Delidding 4790K thread

I delidded the 4790K. It's not done yet, but I am typing on it. Please keep in mind when I repasted, I only did the die and the lid. For heatsink, I just reused the crud already on it, so this repaste job is atrocious.

Just made sure it's working, I didn't break the CPU. Next I will be going to MicroCenter to either get ICD7 or Antec 7 TIM.

Preliminary results are already improved. Before I was having thermal throttling 30% using the x264 benchmark during Pass 2 at 100% usage and 3.9ghz, no thermal throttling, sitting at 89-92C @ 100% usage, 4.2ghz.

Be very careful delidding if you choose to do so, watch a video or a guide. There are capacitors on one side of the die, don't chop them off. Will post pics when I repaste.

 

nice, good job there zymphad looking forward to the pics and proper results

 

Can't wait to see the results. Considering sending mine off to get it done. A bit timid to do myself.

 

Delid


Lid back on repasted with Antec Formula 7

 

XTU stress test was around 84-87C. In Crysis before was around 92-94C, now it's around 82-86C.

 

What method did you use to remove the lid? How was the stock paste job?

 

May I ask... Why replace the lid? Why not leave it off?

 

problematic to do a bare die mount in batman, since the lower z height due to the missing lid would cause a misalignment of the shared cpu/gpu heatsink...

10C drop is nice btw

Sent from my Nexus 5 using Tapatalk

 

@Zymphad - did you delid with heat or block and vice method?

 

heat method? dont u mean razor? or is there a third one im not aware of....?

Sent from my Nexus 5 using Tapatalk

 

Did a CPU stress test and then used a razor right after. With Antec Diamond paste noticing the min temp isn't that great, but max temp is improved. I also have NH-1. But may give coollaboratory liquid a try for the lid.

 

aaah gotcha so the razor method is easier when stressing the CPU first hm? interesting

 

Will post back later with coollaboratory liquid ultra, but that is going to be a while, the shipping is slow according to Amazon.

 

This Friday I will be sending my chip off to a guy doing delidding service on eBay. He charges $50 and has a one day turn around including thermal testing and what not. He uses coollaboratory ultra also. He said if I choose to do this myself and use the same paste to make sure to seal the onboard vrm as that has been his only mistake in the past. I have never done this so I can't say if that's standard practice but figured I would share none the less. I'm going to take any and every temp measurement I can and post before and after. Is there any reason I shouldn't trust this guy or go through with the job through him? Besides the prob. One week down time he does guarantee his work and if the chip fails he will buy a new one.

 

in the beginning i was also thinking about using that guy´s services, but now im gonna do it myself i dont see a reason not to trust him, his ebay rep is positive through and through please do keep us updated though!

 

Will redo paste by following Antec instructions this time haha. Still have coollaboratory on order.

 

@Zymphad what are you using to reseal the CPU? Anything special or just some RTV black?

Edit
@jaybee83 I would imagine if he is using CLU he would want to seal the vmr still. Usually from what I have read people who seal the Vmr do so with the same silicone they use to seal the chip. That was my main reason for asking.

@pukemon the thickness of the sealant will almost negligible since you apply a pressure causing the IHS to displace the extra sealant. Yes it will cause the IHS to sit ever so slightly higher but as long as you are making good contact with a slight pressure on the chip it should not make a difference. The IHS is designed to reduce contact resistance. Going from the small chip to the large heat sink increases what is called the spreading resistance. Increasing this resistance will lead to poor heat transfer as it is based on the entire resistance network. To battle the spreading resistance that is where the TIM comes into play as well as the heat spreader. Basically as long as everything is in contact properly the difference of the chip sealed or left not sealed would be not noticeable as long as pressure exist the increase of pressure will not allow for better heat transfer between the IHS and heat sink. This is my understanding from what I have gathered from advanced heat transfer class I took regarding heat sink design

 

i guess nothing, just clamps the lid onto the die with the cpu socket clamp that way u can repaste easily

Sent from my Nexus 5 using Tapatalk

 

Yeah, but that sealant is what 1-3mm thick? Will you lose any clamping force or heatsink not make contact?

