Cracking Issues with Skylake and LN2 Benching? Der8auer May Have the Solution
Skylake CPUs can be notoriously susceptible to the issue of cracking when benching with LN2, a phenomenon where the heat conductivity between the pot and the HIS is reduced by trapped air between the IHS and the CPU PCB. It’s usually prefaced by an audible ‘cracking’ noise. Deliding the CPU and replacing the thermal paste with better quality paste can help but can also cause problems if the black glue that holds the lid in place is removed or damaged. With this problem in mind der8auer has come up with a possible solution.
Der8auer’s idea is to add a thin spacer between the IHS and the PCB to reduce the space and thusly the potential influence on cracking. Der8auer, known to most of us simply as Roman, being the community spirited guy he is kicked off a forum thread on HWBOT to share his idea with the community, also offering to provide IHS spacers to any interested parties for a nominal fee so they can conduct their own research on the issue.
“It's just a guess but I think the space between IHS and PCB has some kind of influence on the cracking behavior. I made some spacers out of stainless steel. The thickness is 0.10 mm and you can stack them e.g. if you want to use 0.30 mm spacer. Just fix them with a little bit of TIM.”
On the issue of eliminating the space between PCB and IHS, Joe_cool also proposed an idea using copper blocks instead. Clearly there are plenty of overclockers out there looking to resolve cracking on Skylake chips.
Check out the forum post and follow the discussion here.
If you benched Skylake on LN2 already, I'm sure you had some kind of cracking issues. We already tried a lot to improve the cracking. Mainly by changing the thermal paste.
However, also the black glue between IHS and CPU PCB seems to help. At least it was much worse for me whenever I removed the black glue.
It's just a guess but I think the space between IHS and PCB has some kind of influence on the cracking behavior. I made some spacers out of stainless steel. The thickness is 0.10 mm and you can stack them e.g. if you want to use 0.30 mm spacer. Just fix them with a little bit of TIM.
This is a pure research project - nothing I'm trying to make money on. I have 50 spacers here and paid 3,75 € per spacer. Would be cool if some guys join in and help me investigate further. I can ship a pack of 5 spacer worldwide for 23 €.
Count me in, you can ship me a pack and I can mail to anyone in the US wanting to try it.
Count me in
I have never had a single crack on either of my Skylake chips and both have the black glue on them still... so I totally agree with you here. I imagine the thermal paste is trying to expand as the chip rapidly starts putting out heat. If there isn't room for it, the pressure builds until you get the crack at which point the pressure from the paste trying to expand is greater than the pressure holding down the pot. The crack is literally the sound of the paste exploding. Scratching the die makes sense here as it gives the paste a bit more room to expand.
I'd take you up on a pack but I don't want to remove the black glue on mine lol
I'm in, but doubt if we even need more than 3 or so.
So if we can determine the good spacing height... we get a fully copper IHS from Der8auer with the right dimensions
That's the goal
omg gib plz
Count me in Roman.
Good stuff!! want it!
ooooo cool idea.... if only the shipping to SA was not 45 euro. i would so be interested in testing as well
no no. I will ship then in a small letter. 23€ already includes the price for 5 spacers + shipping.
PayPal is ready.
I like that idea
Maybe Roman could make a more functional IHS with fins on the side
In for IHS!
Oh ok that's ok then if there is still left... please me please
I will keep 10 myself for my own testing so I can send out 8 packs of 5 spacers.
DrWeez
Gunslinger
RULE
nacho_arroyo
xxbassplayerxx
Leeghoofd
XtremeAddict
One more can join the party
Okay Splave took the last kit. Will contact you guys soon with details
Let me know when they become available again and I'll buy 5x from you.
I sense something fishy here
^check your upper lip mate
Been there done that, what's next
Hey Roman, i will test this little thing next week @ cascade.
And i have also some more advanced ideas for milling...
I will let you know if i can get some improvement with this.
Because i will not start a mass production of anything like that, i do it only for myself.
Hey! I had a similar idea but I'm not sure about the correct dimensions. That's why I made the small spacers so we can test to find the correct spacing
I will start with the normal 0.35mm deep milling in ihs and can go deeper anytime...
