Naked! Intel Core i5-7600K Kaby Lake How To @ [H]

The reason a shim is needed is 2 fold with Skylake onward.

First you dont want the cooler tipping on the edge of the core as it could chip it. Dead CPU.

Second the substrate the CPU die is mounted on is very thin. It bends under pressure from the centre when there is no heat spreader.
The shim allows the centre to bend in a small amount before the heatsink makes contact with the shim, then putting pressure on the shim as well.
This prevents all the pressure concentrating on the die.
 
Awesome :)

On the last run before shaving the corners of the socket, did you install the mounting screws on backwards on purpose?
 
And the $64K question:
How did the temps compare between all of those variations (stock, delid/relid, no lid) with the waterblock?
 
like it or not? yes I do like it and have been waiting for years for [H] to catch up on YT. I know its hard with the limited staff but I'm sure you can make it work. [H] prob has one of the biggest followings if not the largest(don't know the stats) of all the tech sites and YT exposure will add to it. plus you could always monetize and maybe use that for more video production/staff, maybe...
 
The reason a shim is needed is 2 fold with Skylake onward.

First you dont want the cooler tipping on the edge of the core as it could chip it. Dead CPU.

Second the substrate the CPU die is mounted on is very thin. It bends under pressure from the centre when there is no heat spreader.
The shim allows the centre to bend in a small amount before the heatsink makes contact with the shim, then putting pressure on the shim as well.
This prevents all the pressure concentrating on the die.

I feel you on the "need" to put a shim on it if you are from the good old T-Bird days ;)
But the reason why those usually broke where because of the CPU retention clip, the force on the die when mounting the heatsinks where on the one edge to start with if one was not very carefull.

When mounting a waterblock on modern cpu's it's quite hard to tip it if you mount the block correctly.

@ Kyle - Could you measure the thickness of the substrate when you remove the cpu next time if you have callipers ? could be interesting to see the thickness and compare it to the older models.
 
I feel you on the "need" to put a shim on it if you are from the good old T-Bird days ;)
But the reason why those usually broke where because of the CPU retention clip, the force on the die when mounting the heatsinks where on the one edge to start with if one was not very carefull.

When mounting a waterblock on modern cpu's it's quite hard to tip it if you mount the block correctly.

@ Kyle - Could you measure the thickness of the substrate when you remove the cpu next time if you have callipers ? could be interesting to see the thickness and compare it to the older models.
Yes, but without the IHS, you lose some spreading pressure on the die and CPU board. they made that thinner in last few versions. SO it flexes more without IHS.

Kyle I like these en-depth review with vid , normal reviews can be found all around but these types are interesting.
 
I feel you on the "need" to put a shim on it if you are from the good old T-Bird days ;)
But the reason why those usually broke where because of the CPU retention clip, the force on the die when mounting the heatsinks where on the one edge to start with if one was not very carefull.

When mounting a waterblock on modern cpu's it's quite hard to tip it if you mount the block correctly.
There will always be someone who can prove you wrong.
In Kyles video he wasnt scared of chipping the edge of the die because it has a sprung bound base of pins which would orient it correctly.
For his block that works well but there are other blocks whose tightening mechanism wont let the block sit flush to the surface before you tighten and are not so easy to match the tension on each screw or even know if you have tightened it enough. (Skylake needs less pressure, older coolers can apply too much when locked down)
And as Edkiefer and I mentioned, the shim helps spread the load so the substrate doesnt bend.

@ Kyle - Could you measure the thickness of the substrate when you remove the cpu next time if you have callipers ? could be interesting to see the thickness and compare it to the older models.
Internet to the rescue!
http://www.anandtech.com/show/9505/skylake-cpu-package-analysis
 
@ Kyle - Could you measure the thickness of the substrate when you remove the cpu next time if you have callipers ? could be interesting to see the thickness and compare it to the older models.
Kaby Lake - .75mm
Skylake - .75mm
Haswell - 1.10mm
Ivy Bridge - 1.05mm
 
Wow, 0.75mm is only 0.0295", thinner than most spark plug gap.