 

it wont be that much of a difference, but rather enough for the IHS to make better die contact without misaligning the heatsink

Sent from my Nexus 5 using Tapatalk

 

But what if the IHS does t make good contact with heatsink is what I'm saying if no sealant or something to make up for it. Come to think of it, probably fujipoly pads wouldn't be a bad idea. Get the benefit of making up for loss of z axis and heat dissipation.

 

The IHS is designed to be sealed using some sort of silicone sealant. There should not be any issue with it not making contact with the internals. I don't know about others but putting fujipoly pads between a metal based Tim eliminated the benefit of the Tim as the pads would be the lowest common denominator. I think you may be overthinking it.

 

I used nothing, as instructed by every guide I could find.

Also.changed order to Coollaboratory's Ultra Metal pad for Notebook so can use on GPU and copper heatsinks. Read the regular TIM corrodes copper.

 

I am getting mine pasted with CLU liquid I've also read it can corrodes copper but in further reading I found numerous users removing the lid after 1-2 years to prove that is false. Also watched a few YouTube videos disproving the same thing. Since I couldn't find any videos proving it does I figured that is overblown as CLU is so widely used there would be more info on it if it did corrode copper.

Edit: there are reports of it corroding aluminum but the IhS that Intel uses is copper so that is a non factor.

With that said I have the option of getting the chip back unsealed but just delidded. Are there any temp numbers that back up the fact that leaving it unsealed is the best route? Or is that just so when you re paste again you don't have to delid again?

Edit: @pukemon that's not a bad idea...I'm not sure if that would provide any benefit over silicon since the heat is not being transferred to that region. I can think of pros and cons.
Pros being that I would be much easier to delid in the future. Cons being that when the IHS heats up I would think the silicon provides better insulation from it being transferred back to the chip where as the pads would increase heat being sent back to the board.

 

I didn't mean using fujipoly on the die. I meant using fujipoly instead of a sealant for the IHS.

 

Copper shim vs the lid?

 

@pukemon: even if u took the thinnest pads at 0.5mm their thickness would probably still be larger than that of the silicone sealant. besides, the common argument is that the main factor reducing temps in a delidded cpu is NOT the exchange of the TIM but rather the tighter contact of the IHS with the die due to removal of that silicone sealant. also, zymphad's and bigspin's experience suggest that batman doesnt have any problems using delidded cpus as long as they still have the IHS attached

@baii: yep, thats the idea some users had, to solder a copper shim to the heatsink and mount the cpu baredie

Sent from my Nexus 5 using Tapatalk

 

Eh they refused to cancel so getting both and will have to RMA one. Lame.

 

so you're saying no sealant so IHS is directly on die? or the other way you mentioned adding a shim to the heatsink and not using the IHS? if so then i understand. just from all the videos i've seen that sealant looked pretty thick.

 

Finally a thread worthy of the Clevo section... it has been so many years since I had the pleasure of going through such a nice conversation.

 

@pukemon here is a good video showing the chip after sealant. It's not English but it will do the job

 

not necessarily solder, used to be paste-shim-paste. But I guess the limiting factor probably is the paste more than the piece of metal.

 

oi! u havent visited the batcave yet i see

Sent from my Nexus 5 using Tapatalk

 

I wonder if anyone will mod the backplate to work with a de-lidded cpu minus the IHS.

 

That would be amazing, although you have to be careful so that the heatsink's pressure doesn't break the CPU die.

As an advice for de-lidding CPUs, go to a local computer repair shop and ask for broken CPUs. They should give them away for free and you can practice de-lidding on them. At my former workplace we had many broken CPUs around and I practiced de-lidding on them.

 

The issue is still the height difference, how would you mod it so the heatsink will touch the die?

