Delidding troubleshooting.

[AverageTechGuy]

so i have an issue here and this forum is the only place i trust the answers from. So i delidded my i7 7600k and replaced the old thermal compund with artic silver 5. Which i read this to be one of the best compounds. Now my cpu is hitting thermal throttling whenever i run cinebench and im idling at like 65C and random spikes to like 80C without anything running. what could i have messed up to cause this issue? prior to this procedure i would never go above 80C under any circumstance. i was just trying to get better thermals for possible overclocking. Oh and the CPU is at 4.4GHz at the moment. 

[WoodenMarker]

What cooler are you using? Have you tried reseating it?

What's your core voltage set to?

[App4that]

Did you relid? Are you sure the cooler is seated properly?

[JPStone]

If you applied to much AS5 compound it could have insulated the die and caused the opposite effect as the the intended purpose of transferring heat away from the CPU. I use AS% myself and its a heavy compound so a little goes a long way. I've never delided and replace compound but I know this can happen even if you apply to much onto the heat-spreader so maybe, if you can do so safely without damaging the Die, try and remove compound and try putting less. Then again it could be you didn't put enough. Tough one. Good luck.

[AverageTechGuy]

honestly i was thinking the same thing. i may have put too much AS5 on. this was my pretty much first time dealing with it and it is a lot runnier than other compounds i have used. so i think that will have to be my first step. i do appreciate the quick feedback. and i also fooped the model number of the cpu. its a 6700k not a 7600k. although i dont think it matters much in this case. Also, to answer everyone else and to possibly trigger a different cause, my voltage is set to auto. usually is around 1.24 volts or something around there. and my cooler is a thermaltake 240 riing AIO, and yes i resat it to no avail. Again i appreciate the super quick response. really helped alot. i will get back to you guys/gals tomorrow with the results of me redoing the AS5 on the die. 

[MageTank]

14 minutes ago, AverageTechGuy said:

honestly i was thinking the same thing. i may have put too much AS5 on. this was my pretty much first time dealing with it and it is a lot runnier than other compounds i have used. so i think that will have to be my first step. i do appreciate the quick feedback. and i also fooped the model number of the cpu. its a 6700k not a 7600k. although i dont think it matters much in this case. Also, to answer everyone else and to possibly trigger a different cause, my voltage is set to auto. usually is around 1.24 volts or something around there. and my cooler is a thermaltake 240 riing AIO, and yes i resat it to no avail. Again i appreciate the super quick response. really helped alot. i will get back to you guys/gals tomorrow with the results of me redoing the AS5 on the die. 

Since you are using an AIO, try tilting the radiator to make sure the pump isn't airlocked. If you didn't relid using sealant, make sure the IHS did not move when you lowered the retention arm on the socket. Also, do be careful using AS5 for compound on your die. While it's good paste for use in between the IHS and heatsink, it's not designed to handle heavy thermal cycling on a bare die, where temperature shifts are extremely dramatic. In my testing, it pumps out pretty fast.

 

Standard paste that works well on the die in my personal tests were Gelid, Kryonaut, Phobya Nanogrease and KPx. You can also use Shin Etsu or Dow Corning, which is basically the stock paste used on GPU's and CPU's, as they are actually designed with durability in mind. You might be hesitant to put the same thermal compound that you just removed back on, but believe me when I tell you the excessive silicone adhesive did far more to hurt your temperatures than Intel's choice of paste. I've seen a solid 5C drop just by repasting with TC-1996.

[AverageTechGuy]

What would your opinion be on thermaltakes thermal paste? I think I have the tg-7 or something like that. Would that be ok to use on the die? And I will try the radiator thing first to make sure. And I also did reseal using red rtv as recommended in other videos including Linus's 

[MageTank]

1 minute ago, AverageTechGuy said:

What would your opinion be on thermaltakes thermal paste? I think I have the tg-7 or something like that. Would that be ok to use on the die? And I will try the radiator thing first to make sure. And I also did reseal using red rtv as recommended in other videos including Linus's 

I have never used thermaltakes thermal compound, so I cannot attest to it's durability. The main thing to look out for is the paste turning into a chalky dust and cracking apart. If you are not resealing the CPU, I'd say go for it, and if you notice thermals getting worse over time, investigate the paste. While this article is old, it's certainly a good read and is applicable today: https://www.electronics-cooling.com/2007/11/reliability-testing-of-thermal-greases/

 

Most pastes are designed to perform well, and to be budget friendly while doing so, but not every paste is advisable to use on the die. It's just something to keep in mind if you run into further issues in the future. I certainly do not suspect your choice in paste to be causing you any issues at the moment, as it would take far longer for the effects of pumpout or drying to take place, and I've certainly never seen a paste dry up in such a short span of time.

