7700K for Delidding!

Bandalo said:
I honestly can't test my theory either, not without a stack of Kaby Lake samples and some good x-ray equipment. I'm just talking the facts of heat transfer. That's the fun of the discussion though.
Click to expand...

Look at debauer's delidding of a 6950x. It's very surprising his findings with replacing the solder.
 
thesmokingman said:
Look at debauer's delidding of a 6950x. It's very surprising his findings with replacing the solder.
Click to expand...

I just watched the video, and while I don't speak German, I do have some observations.

I think my point is still valid...without a cross-sectional view, it's hard to determine the actual gap between the die and the IHS in a given CPU. That gap has FAR more to do with the heat transfer than the TIM itself, be it a thermal paste or solder. Now solder DOES have great heat transfer characteristics, but it's still more material to transfer through.

There's two factors that determine the temperature delta here. One is the total thickness between the die and the heatsink. So in a typical system you have this:

DIE -> internal TIM -> IHS -> external TIM -> heatsink surface

The second factor is the actual materials used for the TIM and the IHS. The thinner the TIM layer is, the less the material used actually matters. So if you have a 1mm gap, using solder or a high-quality TIM will have a large impact on temps of the die. If you have a 0.001mm gap, you could use whale snot and still have good die temps.

It's VERY, VERY hard to judge these things based on any online "tests", because the people doing the testing have no way to measure the gap before and after the tests are completed. If you could ensure the gap was the same, you could make a reasonable comparison between the effectiveness of Intel's TIM and another aftermarket brand. But most review sites lack the tools (i.e., an x-ray or some sort of non-destructive test method) to measure those gaps with any accuracy.
 
Can't wait to see the results here. Been researching this with mixed results from around the web.
Question I have is if anyone knows how well Liquid Metal holds up.. my understanding is the stock Intel TIM should last about 10 years.
 
Bandalo said:
I just watched the video, and while I don't speak German, I do have some observations.

I think my point is still valid...without a cross-sectional view, it's hard to determine the actual gap between the die and the IHS in a given CPU. That gap has FAR more to do with the heat transfer than the TIM itself, be it a thermal paste or solder. Now solder DOES have great heat transfer characteristics, but it's still more material to transfer through.

There's two factors that determine the temperature delta here. One is the total thickness between the die and the heatsink. So in a typical system you have this:

DIE -> internal TIM -> IHS -> external TIM -> heatsink surface

The second factor is the actual materials used for the TIM and the IHS. The thinner the TIM layer is, the less the material used actually matters. So if you have a 1mm gap, using solder or a high-quality TIM will have a large impact on temps of the die. If you have a 0.001mm gap, you could use whale snot and still have good die temps.

It's VERY, VERY hard to judge these things based on any online "tests", because the people doing the testing have no way to measure the gap before and after the tests are completed. If you could ensure the gap was the same, you could make a reasonable comparison between the effectiveness of Intel's TIM and another aftermarket brand. But most review sites lack the tools (i.e., an x-ray or some sort of non-destructive test method) to measure those gaps with any accuracy.
Click to expand...

There are two vids, one german and one in english. It's interesting just to keep in your head, but as I wrote before I'm not about to tell ppl what they can and cannot achieve when they're seeing big ass gains. Also, height doesn't matter with solder as it is fused to the IHS when soldered effectively becoming one piece imo.
 
Here is Roman Hartung's (der8auer) new model.....Delid-Die-Mate 2

 
thesmokingman said:
There are two vids, one german and one in english. It's interesting just to keep in your head, but as I wrote before I'm not about to tell ppl what they can and cannot achieve when they're seeing big ass gains. Also, height doesn't matter with solder as it is fused to the IHS when soldered effectively becoming one piece imo.
Click to expand...

It is effectively one piece which removes the interface layer, but there's still a total "thickness". If I'm transferring a fixed amount of heat through a metal, I'll have a certain delta through a given thickness. If I have a thicker piece of metal, the temp delta increases.

The height DOES matter then, even with an excellent conductor like solder.

My overall point on this whole topic is that the problem most likely lies with the manufacturing tolerances for the IHS to die interface, and not "crap Intel TIM".
 
Bandalo said:
if you removed the lid, scraped the old glue off and simply remounted the lid with the stock TIM, you'd see almost the same improvement in performance.
Click to expand...

yeah, I've asked Kyle if he would be willing to test that

that would be really be interesting

ehh now too late

though you could try it again with the worst TIM you can find
 
Bandalo said:
It is effectively one piece which removes the interface layer, but there's still a total "thickness". If I'm transferring a fixed amount of heat through a metal, I'll have a certain delta through a given thickness. If I have a thicker piece of metal, the temp delta increases.