I will use the CLU on both GPU and CPU, read on NBR some have used it with copper heatsink too.

The guides I read specifically said to remove as much of the sealant to allow for the best contact possible between IHS and the die.

 

@Zymphad: by soldering a copper shim onto the heatsink and thus making up for the height difference...

also, i dont think theres any chance of the cpu die breaking in this case, since we dont have the same pressures seen in desktop heatsink mountings

Sent from my Nexus 5 using Tapatalk

 

@Jaybee - you won't know for sure unless you try...adding a thicker cooper shim will increase the pressure.
I am all down for this, just that I would first try with a cheap CPU for safety reasons.

 

All this talk has got me thinking. As you know I have sent my CPU off to get delidded. I've been bouncing ideas around and could use a bit more advice. I'm considering instead of getting the entire chip sealed to ask the guy to only use a small amount of sealant at the corners of the IHS. This would allow me to open the chip up later much much easier. Are there any negative effects to doing this?

Edit1: also I'm confused about this copper shim idea. By introducing the copper shim yes you in fact increase contact pressure but that will be offset by the addition of the contact resistance between the shim and chip and the shim and IHS. Since the IHS is made of copper and makes contact with the chip just fine. I just don't see the benefit to the shim idea. Maybe someone could explain it to me. I also don't get the theory behind not using th heat spreader. The heat speeder is there to smooth the transition from small CPU chip to large heat sink. This change in size causes massive spreading resistance that if not accounted for can dominate the resistance network causing poor heat transfer. Removing the IHS would drastically increase this resistance so once again I'm not sure the benefit. Of course all my information is what I have gathered in heat transfer classes and not practice so if someone could explain the difference they see in temps I would be very interested.

 

with IHS: chip - thermal paste - IHS - thermal paste - heatsink

both the IHS and the heatsink are made of copper = high thermal conductivity

so in this case, the bottlenecks are the two thermal pastes. best to use high end performing ones in both cases to maximize heat transfer. how can we do that? by delidding the cpu and exchanging the TIM under the IHS, additionally to the TIM between the IHS and the heatsink.

now, what would be an even better scenario? a scenario, where one of the two bottlenecks, namely the TIM, is removed! easiest way would be to just remove the IHS, paste the chip and mount the heatsink, right? in theory yes, but in the ZM series this would cause the shared GPU/CPU heatsink to misalign due to the missing IHS height. thus we need an alternative...

enter the copper shim! using a shim only makes sense if you solder it to the heatsink, thus making sure that the thermal conductivity is based on metal alone and NOT a second TIM! this way, u have both the advantages of using just one instead of two thermal pastes (= one bottleneck gone) and a properly aligned heatsink, if u use a copper shim with the same thickness as the removed IHS

now clear?

 

@jaybee83 thank you very much for the information. I get the concept behind why it is done. What I do not understand is why people eliminate the IHS. Eliminating this will drastically increase the spreading resistance. This is a very important part of the thermal resistance network often overlooked. Here is an article to help explain what I mean.
http://www.electronics-cooling.com/1998/01/calculating-spreading-resistance-in-heat-sinks/

Adding the heat spreader will smooth the transition lowering the resistance to the heat sink aiding in better heat transfer. By using the shim method it will raise the spreading resistance a lot negating any benefits of losing one of the bottle necks and possibly decreasing the over all heat transfer. Please do not take offense to me asking these questions as I am just trying to learn. Adding the shim goes against everything I have learned in obtaining my PhD in Mechanical engineering. In theory it seems like it should work fine but much like increasing the size of the heat sink can lower the overall heat transfer due to the increase in the spreading resistance I think that is how it will behave here as well. By no means am I saying anyone is wrong but I guess I just wanted to add my view and see how it actually works. If you or anyone else is interested I have a matlab code that I have written that will calculate the heat transfer from a CPU. You input the heat sink parameters and it will auto spit out an optimized design. If you play around with it you can see that one wrong move and you can raise the spreading resistance almost double or even more causing the temps to raise by a great deal.