 

So if it were me in your current situation, I would start with rocking the radiator to make sure no air found it's way to the pump, as that would be far easier than pulling the block off and removing the IHS again, and if that does not solve things, investigate IHS contact with the die, and visual characteristics of the paste (did it make proper contact, did it get good coverage on the die, is it still in a malleable form, etc). 

[AverageTechGuy]

Thank you so much for the info. Really helps a lot. As I stated earlier I will try everything Tonight and post back tomorrow to let you all know whether or not it's fixed. 

[AverageTechGuy]

Ok so after 2 more failed attempts at fixing my temp issues I finally figured it out. Sure enough it was the thermal paste j was using. I applied some of the other paste I had and my temps never go above 65c on a stress test with a 4.5ghz overclock. Thank you guys once again for helping me with the issue. And a huge thanks to magetank for bringing up the possibility. 

[AngryBeaver]

19 hours ago, AverageTechGuy said:

so i have an issue here and this forum is the only place i trust the answers from. So i delidded my i7 7600k and replaced the old thermal compund with artic silver 5. Which i read this to be one of the best compounds. Now my cpu is hitting thermal throttling whenever i run cinebench and im idling at like 65C and random spikes to like 80C without anything running. what could i have messed up to cause this issue? prior to this procedure i would never go above 80C under any circumstance. i was just trying to get better thermals for possible overclocking. Oh and the CPU is at 4.4GHz at the moment. 

Also delidding your CPU and NOT going back with liquid metal is pretty much pointless. Yes you might see minimal gains if done correctly, but the risk isn't worth the rewards unless you are using a liquid metal solution.

[MageTank]

12 minutes ago, AngryBeaver said:

Also delidding your CPU and NOT going back with liquid metal is pretty much pointless. Yes you might see minimal gains if done correctly, but the risk isn't worth the rewards unless you are using a liquid metal solution.

This is not true, not in the slightest. Traditional paste works perfectly fine when delidding. It has never been Intel's choice of paste that caused high thermals, but rather the resealing process which uses abundant silicone adhesives, causing poor contact between the die and IHS. I still saw a 5C improvement using the exact same TC-1996 Dow Corning paste that Intel uses when delidding my CPU. Granted, the thermal benefits are not as dramatic as liquid metal, the benefits still exist nonetheless.

 

Another huge advantage of delidding is per-core uniformity, where you are able to lower the difference in temperatures between your hottest and coldest cores. I've seen chips that would have a 15C difference in temperatures between the hottest and coldest cores. This was especially true for Haswell V1, which saw some of the worst differences in temperatures.

 

Risk vs reward is a completely subjective concept. What might seem risky to you, might not be risky to me. What might be "worth it" for you, might not be worth it for me. Liquid metal imposes more of a risk, as it requires more precautions before going through with the application, but yields a higher reward as a result. Traditional paste requires none of the prep work, is cheaper, and is readily available in most tech stores. If one is in need of an immediate boost in thermals, I see absolutely nothing wrong with delidding and using traditional paste. 

[AngryBeaver]

7 minutes ago, MageTank said:

This is not true, not in the slightest. Traditional paste works perfectly fine when delidding. It has never been Intel's choice of paste that caused high thermals, but rather the resealing process which uses abundant silicone adhesives, causing poor contact between the die and IHS. I still saw a 5C improvement using the exact same TC-1996 Dow Corning paste that Intel uses when delidding my CPU. Granted, the thermal benefits are not as dramatic as liquid metal, the benefits still exist nonetheless.

 

Another huge advantage of delidding is per-core uniformity, where you are able to lower the difference in temperatures between your hottest and coldest cores. I've seen chips that would have a 15C difference in temperatures between the hottest and coldest cores. This was especially true for Haswell V1, which saw some of the worst differences in temperatures.

