperature increased again to 72 degrees ... after that I changed to MX4 paste, same story. I decided to leave the cpu without IHS, bare die / naked die, nothing, same problem, and I tried this even with Thermalright Chill Factor 3. i've got some ideas that would cause (dried paste?, or the paste not ok with that extreme temperature on die), but I want your opinion. I took noctua nh-1 paste, I have not tried it yet , as I think it will be a lost cause too. I thought about buying gelid extreme to see how it goes, if is not ok i will eventually go with coolaboratory ultra ,and ihs back , and between ihs - die go with CLU, and between IHS - HS to put gelid extreme.Awaiting your responses , Thank you

 

[Idontcare]

You need CLU or ICD for bare-die application.

 

[loccothan]

Im Using Arctic Silver no.5 and its OK to Me

gifs deleted

 

[bloodindark]

Im Using Arctic Silver no.5 and its OK to Me

with bare die?you have the same temps as in the first day?

 

[bloodindark]

You need CLU or ICD for bare-die application.

ok , but what i am more interested in ,is why i have such behaviour , after only 2-3 days , temps increased by 7-10 degrees , if someone can explain to me , please

 

[Idontcare]

ok , but what i am more interested in ,is why i have such behaviour , after only 2-3 days , temps increased by 7-10 degrees , if someone can explain to me , please

It is called the "pump-out effect" and this thread has a bunch of references to it earlier on if you skim through.

And here is another thread with discussion on the matter as well, special attention paid to this post.

Both ICD and CLU avoid the pump-out effect.

 

[bloodindark]

gelid extreme has pump out effect?and thank you for your response , i thought saw , i read the entire thread , but no one said pump out effect occurs within days you all talked about weeks/months

 

[TerryMathews]

gelid extreme has pump out effect?and thank you for your response , i thought saw , i read the entire thread , but no one said pump out effect occurs within days you all talked about weeks/months

The rate of pump out is relative to how much heat the chip pumps into the TIM. Think like boiling water on medium-low vs high on your stove.