Without glue on cpu i think you don't need more than 0.5mm.
In my opinion, we have to focus more on the backplate then, to get the optimal pressure.
_
That could work out good aswell
How will you test exactly?
First i can test improvement at my -115°c idle cascade only, but last 2 weeks at march i have time to test it with let say 200l ln2.
Till then i will make some different backplates and maybe some bigger fullcover ihs copper blocks for socket mounting.
_
Ship me one boss
First step, direct from milling machine...
looks great
Finished with microstructure surface.
I have to say,this looks wonderfull
Awesome, Marc. If this works with your cascade next stop gold
Since we are talking about this,has anyone got any opinion how a silver IHS would perform? given that it has 10% higher conductivity than copper i think it could be an alternative to copper? Don't know the costs,but does it matter when we talk about overclocking?
Looks great
Will be interesting to see how it performs.
@alex: Silver would perform better, yes. It's just a guess but 200-300€
If gains are sensitive i can see a winner
Thx guys!!
With little gimmick.
This reminds me very much of your legendary mem pot
- stylish 
Would be interesting to take this and expand on it, instead of a just the IHS, make it into the base of a 2 piece pot, which would reduce 1 copper to TIM contact point. The retention bracket would be the next problem I guess.
..
Awesome idea, I'm interested in how heat transfer capabilities are like compared to the stock IHS. Waiting for results
Tip: next time don't use comic sans font, Its horrible
Don't worry guys, Roman will test it this weekend @ ln2.
_
Nobody test the pot directly on CPU die, with some plate protection to retain the chip in the socket? I've see some stuff like that on FB.
Plate only wont work. You have to cut the edges of the socket because the edges are higher than the die.
Naked die works with AIO CooerMaster Nepton which has big base. Don't see why it would be different for LN2 pot.
No retention bracket of course.
PS: On a second look you might be right though. I was getting good contact, but have overlooked this fact. The die is very thin.
I guess you have measured it.
PS2: I think it is just enough. At least on the OCF. The die is a little above corners of socket, but almost the same height. Will even act like a stopper to prevent excessive clamping force. Not sure it will be enough though. Had that idea to try it, but still haven't.
This should be a solution to avoid cut edges?
Or the edges still touch the pot?
You can get it to work sometimes but it will have a bad contact. Solution would be to cut the edges and use the aquacomputer spacer.
Or using custom made
What if you rotate the pot by 45 degrees (e.g. F1EE has relatively small base)? Then it should work without cutting corners.
Will have to fasten it with something else though. Or make a custom "diagonal" mounting kit.
Hi guys
i'm using this method but strangly those edges were not a problem at all [maximus viii gene],
i'm just using some thick Tape,they fix cpu on socket and also help the die to handle pot pressure,thats a messy method but works fine in short times.
pot directly on die[gelid between them] = cinebench 6200 and xtu 6140mhz and it was good for a chip that can not pass cinebench more than 5900 and 5800 xtu with "pot on ihs" method
i had not grizzly and i was using a [maybe] rotten Gelid , but i used the same thermal paste for those bench sessions.
would love to try one, been using tape and all sorts of things.
says:
if it is just a trial,u can cut in half that and try:D alot of thickness from 0.02mm to 1mm
EK-SF3D Inflection Point EVO will fit to touch only the DIE. It has small base (round one), which is much smaller than IHS of CPU
I got the chance to use an EVO last week... even on a C2D it's hard to hold a temperature with that. It was swinging all over the place.
Vince has spoiled me
XA has used them a while. Maybe hes quick-pour mcgraw. If you've only used an f1, then try the venom, you'll see what i mean. Much quicker response.
Yeah, @steponz was saying the same thing.
I like to be lazy with my pouring so the Dark is perfect for me
Ive got both, personally prefer the venom, even on haswell with the precise cold bugging.
Kit arrived, testing this weekend
For those who want to try to help BUT don't have Roman spacer, you can try with paper sheet
. Easy to cut, cheap and you can use paper even with black glue.
Classic paper is less than 0.1mm so if you want more can use more sheet or find high quality paper.