 

Risk vs reward is a completely subjective concept. What might seem risky to you, might not be risky to me. What might be "worth it" for you, might not be worth it for me. Liquid metal imposes more of a risk, as it requires more precautions before going through with the application, but yields a higher reward as a result. Traditional paste requires none of the prep work, is cheaper, and is readily available in most tech stores. If one is in need of an immediate boost in thermals, I see absolutely nothing wrong with delidding and using traditional paste. 

If you read my post I state that when done correct you can see some minimal gains. 5c is a minimal gain. If you were to look at the gains with most LM applications you would see they are in the 15-25C range (depending on chip).

 

Liquid metal isn't hard to get either, most order it online and have it 2 days later. The application process isn't THAT much more risky. I mean you are correct in the sense that it is conductive, but if you follow proper precautions by using either nail polish or liquid electrical tape... then that risk shrinks drastically. Then there are also methods using just plain tape that make the application process easier as well.

 

Delidding a chip isn't as risky as it use to be, but it is still a risk and the killing of your warranty. So if you are going to take those risks then using a paste over LM doesn't make much sense (outside of Ln2 cooling).

[App4that]

3 minutes ago, AngryBeaver said:

If you read my post I state that when done correct you can see some minimal gains. 5c is a minimal gain. If you were to look at the gains with most LM applications you would see they are in the 15-25C range (depending on chip).

 

Liquid metal isn't hard to get either, most order it online and have it 2 days later. The application process isn't THAT much more risky. I mean you are correct in the sense that it is conductive, but if you follow proper precautions by using either nail polish or liquid electrical tape... then that risk shrinks drastically. Then there are also methods using just plain tape that make the application process easier as well.

 

Delidding a chip isn't as risky as it use to be, but it is still a risk and the killing of your warranty. So if you are going to take those risks then using a paste over LM doesn't make much sense (outside of Ln2 cooling).

90% of the gains from delidding are getting rid of the gap made by the adesive which holds the IHS in place, like Mage said. We use LM because we're rebels who don't care about risks. Man. 

 

You also don't void your warranty by delidding, if you send a lidded CPU back. See rebel. 

[MageTank]

2 minutes ago, AngryBeaver said:

If you read my post I state that when done correct you can see some minimal gains. 5c is a minimal gain. If you were to look at the gains with most LM applications you would see they are in the 15-25C range (depending on chip).

 

Liquid metal isn't hard to get either, most order it online and have it 2 days later. The application process isn't THAT much more risky. I mean you are correct in the sense that it is conductive, but if you follow proper precautions by using either nail polish or liquid electrical tape... then that risk shrinks drastically. Then there are also methods using just plain tape that make the application process easier as well.

 

Delidding a chip isn't as risky as it use to be, but it is still a risk and the killing of your warranty. So if you are going to take those risks then using a paste over LM doesn't make much sense (outside of Ln2 cooling).

5C is not minimal by any means. 5C is quite substantial for most overclockers. As for your liquid metal argument, I live in the states and have yet to find a place that ships it in 2 days. CLU is not an Amazon Prime or Newegg Premier item, nor is Conductonaut. I've delidded a ton of CPU's, and have gone through countless tubes of liquid metal, so it's extremely safe to assume I've gone through this order process several times over. My local Micro Center is 10 minutes away, and has Kryonaut on the shelf. For a same day solution, there is absolutely nothing wrong with Kryonaut, which in my tests, still yielded an 8-10C drop in temperatures on average.

 

The warranty aspect is also extremely situational. Sure, it technically voids warranty to delid your CPU, but so does using aftermarket coolers, any overclocking (both CPU and even out-of-spec XMP profiles on memory) and even then, you can still get warranty service if you put it back together properly. @Lays killed a chip under LN2 and had liquid metal on it. The underside of the IHS was stained, he still replaced it with standard paste (not even Dow Corning or Shin Etsu paste) and glued it back together, got his replacement without a single issue. Obviously this is going to be a case by case thing, and be dependent entirely on who reviews the process, but it's still possible to get it serviced.