Edit: Jaybee I misread your post a little. Going to leave the above for a theory of heat transfer but you are correct as long as the size is the same as the IHS is the same and you solder the shim to the CPU this would maximize the heat transfer. In reality if you tapper the shim to be wider on the heat sink side theoretically you would increase heat transfer even more.

At any rate I would like to thank you guys for answering my questions as I'm really enjoying learning all this stuff.

 

If I recall, the problem with Ivy Bridge and Haswell isn't the TIM (Intel uses DuPont paste--good stuff), it's the gap between IHS and CPU due to too much sealant used to attach the IHS. Delidding and then removing the excess glue from the edges lets the IHS sit lower on the CPU, improving heat transfer.

 

We'll find out soon enough since I'll be switching to liquid metal instead of paste. With the amount of CPUs Intel sells, surely they could leverage enough product to get liquid metal at ridiculous low prices and use that instead.

 

No you would be better off removing the backplate, inserting a material the thickness of the IHS adds to the cpu and then putting it back (I suggest the use a micrometer to measure it properly), this will make the mounting holes sit closer to the motherboard and allow perfect mating to the bare die.

That way there is no extra thermal layers to deal with and would be the professional looking solution.

 

@octiceps: that is correct, removing that lil bit of extra thickness due to the sealant, the increased pressure between die and IHS causes the chip to make better contact, thus increasing the thermal conductivity to the heatsink

Sent from my Nexus 5 using Tapatalk

 

Can you use a picture with indicators of what you mean? I'm not understanding what you are saying, a diagram would help.

 

So... I used Coollaboraty's Liquid Metal Ultra for both GPU and CPU.

At least know nothing fried. It's working. I'm running Prime95. Turned on maximum fan, Fn + 1 if you didn't know and it reached 88C on P95 after 5 minutes. My conclusion is this heatsink/fan system can't cope with P95, P95 is just a crappy software made to destroy your CPU.

Also it's just luck. Some of you have better CPU than I do. De-lidding and using CLU didn't produce the results I wanted. It does idle lower, but not by much. And the overall maximum heat is little better, so there was improvement, but not what I was looking for.

Would I recommend it? Not really. I believe if we had better heatsinks like desktops, yes. But for P770ZM I believe it's the lottery of what you get. CLU will delay getting to the max heat, but ultimately this cooling system can't cool fast enough to keep up with CLU, so eventually the maximum heat just builds and builds until reach the threshhold. That's my impression.

Even for GPU, it took longer, but at 165/250 OC it still reached 75C with CLU. It seems there is a limitation to how fast our cooling system can dissipate heat and it's not the fault of TIM as much as binning lottery.

 

Yes we continue to use it to stress the CPU but it has been repeatedly reported that prime 95 does not work well with Haswell or ivy bridge chips. I don't know why...

 

Well it's old crappy hardware that doesn't take advantage of SSE4.2 and AVX2 and other tech that Intel has, I would imagine the calculations P95 is doing is very inefficient for current hardware.

But some here have been reporting they idle 46C. And sorry, delidding and CLU if yous was idling around 54-56 like mine was won't do it. I idle around 51-52C with a low of 49C after 23 minutes, so it will help but it's Intel's lottery that you are battling, not Clevo's thermal system or TIM. CLU is the winner of my testing between Gelid vs Diamond vs CLU though.

 

for proper idle temps, leave ur machine switched on overnight while recording with hwinfo or similar. then ull get those mid to high 40 temps on average otherwise its correct, when doing light work like browsing or office and then leaving the machine alone for a short period of time ull get something in the low to mid 50s.

what was ur application method for CLU? ive seen user reports indicating that its best to apply clu onto both the heatsink/IHS and/or IHS/chip for best results

Sent from my Nexus 5 using Tapatalk