I didn't try but it should works
Great idea Roman
wow, very interesting information.
Hi!
Seems like I'd found a good solution for fullpot benching.The idea was to not put too much pressure on the cpu circuit board as we know its thin so changes shape easily. I was thinking about the possible cause of cracking for two days straight and it seems I was right in one way.It can be caused from the deformed pcb not making even contact with the ihs.
In this session I put one layer of dielectric tape next to the die to support the ihs and not push out all of the TIM because I was thinking that its better if that whole thing freezes into 1 piece solidly.
So my easy solution is to put some tape under the ihs to support it a bit,and keep the factory distance from the die,apply more paste than you would normally,and put light pressure on the pot.It depends on which pot you have,but I have an Xtreme ONE which is a 4,4kg beast so It doesnt need any additional force.OH and I put a little mx2 paste under the side of the IHS like the factory glue,so It froze together solidly,and no crack until -180c which gave me additional 100-150mhz 4c/8t benching..
Next thing we should think about is how to support the inner section of the cpu,so it doesn't bend into the socket,so we can use more force on the pot which could make even better contact.
just my two cents..
Yea that basically confirms my testing as well. But we have to know which distance will lead to the best results. Ideally we will have a special IHS with the perfect distances to avoid cracking.
Not sure if bending is causing any more problems. I think it's more the pressure and thickness of the TIM layer.
Did you put the layer also around the whole IHS?
Good testing!
So here's my finding on my 6700K... Absolutely no crack the three times I froze it. I delidded it, left all black residue, used Kryonaut under the die and between pot, and tightened all down tight.
All of that, and I was running full pot, -192 or so, without issue.
Wow very intresting discussion. I have an idea but i don't know how to make it. How about make IHS for Skylake with this bracket. So IHS will tight and no air in. No accident IHS moved because you use heavy pot. Even if you use heavy pot maybe this design will help to make a "perfect" ihs for skylake. Maybe you can use the paste under IHS many times
This morning is starting pretty shitty. Roman's cpu is dead 00
I'm currently testing another one:
with 4 spacer there's not much contact between the die and ihs.
with 3 spacer the Grizzly broken @ -170°c
Now i'm gonna test with 2.
Thanks man!
yep,firstly I tried with these two little pieces then went further and covered the whole pcb.It needs a layer of tape anyway to keep the moisture away from the socket pins.Only thing I was care of is not to overlap the tape so there won't be a similar surface.Looks like it worked,but slow pouring is needed.Only went down to full pot when tried to run 6.3 hwprime at 1.85v.
was also playing this game haha. Had circular IHS's made. Problem was the pip in the middle.
Gonna have to do that last part on a mill instead of a lathe.
From my test the spacer doesn't help the cracking stuff. Have you guys found something different?
I'm still testing on watercooling... I ran before like Wizerty suggested with one layer of scotch black tape... so I guess 1-2 spacers is plenty ... rest might be overkill...
I dont care about the snap anymore becauses once I get it i just heat to 60ish degrees in wprime 5ghz 1.60v and till pot temp hits around -20c its fine again.
Im happy to hear it because I only get it once a session usually.
my best xtu run is after a snap and using this method.
Well if you don't have enough paste it won't work again Allen... I glued back the IHS in the beginning and it took me a while to figure out my clearance was not enough. So if I had the crack the paste was literally a goner... -120 pot temps was +40 on idle, heating up made it only worse... So the clearance needs to be correct... It will always crack in my book after some time, hence why many re-apply the paste before doing a new session...
2nd session without cracking,with the method I described.
Interesting concept to take up space missing from cutting the glue.
I've seent his with AMD processors that had been de-lidded and decided to just leave the IHS plate off.
However I have lapped the bottom edges around the IHS plate where it meets the processor PCB to be sure it touches all the way round four corners.
Very cool, hats off!
~ShrimpBrime~
Is it not just easier to mount while ambient and warm to +60, then pull down?
Why do you cut the plate in the size of the old one? It is necessary to heat spreader was the size of the PCB - would cover the corners. Those. should be a new heat spreader type of http://abload.de/image.php?img=imageidsya.jpeg but larger size
It's elementary strength of materials. Lever arm, strength bending...
