Delidding the Intel i7 7700K CPU

[Jade_Monkey]

Hi everyone,

 

I've been experiencing some absurd temperatures in my CPU at stock settings due to Intel's horrible thermal paste solution.

I figured i would share my experience delidding the CPU just in case anyone was interested in doing it too.

Just to clarify because the name is a bit misleading, delidding consists on removing the metal cap that covers the CPU, cleaning the stock thermal paste, applying a better one, and re-installing the metal cap again.

 

Temps (no overclock, Closed water loop):

Before:

• 89C (and some spikes above that too) while gaming.
• Around 45 -50 C when idle

After:

• 51 C Max during stress test
• 33 C idle

The process is pretty straightforward, I used the following items:

1. Rockit88 delid tool. (Shipped really fast).
2. Arctic Silver kit to clean the paste and to reapply paste OUTSIDE the CPU
3. Thermal Grizzly Conductonaut for the INSIDE of the CPU
4. J-B Weld 31314 Red High Temperature RTV Silicone to Seal the CPU (make sure to leave a small unsealed part for thermal expansion). You can use Superglue instead of silicone, so this step is optional.

 

The process takes about an hour and the results are immediate. Here are the instructions from Rockit's website.

 

Just in case you go for it, do not use any sharp tools around the CPU to remove the old silicone. Use the Arctic silver kit #1 liquid and your fingernails.

 

The results are amazing even when you do a mediocre job at it, so anyone suffering from high temps with Intel CPUs I do encourage to try it if you have a minimal knowledge of how to install a CPU.

[dburne]

 

Wow, some very nice results from your efforts, way to go!

[Jade_Monkey]

Yes, my guess is that mine was worse than average. Now I have some room to overclock it.

[TG-55Panthercules]

Congrats on your results!  (and courage   ) 

 

I've got all the tools and parts now and am just waiting till I work up the nerve to do this to my 7700K.  Some questions though, since you've already done it:

 

1.  While what you said in your steps 2 and 3 match other things I've read, I'm curious why it is that if the conductonaut is so good for use between the CPU and lid, why wouldn't you also use it between the lid and the cooler instead of the arctic silver?

 

2.  Other things I've seen have talked about using cottom swabs/isopropyl alcohol to clean off the old thermal paste (and IIRC that's what I used the last time I had to remove an old cooler and install a new one).  What do the cleaners in the Arctic Silver kit have in them other than the alcohol?  (I've got the Arctic Silver paste left over from my recent new build/upgrade, but not the cleaners.)

 

With my latest settings (and new Noctua fans) I'm not having any real problem temps or fan noise issues at this point (temps staying in mid/low 70's during gaming while OC'd to 4.8), so I'm not sure if I'm gonna do it yet, but it would be nice to shave maybe 10-20 degrees off just for stability/longevity considerations.

Edited October 6, 2017 by TG-55Panthercules

[CanadaOne]

Hi everyone,

 

I've been experiencing some absurd temperatures in my CPU at stock settings due to Intel's horrible thermal paste solution.

 

 

 

 

What kind of CPU cooler are you using?

 

Woudn't a good liquid cooler, like one of the Corsairs, handle the temperatures well enough?

 

Edit: Sorry just saw the "closed water loop" part of your post. And that isn't enough?

Edited October 6, 2017 by CanadaOne

[chiliwili69]

 

 

Temps (no overclock, Closed water loop): Before: 89C (and some spikes above that too) while gaming. Around 45 -50 C when idle After: 51 C Max during stress test 33 C idle

 

Thanks for posting your experience with that delicate process and being brave enough to do it yourself. Congrats!

Now you can go to really good OC.

 

When you said 89C while gaming, do you refer to BOS? BOS in VR?

 

Also, 51C during stress test is quite good. What stress test? an what temp you have while gaming now?

 

I am going to install liquid cooling on my rig (NZXT Kraken X52) and wanted to know what temps should I have with BOS with no OC.

[SCG_Fenris_Wolf]

Thank you for sharing.