 

My point is, both traditional paste and LM work fine for delidding. Both are superior solution than leaving your chip stock with that excessive adhesive under the die. Regardless of precautions, liquid metal is still more risky in the grand scheme of things. If you spill non-conductive traditional paste on a PCB, it won't bond to any contacts and risk shorting something out. If you spell liquid metal, you need to immediately clean it and hope that none of the gallium that you wiped off, found its way elsewhere.  Liquid metal requires precaution and precision, whereas traditional paste is a "set and forget" procedure (assuming you chose decent paste that can handle intense thermal cycling). 

 

I don't have anything against liquid metal, have been using it on my die for a few years now and it's my go-to solution for every CPU I own, even the locked ones. I just don't want to see people telling others that it's pointless to delid without using liquid metal, because that's simply not true.

[AngryBeaver]

9 minutes ago, MageTank said:

5C is not minimal by any means. 5C is quite substantial for most overclockers. As for your liquid metal argument, I live in the states and have yet to find a place that ships it in 2 days. CLU is not an Amazon Prime or Newegg Premier item, nor is Conductonaut. I've delidded a ton of CPU's, and have gone through countless tubes of liquid metal, so it's extremely safe to assume I've gone through this order process several times over. My local Micro Center is 10 minutes away, and has Kryonaut on the shelf. For a same day solution, there is absolutely nothing wrong with Kryonaut, which in my tests, still yielded an 8-10C drop in temperatures on average.

 

The warranty aspect is also extremely situational. Sure, it technically voids warranty to delid your CPU, but so does using aftermarket coolers, any overclocking (both CPU and even out-of-spec XMP profiles on memory) and even then, you can still get warranty service if you put it back together properly. @Lays killed a chip under LN2 and had liquid metal on it. The underside of the IHS was stained, he still replaced it with standard paste (not even Dow Corning or Shin Etsu paste) and glued it back together, got his replacement without a single issue. Obviously this is going to be a case by case thing, and be dependent entirely on who reviews the process, but it's still possible to get it serviced.

 

My point is, both traditional paste and LM work fine for delidding. Both are superior solution than leaving your chip stock with that excessive adhesive under the die. Regardless of precautions, liquid metal is still more risky in the grand scheme of things. If you spill non-conductive traditional paste on a PCB, it won't bond to any contacts and risk shorting something out. If you spell liquid metal, you need to immediately clean it and hope that none of the gallium that you wiped off, found its way elsewhere.  Liquid metal requires precaution and precision, whereas traditional paste is a "set and forget" procedure (assuming you chose decent paste that can handle intense thermal cycling). 

 

I don't have anything against liquid metal, have been using it on my die for a few years now and it's my go-to solution for every CPU I own, even the locked ones. I just don't want to see people telling others that it's pointless to delid without using liquid metal, because that's simply not true.

First off, LM when used correctly does not pump out or dry out. It also handles the drastic temp shifts of the DIE much better than any paste.

 

It also being a liquid will allow for a much thinner application than any paste. So that "gap" is going to be less of a concern. You just cannot get as thin of an application out of a paste as you can LM and as we all know paste is way less conductive of heat than even the IHS. Now if we were talking about solutions for application between the IHS and Block then you would have a great argument. When it comes to ln2 solutions you also would have a good argument. That being said I have purchased LM off amazon with prime shipping in the past. Really just depends on stock levels.

 

Let me also clarify in that I am not saying it doesn't work fine for delidding. I am also not saying it wouldn't be beneficial over a stock setup. I am merely saying that if you are willing to take the risk and time to delidd, then you are basically shooting yourself in the foot by not going that extra step with LM. Yes, 5c can help stabilize an OC, but a 15+ c drop can take you another step in an overclock sometimes. Plus if you are running a high end cooling solution you want the heat to transfer from DIE to your solution as quickly and efficiently as possible.

[App4that]

10 minutes ago, AngryBeaver said:

First off, LM when used correctly does not pump out or dry out. It also handles the drastic temp shifts of the DIE much better than any paste.

 

It also being a liquid will allow for a much thinner application than any paste. So that "gap" is going to be less of a concern. You just cannot get as thin of an application out of a paste as you can LM and as we all know paste is way less conductive of heat than even the IHS. Now if we were talking about solutions for application between the IHS and Block then you would have a great argument. When it comes to ln2 solutions you also would have a good argument. That being said I have purchased LM off amazon with prime shipping in the past. Really just depends on stock levels.