Not to put words in his mouth but I took it as him saying he does this after a crack, so running cold for a while and this is his method of re seating the thermal paste so he can keep benching
Because it's a first test only. I build that because i need more mass (atm ca. double weigt) ihs for my cascade. Other option for me was to make more mass evap... but that's not so easy for testing.
_
A heat spreader the size of the PCB would definitely help prevent PCB bending--great idea!
Your picture I showed an example.
My main idea is to move the pressure points to the edge of (to the point of contact CPU PCB with the socket)
Reducing the pressure on the center (CPU crystal) not the main, and of itself not solve the problem.
In theory. I do not have subge a processor and guided only by a pictures deformed processors, logic, and knowledge of subject "strength of materials". Well, and the experience of working with similar problems.
I know what you mean, i have myself a lot of other idea's to solve the prob, but not for public now.
this
Leeg you need use more TIM if heat up method doesnt work to fix it
give this a try!
Paste is cheap....
Cheaper than dead cpu ^^
Time to buy stock in Thermal Grizzly
This goes against everything I understand about thermal paste. remember when it was just for filling the microscopic gaps?
So i think i do it right from the first. With big glop of paste my temp is much lower with custom water cooling. Now i just need GC Extreme (because no grizzly here).
says:
I've been using splaves paste metherd as well works great,full pot ever session with no crack
That's the funniest application of thermal paste I've seen. Remember all those article and forum threads where people discuss if you need a blop, a cross or a square? Nope, you need a pile of paste.
Ninja sales tactics by Roman
Yeah haha
Now its more about "creating" more "connection" points between IHS and die so paste can handle the "stress" without cracking
says:
I
0.15 is the last that fit easy....so i think that a tickness of 0.20 should be fine....but a metal tickness not as said wizerty in paper....paper is squeezable
Sent from my E2353 using Tapatalk
What about the gap in the glue which leaves a breathing hole for expanding and contracting of the paste and it's ingredients?
I only assume that the hole which is there in the normal glue would be some sort of breathing gap for the natural movement of the tim.
Just a wild guess tho.
As for what I understand, all things expand and contract when put under hot or cold. So why wouldn't it be the same for the tim. And that could be the Kraken.
why not drill a hole in the top corner of the IHS like x99
Should probably just drill a hole through the chip itself
Sounds like a lot of drilling to me ;-)
Hey Roman, what's up??
I send you one of my copper ihs for testing and till now not any response!?
If you don't need it, pls send it back to me!
_
I have been using strips of electrical tape for months now and it never seems to be consistent. At one point no crack, same exact method next time and crack. But I have tried 4 strips on each side, I have also tried 2 pieces high on each side. I have also tried glueing the ihs back everytime and re-delidding.
Then I will bench and think I fixed it and do the same thing again and it cracks. It's really strange and hard for me to figure out. I have also been using a lot of paste but not as much as splave showed. Will try that this weekend with new chip.
I also tried with i3. delidded and glued back together with a little higher glue and a ton of paste and I had terrible cracks.
Some of my best results have been cracking it on purpose and heating up with xtu and then benching more.
its definitely the most frustrating thing I have come across since starting xoc.
Have an issue, hopefully someone has an idea what the **** is going on, cause I can't figure it out. So I recently delidded 2 chips using delid tool. Scraped glue off pcb and ihs with credit card, as I've done every chip in past. Re-timmed 6320 and ran without any issues. 2nd time I tried running 6320, mobo wouldn't boot. Fast forward to 2 weeks later, today. 2 different mobos. Both 6700k and 6320 will not post. Both look pristine without bend, scratch, or other defect. Multiple mountings and still no post. I've seen codes 14 and 00 on Gene and 04 on OCF. Multiple sticks and configurations of ram, cosm cleared dozen times, no post. Why would this be happening? Never had any issues like this before with delidded cpu. I even checked 1st skylake i bent and it boots no problem. So why would 2 chips using safe method, appear dead?? I'm at a loss.
You've been running on OCF or what board?