 

Wili, I wonder if this means we can run the CPU at 4.9 or even 5.0 with AVX instructions getting called (VR IL-2). May try it later this month.

[Jade_Monkey]

Congrats on your results!  (and courage   ) 

 

I've got all the tools and parts now and am just waiting till I work up the nerve to do this to my 7700K.  Some questions though, since you've already done it:

 

1.  While what you said in your steps 2 and 3 match other things I've read, I'm curious why it is that if the conductonaut is so good for use between the CPU and lid, why wouldn't you also use it between the lid and the cooler instead of the arctic silver?

 

2.  Other things I've seen have talked about using cottom swabs/isopropyl alcohol to clean off the old thermal paste (and IIRC that's what I used the last time I had to remove an old cooler and install a new one).  What do the cleaners in the Arctic Silver kit have in them other than the alcohol?  (I've got the Arctic Silver paste left over from my recent new build/upgrade, but not the cleaners.)

 

With my latest settings (and new Noctua fans) I'm not having any real problem temps or fan noise issues at this point (temps staying in mid/low 70's during gaming while OC'd to 4.8), so I'm not sure if I'm gonna do it yet, but it would be nice to shave maybe 10-20 degrees off just for stability/longevity considerations.

 

 

Based on your temps, I wouldn't bother delidding.

 

1- It crossed my mind too but I decided not to find out by trial and error. I think it has to do with the fact that conductonaut is a liquid metal, not a paste, and it doesnt stick to one place.

2- it's the same thing, just prepackaged for convenience.

[TG-55Panthercules]

Well, I finally decided to go ahead and give it a try, and delidded my 7700K yesterday.  Process with the Rockit 88 tool was super simple, just like the videos I saw on you-tube.  I let the thing cure/set up overnight, and re-installed it this afternoon.  Got a momentary scare when it threw me into BIOS on the initial boot-up, with a message saying a new CPU had been installed.  But I checked everything and all the BIOS settings seemed to be still correct, so I went ahead and exited and completed the startup process.

 

Everything seemed to start up just fine.  I ran the ROG Realbench stress test for 10 minutes, and temps hovered in the low 70s with a max of 75 (this is the one I had to stop right away before because temps were hitting the mid-90s).  Checked with MSI afterburner during some gaming sessions, and temps were mostly in the mid-to-high 50s (whereas they were mid-low 70s before).  So, it looks like I gained about 15-20 degrees, pretty much in line with what was being reported in the you-tube videos and articles I saw during my research.

 

Hardest part was doing without the PC for the 24 hours, and even that was actually refreshing for a change.     Overall, a very positive experience

Edited November 7, 2017 by TG-55Panthercules

[=TBAS=Sshadow14]

why didn't intel just fix the cpu So it operates at normal temps before release.

Good cpu desktop idle should be around 25-30*C (room temp) on good liquid cooled custom loop
With max Load temps around the 50*C Mark. (but it is cruddy intel)

Edited November 7, 2017 by =TBAS=Sshadow14

[dburne]

why didn't intel just fix the cpu So it operates at normal temps before release.

 

Good cpu desktop idle should be around 25-30*C (room temp) on good liquid cooled custom loop

With max Load temps around the 50*C Mark. (but it is cruddy intel)

 

That's about the temps I get with my 4820k with the Corsair H100i closed loop cooler.

 

What the heck did they do with the 7700k...

[TG-55Panthercules]

why didn't intel just fix the cpu So it operates at normal temps before release.

 

Good cpu desktop idle should be around 25-30*C (room temp) on good liquid cooled custom loop

With max Load temps around the 50*C Mark. (but it is cruddy intel)

 

Even delidded, I'm not getting anywhere near those temps (idling around 40 and hitting 55-60 at gaming max loads, low 70s at stress test max loads).  Of course, that's already OC'd to 4.8.  Not sure what stock temps would be - guess maybe I'll have to run some more tests to satisfy my curiousity now

[SCG_Fenris_Wolf]

Read up on that, and to answer your question sshadow, it's about longetivity of the paste used, and the way it can get applied. Liquid metal lacks the former and creates issues in mass production in the latter.