 

Let me also clarify in that I am not saying it doesn't work fine for delidding. I am also not saying it wouldn't be beneficial over a stock setup. I am merely saying that if you are willing to take the risk and time to delidd, then you are basically shooting yourself in the foot by not going that extra step with LM. Yes, 5c can help stabilize an OC, but a 15+ c drop can take you another step in an overclock sometimes. Plus if you are running a high end cooling solution you want the heat to transfer from DIE to your solution as quickly and efficiently as possible.

I think the issue here is unless you go direct die, you're using TIM anyway. I've got LM on my die to IHS, and paste between the IHS and block as you can't use LM between the IHS and block.

 

See the issue? You're using paste regardless, unless you go direct die. The LM doen't have the largest effect. The delidding does. 

[MageTank]

Just now, AngryBeaver said:

First off, LM when used correctly does not pump out or dry out. It also handles the drastic temp shifts of the DIE much better than any paste.

 

It also being a liquid will allow for a much thinner application than any paste. So that "gap" is going to be less of a concern. You just cannot get as thin of an application out of a paste as you can LM and as we all know paste is way less conductive of heat than even the IHS. Now if we were talking about solutions for application between the IHS and Block then you would have a great argument. When it comes to ln2 solutions you also would have a good argument. That being said I have purchased LM off amazon with prime shipping in the past. Really just depends on stock levels.

 

Let me also clarify in that I am not saying it doesn't work fine for delidding. I am also not saying it wouldn't be beneficial over a stock setup. I am merely saying that if you are willing to take the risk and time to delidd, then you are basically shooting yourself in the foot by not going that extra step with LM. Yes, 5c can help stabilize an OC, but a 15+ c drop can take you another step in an overclock sometimes. Plus if you are running a high end cooling solution you want the heat to transfer from DIE to your solution as quickly and efficiently as possible.

I don't recall ever saying LM pumps out over time, but it can certainly dry out depending on a few circumstances. I've tested both CLU and conductonaut on laptops with uneven surfaces for their heatsinks, and have personally seen the liquid metal turn into a sandlike texture. I even have pictures of what that looks like from my brothers XPS 13:

Spoiler

 

As for the discussion on thermal conductivity of paste, it has less to do with raw conductivity, and more to do with proper contact and mounting pressure. The silicone adhesive being too thick prevented proper contact of the paste between the die and IHS, which is why you see such a dramatic improvement when just removing the glue, and using the exact same paste. If thermal conductivity was the be all, end all example as to what makes paste better, we'd see a far more dramatic difference when using the liquid metal between the IHS and cooler. As you've already stated, it's not really applicable in that situation, as it makes a minor improvement at best. The reason liquid metal shines the most on the die, is due to the fact that the die itself gets much hotter than the surface of the IHS, and being able to transfer that heat quicker, aids in lowering the thermals as a whole. By the time the IHS transfers most of the heat away from the die, the gains are minimal. 

 

There have been many tests trying to demonstrate the importance of mounting pressure and contact over raw thermal conductivity, and Innovation Cooling has actually ran a few articles on the subject after releasing their graphite pad (which had very mixed results depending on the application of the pad): https://www.innovationcooling.com/case-study-contact-pressure-related-thermal-compound-performance/

 

Here is another test performed by Kimandsally: https://www.innovationcooling.com/initial-desktop-tests-ic-graphite-vs-gelid-gc-extreme/

From this test alone, he saw a 3C difference just by adjusting only mounting pressure.

 

Steve from Gamers Nexus also saw a similar situation when resealing his IHS after using liquid metal: https://www.gamersnexus.net/guides/3105-vrm-overtemperature-sealed-vs-unsealed-intel-delidding

 

By resealing the IHS, you are altering the mounting pressure against the die, and depending on how uneven your seal is, the contact. This can impact liquid metal just as poorly as it impacted the TC-1996 that ships on these CPU's. 