I have both Gene and OCF. Have more experience with OCF. Like I said, tested with known good chip with no issues. Just doesn't make sense. The 6320 ran fine for one session, but the 6700k has not posted after tool delid. Not a scratch or defect on either. Both chips were briefly tested on water before delid and ran fine. Actually very good.
I dont know if this relates to your situation but you mentioned scraping the glue off. I did a perfect delid on my 6700k and it worked great. But I wanted to apply new glue so I scraped off old glue. After scraping off glue cpu wouldnt boot anymore, i tried a few more times to boot and it actually killed my m8e.
There wasnt a single scratch or imperfection on the cpu but I must have twisted it a tiny bit when scraping glue, the pcb is so thin.
Just be careful not to kill mobo's if the cpu's are dead. That really sucks, sorry you have to go thru it
I had bad experience also with a chip without glue on. I tested around 30+ SKL chips now mixed of i3/i7/i5s
and delidded around 10. All 10 tested cold with the glue on and no problems, non of them died on me. Then
I bought a delidded chip without glue on so I used white bond paper as a spacer, 1st day tested on AIO works
and second day powered up my gene and instant 00
* Looks like removing the glue is bad for our health
^ agreed leave the glue ftw on ihs and cpu
I wonder if the die is smashing with glue off... as @Schmuckley showed, they're extremely easy to shatter.
wish i could do the same but i had shitty IHS+contact:
So no way i could leave glue on.....
I found the proper glue just need to find a method to stick it back with a little bit less presure than socket contact,then wait it try then delid again...
After thinking on this all day..... Came home and checked, the first chip i delidded, i had left glue on for some reason. I really scrutinized the recent two chips. Sure enough, both had made direct contact with ihs and the 6700k definititely has crack in it. Both were very promising chips with low voltage. I dont have a lot cash to play with and really, really sucks. Wish I would have seen this thread a couple weeks ago. See If i can scrape together something for a 6320 at least.
what do you mean crack in it?Cracked die ? Don't think socket exercises so much pressure,hmm weird
Yes, crack in die. upload of pics keeps failing. You can see some abrassion on both dies and 6700k clearly has line perpendicular to any sliding of ihs. I'm testing with a air cooler on top without clamping down, so yeah, just socket and retainer pressure. All that from mounting each chip twice.
I also think it's the actual process of removing the glue. With the pcb being so thin it twists ever so slightly.
after removing glue had 00 on m8e and didnt want to believe it and pushed the start button 3 more times and then mobo would never turn on again..
But it does also smash the die. My first chip had really bad marks on the die just from the pressure of ihs on die and not from sliding or anything. One piece of electrical tape on each side was enough to stop that, but it was crazy the marks left in the tape from the ihs, the hold down really squeezes the two together without that glue.
I'm taking a chance.. Heating them up with hair dryer and gluing both chips back together with black rtv gasket. If they don't boot.....
That long perpendicular scratch across 6700k is weird. Can't help but think its a crack. The ihs slides forward when clamping. Also the length of it is out of ordinary.
To avoid sliding it forward, loosen the Torx screws, place the chip and close the clamp, then tighten the Torx down.
You can also just use the Delid Die Mate to glue it back
@ Joe: I gave it to dancop on the day I received it. Will check with him
Yeh i just killed my nice 6700K as well. Crack in the die and 00 post code. Not fucking happy.
Did you have the glue removed?
no. always left it on.
That sucks man, sorry to hear about that. What do you think caused the crack?
Both chips continue to be dead. Nothing worse than losing good chips! I think it would be wise to re-glue after replacing the tim in any event. The glue helps to stiffen the pcb. Without it, the pcb can still flex a good bit. So rules for skylake delid -
-Use delid tool or razor
-Don't twist pcb
-Don't scrap the glue off
-re-glue after replacing tim
no idea. thats the problem. was working fine, and died at 4.5g on aio cooler in 03 nature. wtf man
I talked to @Der8auer in Poitiers and told him I will delid my chip, and try to remove the glue and to fill ALL the space between IHS and die with Kryonaut.
Well, I had a crack at -160, warm up and try again. No crack afterwards and benching full pot for 2 hours. Chip is alive and well. So cleaning the glue had no negative impact for me. I used MSI Z170A Xpower.