[TG-55Panthercules]

Well, didn't see much of a drop in temps when going back to stock 4.2/4.5 CPU settings (maybe 3-4 degrees, into high 30s at idle).  Also, saw a bit more increase when I went to 5.0 for a couple of tests, but still manageable.  Still not sure what I think about what I've read about effects of running AVX over 4.8 for longer periods, so I've dropped back down to my old 4.8 for now.

[SCG_Fenris_Wolf]

What did you read about running AVX over 4.8 for longer periods?

[TG-55Panthercules]

What did you read about running AVX over 4.8 for longer periods?

 

Something about Intel recommending against it, but there wasn't much detail or resource citation in what I saw so I've got to do some more digging.  Wasn't really an issue/possibility before the delidding, since temps above 4.8 were a bit higher than I wanted to live with long term, but now that I seem to have the thermal headroom available I'll have to do a bit more research on that question.

[SCG_Fenris_Wolf]

I received my delidding tools... except for the TG Conductonaut liquid metal paste. That one got delayed. Will put this beneath and on top of the heat spreader. Can't wait to see the results

[TG-55Panthercules]

I received my delidding tools... except for the TG Conductonaut liquid metal paste. That one got delayed. Will put this beneath and on top of the heat spreader. Can't wait to see the results

 

I wondered about why most folks I saw doing videos/reports on delidding seemed to be using the Conductonaut (or other liquid metals) only between the CPU die and the lid, and not between the lid and the cooler (seemed to me if it was so great for one it should be good for the other as well).  The consensus among the sources I found seemed to be that there really wasn't much to be gained in using the liquid metal between lid and cooler as compared to using a good paste like Arctic Silver on top, so that's what I wound up doing.  Seemed like there were some potential concerns about either or both the difficulty in applying the liquid metal on such a large surface (without worrisome overspill) and the difficulty in cleaning off the liquid metal (compared to the paste) if you ever need to change coolers or something, and the one test I saw by one guy who tried doing it both ways seemed to conclude that the performance gain between the paste and the liquid metal on top was pretty negligible.  

 

Not sure I would characterize anything I saw as clearly dispositive though, so it will be interesting to see how it works if you decide to go that way.

 

Good luck!

Edited November 11, 2017 by TG-55Panthercules

[Dakpilot]

With my small research on this subject, that seems to be the way people do it in my small part of the world locally here, metal between lid and quality paste between cooler.

The new intel coffee lake seem also a (potentially daunting) candidate for this process, especially with the higher ambient temps we generally have here

 

Cheers Dakpilot

[SeaW0lf]

I think people use silicone paste in between the lid and the cooler because of what Panthercules already mentioned: the difficulty of applying the metal paste, danger of a short in case of leakage, difficulty in spreading the metal paste and even surface corrosion (I've read of users saying that the Liquid Ultra / Pro erases the CPU data written on the lid overtime. One excellent option is the Thermal Grizzly Kryonaut, which has a slightly worse performance than the metal ones and is considered to be the best silicone paste on the market and can also be used between the die and the lid. Because certain pastes cannot be used between the die and the lid -- for example MasterGel Nano and NT-H1 -- because of thermal shock; drastic temperature changes, which does not happen between the lid and the cooler. It appears that some pastes, although having high thermal conductivity such as the Nano (11 W/m-K), don't cope well with thermal shock.

 

Tom's Hardware has good thermal paste reviews. The closed loop results were with a Quad Q6600 (done in 2017), but it is a 105W CPU and has a soldered die (better than if they were tests with an i7 with TIM paste).

 

http://www.tomshardware.com/reviews/thermal-paste-comparison,5108-8.html

 

Another detail is that the Arctic Silver 5 is a very old, outdated thermal paste, which has cure time and is frequently counterfeited (it is very popular among computer technicians because the customer has often heard of it and would buy it in a heartbeat, many times paying more than it is worth).

 

The best investment would be the Kryonaut or GC-Extreme to apply in between the lid and the cooler.