The height DOES matter then, even with an excellent conductor like solder.

My overall point on this whole topic is that the problem most likely lies with the manufacturing tolerances for the IHS to die interface, and not "crap Intel TIM".
Click to expand...

If that were true, why does CLU get 20c and regular paste get 5c?
 
Bandalo said:
When you're done shaking your head, can you perhaps propose a theory as to WHY you're seeing a 20C difference? Even the worst, most generic thermal grease shouldn't have that much of a temperature delta if the surfaces are properly mated.
Click to expand...

because the original TIM is actually quite thick
I've delidded my 7600k
I would know

which leads me to belive that the 2 metal surfaces actually never made contact

and TIM is not a replacement for metal on metal contact
only a filler

like you said


thing is

from the top of my head, I can't actually remember if the 1 user that tried with AS5 glued it back together as well

by using silicon to glue it back together you could introduce the same "mistake" again
 
magoo said:
pretty cool little tool.

I think I'd simply use Noctura NT-H1 paste for the inside. Probably better than the standard Intel stuff.

I wonder if JB Weld would work or I think JB Weld has a high temp product, too?
Click to expand...

JB Weld....I like Magoo's enthusiasm.
 
Kelter said:
Can't wait to see the results here. Been researching this with mixed results from around the web.
Question I have is if anyone knows how well Liquid Metal holds up.. my understanding is the stock Intel TIM should last about 10 years.
Click to expand...

People are worried that it "dries out"
but have not seen that behaviour between the DIE and IHS

only had it for a year
worked fine temp wise

thing is
someone posted a manual for the liquid metal ultra

and at the end it clearly states that it does solidify, for safety reasons and it shouldn't pose a decrease in performance

right now I'm using a liquid metal from a different brand
 
Crixus said:
JB Weld....I like Magoo's enthusiasm.
Click to expand...

Damn straight.......JB Weld will never fail you.

If I delid my 400 dollar part and don't fuck it up, when I re-lid it I'm going to be damn certain it won't come apart.

My uncle used to use JB Weld to fix parts on his tractors.....
 
thesmokingman said:
If that were true, why does CLU get 20c and regular paste get 5c?
Click to expand...

Again, you can't control the gap between the IHS and the die in this type of environment.

Example - CLU has a thermal conductivity of 38.4 W/(m*K). Arctic Silver 5 is about 8.9 W/(m*K).

The thermal resistance across the die-to-IHS interface would be:

R = "air gap"/ thermal resistance of the TIM + contact resistance between the die and TIM + contact resistance between TIM and IHS

Lets assume for number's sake, a factory installed 0.1mm gap between die and IHS.

So R of the TIM layer would be:

2.6e-6 for CLU and 1.12e-5 for Arctic Silver, so quite an improvement.

But, say you could reduce that gap instead to 0.05mm. Then you get: 1.3e-6 for CLU and 5.6e-6 for AS5.

So at that thickness, instead of CLU being about 10 times better, it's only about 4 times better. The smaller the gap, the less difference the thermal compound will make.

As a final note, that "contact resistance" number can change too based on TIM...the better a TIM flows, the lower those values will be, but they're fairly small compared to the "air gap" resistance.
 
Peter2k said:
delidding on soldered parts only netted a decrease of ~5 degrees

wasn't really popular on sandy
Click to expand...

You're mistaken and got your examples mixed up. Relidding with regular paste TIM gets only around 5-6c drop vs using CLU which as you saw yourself nets 20c. The solder example is not what I'm referring to. Thus, height to height relidding is the same, the only difference is the TIM. Basically, the theory has holes and no one can prove squat. That's why I rather not venture into it and just accept it as it is.
 
thesmokingman said:
You're mistaken and got your examples mixed up. Relidding with regular paste TIM gets only around 5-6c drop vs using CLU which as you saw yourself nets 20c. The solder example is not what I'm referring to. Thus, height to height relidding is the same, the only difference is the TIM. Basically, the theory has holes and no one can prove squat. That's why I rather not venture into it and just accept it as it is.
Click to expand...
Sorry, but what's CLU?
 
thesmokingman said:
You're mistaken and got your examples mixed up. Relidding with regular paste TIM gets only around 5-6c drop vs using CLU which as you saw yourself nets 20c. The solder example is not what I'm referring to. Thus, height to height relidding is the same, the only difference is the TIM. Basically, the theory has holes and no one can prove squat. That's why I rather not venture into it and just accept it as it is.
Click to expand...