 

I certainly agree that if one is planning on delidding, that buying liquid metal in preparation for the delid will yield the best results, I still can understand why someone would rather just use normal paste. It's cheaper in the short term (reason being, you are going to have to replace traditional paste more often than what you would liquid metal) and easily accessible in local stores. The counter argument is, if you are worried about risk, you are likely already ordering a delid tool, and if you are ordering a tool online, waiting for liquid metal to be shipped is not going to hurt. Still, to each their own I suppose.

[AverageTechGuy]

Ok to put everyone to rest here, just by switching to a Better thermal paste and NOT resealing the heat spreader I dropped 20C on average. So yes the paste and sealing method does Matter

[tikker]

32 minutes ago, MageTank said:

I don't recall ever saying LM pumps out over time, but it can certainly dry out depending on a few circumstances. I've tested both CLU and conductonaut on laptops with uneven surfaces for their heatsinks, and have personally seen the liquid metal turn into a sandlike texture. I even have pictures of what that looks like from my brothers XPS 13

I think this is more the LM reacting with the copper to form an alloy rather than the stuff drying out like regular paste though.

[MageTank]

2 minutes ago, tikker said:

I think this is more the LM reacting with the copper to form an alloy rather than the stuff drying out like regular paste though.

I would agree with that, as I have seen exactly what you are referring to on my CPU's IHS, but on this laptops die and heatsink, it took less than 15 days for all of the LM to completely turn into a sandlike texture, and completely ruin thermal performance.  When I eventually got it cleaned, I took a blade to the heatsink and found that the surface was extremely uneven. Uneven surfaces plus poor mounting pressure equals a very bad time, lol.

[AngryBeaver]

2 hours ago, MageTank said:

I would agree with that, as I have seen exactly what you are referring to on my CPU's IHS, but on this laptops die and heatsink, it took less than 15 days for all of the LM to completely turn into a sandlike texture, and completely ruin thermal performance.  When I eventually got it cleaned, I took a blade to the heatsink and found that the surface was extremely uneven. Uneven surfaces plus poor mounting pressure equals a very bad time, lol.

If you reapply the LM after wiping off the powdery leftovers and then reapply without trying to remove the discolored stuff... then it won't have the issue on reapplication. It is actually filling the micro gaps so well that it binds to the copper and it basically causes what you have seen

 

Also to the comment above, yes you do pretty much always use paste between IHS and block, but there is a big reason why this isn't a problem. It comes down to the surface area that is in contact.  The DIE on the other hand is much smaller and they are going to continue to get even more so as manufacturing processes continue to advance and improve.  So the drastic heat changes on the DIE, plus the extremely thin coating it needs, plus how well it fills the little imperfections, and lastly the much higher heat transfer rate... mean that it has huge advantages over paste when used on such a tiny amount of surface area.

[WoodenMarker]

5 hours ago, AngryBeaver said:

Yes, 5c can help stabilize an OC, but a 15+ c drop can take you another step in an overclock sometimes.

5c shows a bigger difference in cooling at lower temps than higher ones. Depending on the tests, both a 5c difference and 15c difference might represent a similar improvement in cooling.

4 hours ago, MageTank said:

There have been many tests trying to demonstrate the importance of mounting pressure and contact over raw thermal conductivity

The Fuma is an example that shows the importance of mounting pressure: http://www.overclockers.com/scythe-fuma-heatsink-review/

The benchmarks and other forum posts show 10c+ difference between low and high pressure mounting. Scythe had to update the cooler with a kit of screw stops to prevent users from killing their Skylake cpus with it.

[MageTank]

5 minutes ago, WoodenMarker said:

5c shows a bigger difference in cooling at lower temps than higher ones. Depending on the tests, both a 5c difference and 15c difference might represent a similar improvement in cooling.

The Fuma is an example that shows the importance of mounting pressure: http://www.overclockers.com/scythe-fuma-heatsink-review/

The benchmarks and other forum posts show 10c+ difference between low and high pressure mounting. Scythe had to update the cooler with a kit of screw stops to prevent users from killing their Skylake cpus with it.

I remember a ton of people removing C clips from their laptop heatsinks to improve mounting pressure. The results are just as dramatic, with a nice 5C difference. I eventually need to do some additional testing on pressure and contact with various coolers. I might also re-run the tests with a lapped IHS and cooler just to see if the same rules apply with a much flatter surface.