 

And in general the problems people are having with the i7-7700K in stock is due to uncontrolled core voltage. I think the core voltage controller went to the motherboard and they usually set it very high on auto. I've seen lots of users running a stock i7-7700K with 1.35V and beyond while gaming. Deliding is good for overclock, but for who is running it stock and is having temps problems, usually it can be easily fixed by setting a fixed core voltage in the BIOS (if the motherboard allows core voltage settings). 

Edited November 11, 2017 by SeaW0lf

[TG-55Panthercules]

And in general the problems people are having with the i7-7700K in stock is due to uncontrolled core voltage. I think the core voltage controller went to the motherboard and they usually set it very high on auto. I've seen lots of users running a stock i7-7700K with 1.35V and beyond while gaming. Deliding is good for overclock, but for who is running it stock and is having temps problems, usually it can be easily fixed by setting a fixed core voltage in the BIOS (if the motherboard allows core voltage settings). 

 

I'm having a little trouble digesting all the info I've seen so far about the relationship or role of core voltage (vcore?) in overclocking.  It sounds like, as a matter of general principle, that it would be best to be using the lowest possible vcore that will provide stability at any given OC level (from stock 4.2/4.5 to 4.8 or 5.0, for example), but I've never really messed with the vcore while I've been doing my various OC tests - I've just set the multiplier manually in my BIOS.  So, I suspect that my vcore is probably set to auto somewhere, but none of the monitoring tools I've been using (CPU-Z, MSI Afterburner) seem to be monitoring/displaying vcore, so I don't really have a view to what's going on in that regard as I vary the OC level.

 

If I go into BIOS and see that my vcore is set to auto, what voltage would it make sense for me to set it on manually for my current 4.8 OC, and how would I know what to change it to if I wanted to try 5.0 again for testing?

[[SRH]Festa_VR_Noob]

Try HWMonitor, by the CPUID guys.

 

That has vcore and much more.

[SeaW0lf]

If I go into BIOS and see that my vcore is set to auto, what voltage would it make sense for me to set it on manually for my current 4.8 OC, and how would I know what to change it to if I wanted to try 5.0 again for testing?

 

It will depend on the chip, and it can vary a lot. I would say from 1.25V (a premium chip) to 1.35V or even more if the chip is bad. You can check the clock and voltages people are getting on this overcloking guide. It is a renowned forum for overclocking and cooling. There is a guy OCing the i7-7700K at 5.2Ghz with 1.3V. I imagine he could OC to 4.8Ghz around 1.2V or even less.

 

I'm a bit out of touch with overclocking, but in general the main parameters are clock, voltage and LLC (Load-Line Calibration), which prevents the voltage to drop and crash the system while the CPU is being pushed (during stress test for example). In stock settings the LLC will be on auto, but when overclocking you have to set it manually, usually above average. My motherboard has auto / medium / high / turbo and extreme (or something like that). When I OC my CPU to 4.6Ghz I set LLC to high or turbo. It raises the voltage, but it can pass OCCT with no crashes. Or you can leave LLC on medium or high and raise the voltage. It is a compromise, but in both ways you are going to have the voltage about the same.

 

And to find the right voltage you have to test with a stress test. I consider Prime a bit too much. I use OCCT (linpack test is what people generally use). You set a voltage on the higher side, let's say 1.350V for 4.8Ghz. Then you run OCCT for 10/20 minutes. If the voltage passes the test, drop it to 1.340V and repeat OCCT. And you do that until OCCT fails. Then you have an idea of where the voltage stands. And you start to tweak the LLC and voltage in increments of 05. Then you test for longer times on OCCT (1-2hs). Some people leave OCCT running overnight. When everything is set and done, I like to raise the voltage 05 or even 10 for good meajure.

 

And there are other settings as well, and then there is the memory. It is always good to OC the CPU with the memory on stock (no XMP). Then you OC the memory after.

 

But I'm saying all this for my Ivy Bridge. Kaby Lake might be different because they change things with the generations, switch controllers from the CPU to the motherboard and vice versa. When I started overclocking my CPU I read almost 200 pages of the Ivy Bridge OC guide (on the same forum I linked). I asked a couple of questions and then my first OC was a breeze. In a week I saved a couple profiles on the BIOS (4.4Ghz and 4.6Ghz). 