I'm referring to the old sandy days

delidding isn't new

it just didn't net huge improvements back then

we'll I'm taking the 20 degree in Temps on my 7600k

would still be interesting as to why there's such a huge difference

and again
I did delidd mine

the TIM was thicker then people would tell you to use under any circumstances

usually people tell you to as little paste as possible
like the size of a grain of rice

the TIM under my IHS was wide and thick enough for a whole bunch of rice grains :ROFLMAO:
 
Peter2k said:
the TIM was thicker then people would tell you to use under any circumstances

usually people tell you to as little paste as possible
like the size of a grain of rice

the TIM under my IHS was wide and thick enough for a whole bunch of rice grains :ROFLMAO:
Click to expand...

Retail applied paste is always on the generous side. It's what is called CYA. Take off the cooler on a gpu and guess what, it's over pasted. Not a surprise.
 
Hypergreatthing said:
So you mean the liquid metal tim, not just Liquid ultra, correct?
Like conductonaut (http://www.performance-pcs.com/thermal-grizzly-conductonaut-high-performance-thermal-grease-1g.html) should be the same as CLU, correct?
Click to expand...

Looking at the data sheets, they look comparable. The "Conductonaut" stuff has a 73W/mk conductivity compared to the 38 for the CLU. I think the CLU is slightly more viscous, so it may flow better, but it'd be hard to see a difference outside a lab I'd imagine.
 
thesmokingman said:
Retail applied paste is always on the generous side. It's what is called CYA. Take off the cooler on a gpu and guess what, it's over pasted. Not a surprise.
Click to expand...
But it is squeezing out to the side

while in my example it was a solid area of TIM

usually you'd expect the TIM to be squeezed out somewhere

as in
you want the 2 metal surfaces to have contact and the TIM to be a filler

it looked to me like there was no contact
 
Peter2k said:
But it is squeezing out to the side

while in my example it was a solid area of TIM

usually you'd expect the TIM to be squeezed out somewhere

as in
you want the 2 metal surfaces to have contact and the TIM to be a filler

it looked to me like there was no contact
Click to expand...

Ok, for the last time my gawd man. When relidded with normal tim, the gain is squat almost. Look at Linus' delidding video for an ex. Now when done with CLU, something happens and you get 20c drops. Kinda makes the whole omfg the height is too high theory moot.
 
Bandalo said:
Looking at the data sheets, they look comparable. The "Conductonaut" stuff has a 73W/mk conductivity compared to the 38 for the CLU. I think the CLU is slightly more viscous, so it may flow better, but it'd be hard to see a difference outside a lab I'd imagine.
Click to expand...

I'd think when it comes to liquid metal it's more down to how it handles

I have both liquid metal ultra
and
Phobya LM

they do handle differently

conductonaut is grizzlys version of it
TH021TG_121750_800x800.jpg


Tim_test_update1.png
 
thesmokingman said:
You're mistaken and got your examples mixed up. Relidding with regular paste TIM gets only around 5-6c drop vs using CLU which as you saw yourself nets 20c. The solder example is not what I'm referring to. Thus, height to height relidding is the same, the only difference is the TIM. Basically, the theory has holes and no one can prove squat. That's why I rather not venture into it and just accept it as it is.
Click to expand...

Height-to-Height relidding is NOT the same, and that's my point. Manually scrapping off all the old adhesive and TIM, replacing the TIM and reinstalling the lid almost guarantees a difference in that die/IHS gap. The application PRESSURE is key here as well. When you use a liquid-metal TIM, it's VERY thin and any excess is squeezed out easily when you stick the lid back on. Any other, thicker TIM is going to take more pressure to squeeze out.

Not much difference when using a heatsink, since you squeeze the excess out with the mounting mechanism. In the die-to-IHS interface though, you can only squeeze so hard before the sides of the IHS "hit the floor", and then you can't close that gap any further.
 
Peter2k said:
I'd think when it comes to liquid metal it's more down to how it handles

I have both liquid metal ultra
and
Phobya LM

they do handle differently

conductonaut is grizzlys version of it
TH021TG_121750_800x800.jpg


Tim_test_update1.png
Click to expand...

This second picture really supports the argument here. Difference between the BEST and WORST is only about 10C. I would NOT expect to see a 20-30C different on a delid-relid on a CPU then unless something else changed in the process.
 
Bandalo said:
This second picture really supports the argument here. Difference between the BEST and WORST is only about 10C. I would NOT expect to see a 20-30C different on a delid-relid on a CPU then unless something else changed in the process.
Click to expand...

That was my kaby a few days ago

the DIE and the IHS didn't actually have any contact, only through the TIM
or?