@der8auer any follow up on these or the proposed upgraded IHS?
If they're tested and available I'd be very keen to grab them off you at Computex.
Here we go folks,the first 99.9% SILVER IHS to be tested. Bigger mass than original ,should be very interesting to see how it goes on cold and silver's reaction to extreme cold compared to copper.
Given silver's higher thermal conductivity 406 vs copper 385 i except a few degrees improvement on water, sort of guaranteed,just need to test if on ln2 there will be any gain and hopefully no bigger crack problems.
Will update with new thread once it arrives
Big thanks for James@Rockitcool !
Wow. Have the depths been changed to possibly deal with cracking/contact under the IHS?
Bling bling...
Moved the silver IHS in this thread
should fit better here.
Will be interesting to see if it helps.
@ zeropluszero: We didnt manage to get an reliable data yet.
Nop,same as stock,can be polished inside if needed but i don't know if height is the real issue for cracking,in fact using what Daniel posted i had no crack in last sessions with no spacers at all . The real challenge is how silver will react to extreme cold .
Anything on the spacers?
Still waiting on South African stupid post office
ya after using dancop's method I have been able to bench full pot with no cracks anymore. The big thing for me was I was trying to be cheap and reuse the paste under ihs for a few sessions, now new paste every time and no cracks.
The interesting fact is how the crack/thermal paste degradation occurs.On my backup cpu i get classic crack sometimes,i can quick check a cinebench run with realtemp and see positive temps at -150 pot so i know paste has cracked and shut down platform.
However my best 6700K has different behaviour , no positive temps in realtemp but huge degradation of clocks,can't bench anything over 5900 ,going over will simply freeze.
Also a good tip would be that when u press the lever on the socket i need to seat IHS properly from first try,no matter what. Each time i seated it wrong and did it again without adding paste i had cracks or issues.
[quote name=Alex@ro;443846
Also a good tip would be that when u press the lever on the socket i need to seat IHS properly from first try' date='no matter what. Each time i seated it wrong and did it again without adding paste i had cracks or issues.[/quote]
That's true.
so serious question.....what would happen if someone tryed to solder the IHS to the die on skylake? what would be the possible downside/damage?
edit:
something like this maybe
http://www.indium.com/thermal-interface-materials/solder-tim/
http://forum.hwbot.org/showthread.php?t=145035
thanks, for some reason i didnt find that when i google searched.
so it works, just havent found the right procedure/method of doing it effectively.
wonder if this would make a difference,
found here, http://overclocking.guide/the-truth-about-cpu-soldering/
i might give it a go at some point and see if i can get better results.
anyways, sorry for going slightly off topic.
EDIT:
duhhhhh, if i would have kept scrolling down in that same page they talk about soldering skylake.......man i feel silly now. ANYWAYS
Yea I tried it in cooperation with a professional soldering lab which is specialized for soldering solar cells. But still the result was not good. Considering the amount of money I invested in it, it was totally not worth it haha.
(I wrote the article on overclocking.guide if you didn't notice
)
Actually i did not notice hahaha. Man, im having such a bad day today.....i should just go back to bed.
so has there been any positive results with the spacers yet? or is the final verdict to apply lots of paste, crack it, heat it and go to town?
I don't think you ever want to crack it intentionally.
sorry, poor wording on my part.
should have said something more like. "apply lots of paste, bench untill it cracks(if it cracks), heat it up to fix the crack, continue."
Then yeah! Sounds perfect
So then i guess changing the distance between the die and ihs had no effect on the cracking.
it does but so far I don't have any reliable data yet
ya I just started noticing this the past couple days, makes sense but just didnt realize. 2 nights ago I put ihs down but it was really misaligned so I had to adjust it too much and paste cracked pretty fast.
But last night I put ihs down almost perfectly first shot and not one crack all night, full pot for hours.
Thanks to dancop guide I use the asus cpu tool now which really helps with alignment.
Got my 6100 too almost 6.4ghz hwbot prime but os was a mess, tried stripping win 10 myself but it wasnt right, need more practice.
Please log in or register to comment.