 

It is not complicated; you just have to understand how it is done. But I would learn before going for 5Ghz. Or even 4.8Ghz. Better to leave it all set and neat, because voltage and temperatures at higher clocks is a dangerous mix.

[TG-55Panthercules]

Well, I took a look at that guide, and then I found that the ASUS tool I was using to set my fan profiles also reports/displays vcore, and it says it's at 1.296 with my current 4.8 OC.  That seems about right, and the stress tests (Realbench) are still showing temps staying in the mid-high 60s, so I guess I'll probably just try one more test at 5.0 and then "declare victory" one way or the other and get back to just playing the game.

 

[EDIT]  Loaded up the mainboard's 5.0 OC preset in BIOS and ran some tests. ASUS tool reported vcore just below 1.4, which seemed a tad high but pretty close to what I was seeing in that guide's chart. Stress testing ran quite a bit hotter (about 10 degrees more than with 4.8) - hovering high 70s-low 80s, with max 85, but temps during IL-2 were staying down around 60.  In VR using my relatively maxed out settings and OVRdrop/map turned on (so, basically, how I would normally fly), I gained about 6 FPS AVG on the benchmark track (about 11%, from 52.4 to 58.4).  I guess I'll keep it there at 5.0 for a while and just keep an eye on things.

Edited November 12, 2017 by TG-55Panthercules

[GridiroN]

I recently myself delidded my 7600K, but it did not work. Perhaps some of you will have some experience with this to help me out.
Ok, this is a very long story so I'll try to be concise:

I bought a delid kit for my 7600K from Rockitcool. I've tried to delid 6 times now and none of them have resulted in performance improvements, most of them resulted in extreme performance handicaps.

The first I delid, I fouled up the silicon relid job, so I had to redo it. The next 4 times I tried resulted in serious performance degredations (idles of 50, loads of 90) until I realized I was using the delid tool incorrectly.

I tried a 6th, time, making sure to do everything right: I delidded the cpu, cleaned it up, put 4 small dabs of silicone RTV around the edges, used CoolitLaboratory liquid metal, applied to both the die and underside of the IHS. Left it overnight.

Testing now, my temperatures are pretty much the same as the factory lid job.

Before delid:

Room: 21c
Idle: 45c
Load: 78c

After 6th delid:

Room: 19c
idle: 45c
Load: 80c.

I'm lost as to what the issue could be. The tim between the cooler and IHS is GC Extreme; the best. I think it's possible the H100i GTX bracket is not putting much pressure on the CPU because the threads get tight before they bottom out which leads me to believe they're uneven and not tight, so I've ordered the backplate bracket from the original H100i that works much better to test that.

Any help or advice would be appreciated. I'm totally loss as to how my temps haven't improved even 1 degree. 

 

I've ordered a new tube of Coollaboratory Ultra because I've gone through most of my last one.

Edited November 12, 2017 by GridiroN

[SeaW0lf]

Loaded up the mainboard's 5.0 OC preset in BIOS and ran some tests. 

 

Cool, that's easier. These BIOS are getting fancier with time. Mine only has an auto OC slide to increase clock, but then the core voltage on auto goes ballistic. 

[Lemon]

GridiroN ,  i delided my 6600k oc to 4.5Ghz , with CoolitLaboratory liquid metal too on the die only (not on the plate) , and the CoolitLaboratory metal pad , between the cpu and the noctua cooler, and i went from mid 90 deg to low 70 on cpu intel burn test (set to maximum),   so there must be something wrong on your side, you should have much better results with these products, so if you follow carefully the instruction how to apply the liquid metal, the only problem i can see with what you said in your post it is the silicone , you say  ( Left it overnight ) do you mean you waited long time before to install on the motheboard ? if you did not wait before install it forget what i say here ! maybe your cpu is damaged i dont know !?