20170105_233422.jpg
 
Bandalo said:
Height-to-Height relidding is NOT the same, and that's my point. Manually scrapping off all the old adhesive and TIM, replacing the TIM and reinstalling the lid almost guarantees a difference in that die/IHS gap. The application PRESSURE is key here as well. When you use a liquid-metal TIM, it's VERY thin and any excess is squeezed out easily when you stick the lid back on. Any other, thicker TIM is going to take more pressure to squeeze out.

Not much difference when using a heatsink, since you squeeze the excess out with the mounting mechanism. In the die-to-IHS interface though, you can only squeeze so hard before the sides of the IHS "hit the floor", and then you can't close that gap any further.
Click to expand...

Not sure wtf you're writing now. I suggest you stop being a denier and look at the evidence. Even [H]'s delidding shows the same result. Otherwise, it's just blah blah blah...
 
thesmokingman said:
Not sure wtf you're writing now. I suggest you stop being a denier and look at the evidence. Even [H]'s delidding shows the same result. Otherwise, it's just blah blah blah...
Click to expand...

If you're not interested in the science, that's fine, but I'm not being a denier. I'm fairly sure my post was clear and to the point, and if you're still saying "wtf", perhaps you might ask some questions or re-read it rather than being a dick. If you could show evidence where JUST changing the TIM showed that much temp change, sure. But in every case it's more than just the TIM, it's also changing that gap.
 
Bandalo said:
If you're not interested in the science, that's fine, but I'm not being a denier. I'm fairly sure my post was clear and to the point, and if you're still saying "wtf", perhaps you might ask some questions or re-read it rather than being a dick. If you could show evidence where JUST changing the TIM showed that much temp change, sure. But in every case it's more than just the TIM, it's also changing that gap.
Click to expand...

No, you keep going around in circles. Maybe it makes sense to you, but from here you keep denying what other say and spin your tales. For ex. the height theory, its negated by the fact that relidding has been done with regular paste and clu. The difference is that with clu we find we get 20c average temp drop. You again deny everything stating height to height gibberish is not the same, its impossible to get temp drop of 20c with clu. And yet the evidence is there and repeated over and over again.
 
thesmokingman said:
No, you keep going around in circles. Maybe it makes sense to you, but from here you keep denying what other say and spin your tales. For ex. the height theory, its negated by the fact that relidding has been done with regular paste and clu. The difference is that with clu we find we get 20c average temp drop. You again deny everything stating height to height gibberish is not the same, its impossible to get temp drop of 20c with clu. And yet the evidence is there and repeated over and over again.
Click to expand...

The height difference hasn't been tested by this site...Kyle removed the old RTV/adhesive and applied the liquid metal and reattached. If he does another comparison using a different, lower-grade TIM, that would be a comparison.

When attaching the IHS, the height difference is set by that old adhesive and the new material. The stock adhesive from the photos and videos is all fairly thick, with a good layer between the PCB and the IHS. So simply replacing that adhesive will give you a height difference on the IHS-die gap.
 
Bandalo said:
The height difference hasn't been tested by this site...Kyle removed the old RTV/adhesive and applied the liquid metal and reattached. If he does another comparison using a different, lower-grade TIM, that would be a comparison.

When attaching the IHS, the height difference is set by that old adhesive and the new material. The stock adhesive from the photos and videos is all fairly thick, with a good layer between the PCB and the IHS. So simply replacing that adhesive will give you a height difference on the IHS-die gap.
Click to expand...

Again nothing but blah blah blah.
 
as far as the jb weld that is chemical epoxy and works by binding the surface metal to surface metal... I don't think the metal on top of the core is thick enough not to risk having metal dissolved too deep and etching into the layers of gates... some how having a path of least resistance end up forking into the heat spreader seems a bad idea. Oh I double checked the clean rooms of intel, amd and nvidia were oxygen free before so the paste does likely expand when the cores are glued to the heat spreaders. Which likely causes the paste to expand when the pressure in the room changes. So they are vacuum sealed like a jar of pickles not like a scuba tank. snicker you all can open a jar of pickles right?
 
drakken said:
as far as the jb weld that is chemical epoxy and works by binding the surface metal to surface metal... I don't think the metal on top of the core is thick enough not to risk having metal dissolved too deep and etching into the layers of gates... some how having a path of least resistance end up forking into the heat spreader seems a bad idea. Oh I double checked the clean rooms of intel, amd and nvidia were oxygen free before so the paste does likely expand when the cores are glued to the heat spreaders. Which likely causes the paste to expand when the pressure in the room changes. So they are vacuum sealed like a jar of pickles not like a scuba tank. snicker you all can open a jar of pickles right?
Click to expand...

The IHS is not air sealed, there's a slight opening in the glue for expansion.
 
You must log in or register to reply here.
Back
Top