I didn' think it was that thin.
Yeah, I think using the naked vice method would not be a good deal with these KBL and SL CPUs.
 
Thanks for doing this Kyle, but I still have a dilemma.

When I first paste the die with IHS on top, temps are great on all cores.
After a couple of days to a week (depending on paste), temps on at least one core increase between 9C to 20C (again depending on paste) compared to the others, negating the point of delidding.
When I remove the IHS there are always points on the die where there is practically no paste, it squeezes out.
The tighter I screw down the cooler, the quicker it happens so I currently have the cooler quite loose.

I had an idea that the die might not be perfectly flat or flexes slightly under load, but it could be the mating surface of the IHS isnt perfectly flat.
I'm tempted to go for direct die as its seems more likely that my IHS surface is the problem.
Decisions.
 
Thanks for doing this Kyle, but I still have a dilemma.

When I first paste the die with IHS on top, temps are great on all cores.
After a couple of days to a week (depending on paste), temps on at least one core increase between 9C to 20C (again depending on paste) compared to the others, negating the point of delidding.
When I remove the IHS there are always points on the die where there is practically no paste, it squeezes out.
The tighter I screw down the cooler, the quicker it happens so I currently have the cooler quite loose.

I had an idea that the die might not be perfectly flat or flexes slightly under load, but it could be the mating surface of the IHS isnt perfectly flat.
I'm tempted to go for direct die as its seems more likely that my IHS surface is the problem.
Decisions.
Have you reattached the IHS? I think that might have something to do with it. I generally had better results with the IHS attached well with adhesive and squeezed down tight.
 
Yes, I've tried with and without.
It made no difference other than taking a day longer to reach the same state.
Then I had to delid it again to try different paste.
The best so far is Gelid Extreme, the extra thickness seems to help it stay in place giving the best temp of only +9C on one core.
It settled to that after one week. Another week later it has remained the same.
I would suggest that the problem is that the paste is drying out. From what I get from all of this is that relidding with a non-metal TIM is not a good idea for long period of time. I think it dries out quickly and causes issues where the metal TIMs do not gas off over time.

Have you inspected the TIM to compare what it looks like when your temps start going up?
 
With MX-3 it took 2 to 3 days for temps to rise 15C.
With basic white paste it to 1 day to rise 15 to 20C (depending on amount of paste used).
with Gelid Extreme it took a week to rise 9C. I have used the CPU in another PC since and after refitting in mine again, the +9C on one core remains.

I have had it apart around 10 to 20 times in the last couple of months trying to find what happens.
None of the pastes dried out, way too short a time span, it gets squeezed out when under load and is still goopy as if it came out of the tube.
I tried a huge amount of white paste and it helped by reducing the rise from +20C to +15C.
The worst it has been is +25C on 2 of 4 cores.
 
Nice video. I delidded my Ivy recently with great results, but I'd really like to try running it naked.
 
With MX-3 it took 2 to 3 days for temps to rise 15C.
With basic white paste it to 1 day to rise 15 to 20C (depending on amount of paste used).
with Gelid Extreme it took a week to rise 9C. I have used the CPU in another PC since and after refitting in mine again, the +9C on one core remains.

I have had it apart around 10 to 20 times in the last couple of months trying to find what happens.
None of the pastes dried out, way too short a time span, it gets squeezed out when under load and is still goopy as if it came out of the tube.
I tried a huge amount of white paste and it helped by reducing the rise from +20C to +15C.
The worst it has been is +25C on 2 of 4 cores.

Sounds more like uneven pressure on your IHS against the die. And only use a small amount of TIM it's only ment to transfer heat where there is no "direct" contact between the die and the IHS same goes for your CPU cooler and the IHS
 
Sounds more like uneven pressure on your IHS against the die. And only use a small amount of TIM it's only ment to transfer heat where there is no "direct" contact between the die and the IHS same goes for your CPU cooler and the IHS
Little paste results in hotter temps and more cores getting hotter.
I've tried changing the pressure at each end, it made around 1C change in temp difference between cores.
You can be sure I've tried many variations on the theme of pressure.