 

 But if you waited silicone dry before install it , the problem may be too much space between the die and the plate,

when you use silicone rtv between 2 metal parts, like for a cover valve of a car and you tight all 10-14 bolts there is still a good thickness of silicone between parts, so if you let it dry with no pressure or not enough, you may have too much gap between the die and the plate, and it may definitively be a good cause of bad heat dissipation of your cpu , also be sure you remove carefully and completely the old silicone, you want the die and the back plate as near as possible
 

when i did mine i did not wait to install the cpu and the cooler on the mobo, to have a pressure on it before the silicone dry,  but it is little tricky cause the cpu plate slide a bit (cause silicone is not dry yet ) when you lock the cpu with the cooler retainer, so i had to install the cpu back plate very little offset over the die (about 1mm aprox )so it sit at the right place when it slide a bit when you install the cooler, and i use very little of silicone,   that is what i would do with your i7 for the 7th try..  you are very patient sir ! 

 

hope my English is good enough you understand me, wish you good luck to enjoy low temps on your rig

Edited November 13, 2017 by Lemon

[GridiroN]

GridiroN ,  i delided my 6600k oc to 4.5Ghz , with CoolitLaboratory liquid metal too on the die only (not on the plate) , and the CoolitLaboratory metal pad , between the cpu and the noctua cooler, and i went from mid 90 deg to low 70 on cpu intel burn test (set to maximum),   so there must be something wrong on your side, you should have much better results with these products, so if you follow carefully the instruction how to apply the liquid metal, the only problem i can see with what you said in your post it is the silicone , you say  ( Left it overnight ) do you mean you waited long time before to install on the motheboard ? if you did not wait before install it forget what i say here ! maybe your cpu is damaged i dont know !?

 

 But if you waited silicone dry before install it , the problem may be too much space between the die and the plate,

when you use silicone rtv between 2 metal parts, like for a cover valve of a car and you tight all 10-14 bolts there is still a good thickness of silicone between parts, so if you let it dry with no pressure or not enough, you may have too much gap between the die and the plate, and it may definitively be a good cause of bad heat dissipation of your cpu , also be sure you remove carefully and completely the old silicone, you want the die and the back plate as near as possible

 

when i did mine i did not wait to install the cpu and the cooler on the mobo, to have a pressure on it before the silicone dry,  but it is little tricky cause the cpu plate slide a bit (cause silicone is not dry yet ) when you lock the cpu with the cooler retainer, so i had to install the cpu back plate very little offset over the die (about 1mm aprox )so it sit at the right place when it slide a bit when you install the cooler, and i use very little of silicone,   that is what i would do with your i7 for the 7th try..  you are very patient sir ! 

 

hope my English is good enough you understand me, wish you good luck to enjoy low temps on your rig

 

During one of my attempts, I figured the silicone might have been a problem and tried a delid attempt using no silicone, just the IHS over the die, and that was the hottest attempt. My temps went up to about 90c. I think the floating of the IHS on the chip fouled up the liquid metal, creating air bubbles and what not, so I figured I really need to some kind of adhesive holding the plate in 1 place so it can't swim around. 

[Lemon]

GridiroN

you said you put liquid metal on ihs and die, i only used very thin coat on my die , you should already know it but just in case the paste or liquid only purpose is to fill surface imperfections, the ihs and die must touch each other as much as possible, too much paste do not help for heat dissipation .

also, i did delid mine with a  blade (the very thin ones from pharmacy, btw) so i don't know much about those delid tool, but what i saw from the videos the tools do a pretty high tension on the side of IHS, is it possible after 6th times the IHS can be little convex now and there is a gap between ihs and die ? 

if you d want to check it, you could use cheaper thermal paste or silicone rtv can do the job too for the test . just put a  very very little amount on the middle of the die, i d say no more than a 1mm - 2mm wide dot of paste or silicone,  and install it   on your mobo  and take it out right after and check how to paste expand on the die, you ll have a good visual (spread area) of surface contact. maybe there is better method to do i dont know, i m just trying to help you, there is a heat conductivity problem somewhere between the die and the heat sink fan...