Something is curved, whether its the core or the IHS underside.
The more paste or thicker the paste, the less the temp rise and it takes a bit longer to happen as well.
ie the paste has a harder time getting squeezed out and a lot of paste perhaps helps paste move back in place when it cools.
Something is curved or becomes curved when it gets hot.
 
What ever happened to that new naked die cooler you guys posted about a few weeks ago? have you gotten your hands on one?
 
Little paste results in hotter temps and more cores getting hotter.
I've tried changing the pressure at each end, it made around 1C change in temp difference between cores.
You can be sure I've tried many variations on the theme of pressure.

Something is curved, whether its the core or the IHS underside.
The more paste or thicker the paste, the less the temp rise and it takes a bit longer to happen as well.
ie the paste has a harder time getting squeezed out and a lot of paste perhaps helps paste move back in place when it cools.
Something is curved or becomes curved when it gets hot.


Agreed...
The IHS is probably curved on the bottom, thus the more paste helps.

Where the cores temps even before deliding ?
 
No, they were close to 10C different.

If you recall back in the day, EK waterblocks had a slight bow to them in order to push the IHS onto the die. I am wondering, to your point, that the mating is uneven. If you have calipers you might want to check the height of each edge of the IHS.
 
If you recall back in the day, EK waterblocks had a slight bow to them in order to push the IHS onto the die. I am wondering, to your point, that the mating is uneven. If you have calipers you might want to check the height of each edge of the IHS.
Its the other side of the IHS that could be at issue, the one that mates with the die.

I can set specific angles with my block, I've tried many. Its the Apogee XT.
It tightens down to a height specific to the position of the nut on the thread.
When the block is first applied, the loose nuts make the block float above the CPU with a large air gap until tightened down.
Anyway, with the CPU floating on a bed of pins, a slight mismatch in angle wont cause an upset because the position of the CPU will adjust.
But I found that it doesnt matter what angle or how hard/soft I tighten, the problem always occurs.
The problem is not with the heatsink to IHS contact.
 
Its the other side of the IHS that could be at issue, the one that mates with the die.

I can set specific angles with my block, I've tried many. Its the Apogee XT.
It tightens down to a height specific to the position of the nut on the thread.
When the block is first applied, the loose nuts make the block float above the CPU with a large air gap until tightened down.
Anyway, with the CPU floating on a bed of pins, a slight mismatch in angle wont cause an upset because the position of the CPU will adjust.
But I found that it doesnt matter what angle or how hard/soft I tighten, the problem always occurs.
The problem is not with the heatsink to IHS contact.

I think I could have been more clear. I'm saying the IHS is not laying flat against the substrate and therefore not mating flat with the die. If one edge of the IHS is even a small amount higher than another edge, the IHS will mate unevenly. I think this is why Intel used such crap TIM, because it would be very high viscosity and not move around even when heated and could fill in any uneven gaps between the die and IHS. This is all just speculation but it's an avenue that you could explore.
 
I think I could have been more clear. I'm saying the IHS is not laying flat against the substrate and therefore not mating flat with the die. If one edge of the IHS is even a small amount higher than another edge, the IHS will mate unevenly. I think this is why Intel used such crap TIM, because it would be very high viscosity and not move around even when heated and could fill in any uneven gaps between the die and IHS. This is all just speculation but it's an avenue that you could explore.
I've tried many angles and without using silicone.
When not secured with silicone it floats on the die with the paste lubricating its movement (when loose on the desk).
The IHS doesnt touch the substrate anywhere. It sits as flat as possible on the die.
In case locking it down causes the angle to change as the IHS starts to contact the substrate, I have tried a fleeting contact with the block but the problem still occurs, it just takes a little longer.

Bear in mind I've done a ton of experiments and at no point has it behaved, not once. And the same problem was evident before I delidded.
I've refitted the IHS + paste close to 20 times and made pressure adjustments orders of magnitude times more.
I've gone through loads of different pastes!
 
Back
Top