Extremely Quick Thermal Paste Degradation, 15°C Increase - Bad Batch?

[lexluthermiester]

shaite... are you replacing TIM on a running machine? The LED are glowing... it is on...

Right? That's what I was just thinking.

@Wet_Paint
Do you always apply thermal compound while the system is running? Thermal throttling or not, the temp spikes are going to hurt that CPU die. That might be your problem right there. Always, and I mean ALWAYS, unplug the PSU before working on a system. Better safe than sorry.

 

[crazyeyesreaper]

Looking at the TIM spread its not awful however the bottom right could be better. It seems that somewhat knackered screw is resulting in less than optimal pressure on that side. Pressure means a great deal in terms of heat transfer. A good example is the Scythe Fuma cooler its a dual tower dual fan heatsink that was about $50 but was nearly in the same league as the Noctua NH-D15 and 240mm AIO coolers and that is because it had a much higher mounting pressure. Enough so it could cause the thinner PCBs on the newer CPUs from Intel to bend which was obviously a bad thing. However it did result in a cheaper cooler offering far more performance than it had any right to. So the issue could be a mounting pressure problem if the screws are not seating properly. Try turning the pump (change its mounting orientation and see if the problem screw goes all the way in on a different hole if not you may want to contact Alphacool about the mounting hardware. Another option is to use your hand and apply pressure on the problematic area by pushing on the bracket (not the screw) you can apply greater pressure and see if it makes a difference. At this point without physically being there I can't really help you any further than that. However I will note i have an AIO here from a different vendor different mounting hardware same problem. Performance was not what it should be. Traced it to bad threads took it apart and carefully got the hardware to fully seat outside the case. Then reintalled it. With full pressure on all 4 corners performance improved by about 3C under Aida64 FPU load (extreme torture test).

I repasted it with the alphacool paste which came in the box with the cooler, and did some test runs in Cinebench. Ran each test 3 times and waited 20 minutes after boot.

Hydronaut: 89-93°C, 91°C average, 30.0°C ambient
Alphacool #1: 88-91°C, 89°C average, 30.6°C ambient
Alphacool #2: 89-91°C, 90°C average, 31.6°C ambient





Hope these pictures are good enough.
View attachment 104608View attachment 104609View attachment 104607


Edit: took a picture of the mounting screws, the remaining two look exactly like the one on the left, but the one on the right is impossible to "screw in fully", ive tried and even with a lot of pressure i couldnt move it even little bit, actually i stripped the screw(just a little bit, still can unscrew/screw it as usually but it has some wiggle room).
View attachment 104610

 

[Wet_Paint]

Well first of all these Skylake-X CPU will imply run very hot due to their nature and the fact that they aren't soldered , that 4.7 Ghz overclock doesn't help your cause either.

That been said I have seen the pictures you posted with the TIM applied and nothing looks out of place to me. A blob in the middle is all you need to do , you can spread it however you want in a million different ways it will make no difference , this has been proven time and time again. To me it sounds like there is something wrong with the surface of the block or maybe even the IHS. Maybe the block isn't fully in place.



I think you have your problem right there.



Don't , it's enough.

I read the post after i repasted it, gave it tiny bit more and made the blop more round, now im out of alphacool paste and ive only one "re-paste" with hydronaut left.

While re-pasting it i checked all the nuts on the mounting screws and some of them got loose, tightened them up and all 4 mounting screws are now properly screwed in. Temperatures are still identical, zero difference outside of margin of error (-1°C avg).

Right? That's what I was just thinking.

@Wet_Paint
Do you always apply thermal compound while the system is running? Thermal throttling or not, the temp spikes are going to hurt that CPU die. That might be your problem right there. Always, and I mean ALWAYS, unplug the PSU before working on a system. Better safe than sorry.

Like i said the LEDs run for some time after the PSU is unplugged, I can even attempt to boot without PSU plugged in, fans start to spin up but quickly drain the PSU capacitors and turn off in about 1 second before POST completes.

Looking at the TIM spread its not awful however the bottom right could be better. It seems that somewhat knackered screw is resulting in less than optimal pressure on that side. Pressure means a great deal in terms of heat transfer. A good example is the Scythe Fuma cooler its a dual tower dual fan heatsink that was about $50 but was nearly in the same league as the Noctua NH-D15 and 240mm AIO coolers and that is because it had a much higher mounting pressure. Enough so it could cause the thinner PCBs on the newer CPUs from Intel to bend which was obviously a bad thing. However it did result in a cheaper cooler offering far more performance than it had any right to. So the issue could be a mounting pressure problem if the screws are not seating properly. Try turning the pump (change its mounting orientation and see if the problem screw goes all the way in on a different hole if not you may want to contact Alphacool about the mounting hardware. Another option is to use your hand and apply pressure on the problematic area by pushing on the bracket (not the screw) you can apply greater pressure and see if it makes a difference. At this point without physically being there I can't really help you any further than that. However I will note i have an AIO here from a different vendor different mounting hardware same problem. Performance was not what it should be. Traced it to bad threads took it apart and carefully got the hardware to fully seat outside the case. Then reintalled it. With full pressure on all 4 corners performance improved by about 3C under Aida64 FPU load (extreme torture test).

Even tho the nut wasnt properly tightened the spring on the other side gave it enough force. The only thing it improved were random temperature spikes when idle, which went down by couple degrees.

 

[John Naylor]

1. You are experiencing a problem common to all CLCs in that the rated pump flow is on;y a fcartion of a gpm.... for custom loops, the minimum is considered 1,0 to 1.25 ... more doesn't help but less does. And that rated 0.3 is w/o all the extra tubing, fittings for the GPU so it's far less than the rated flow.

2. Marketing types love to play with numbers. When a package reads 0.30 gpm / 2.0 psi that doesn't mean what it says ... it means it can produce 0.30 gpm @ 0 plsi and 0 gpm at 2 psi in marketing speak. So the reality is you are likely down around 0.1 to 0.15 gpm at best.

3.a If I understood correctly, you have a Open Loop type AIO with a GFX card cooler designed to be connected together and using 2 pumps. I don't understand where / how the NH-D15 fits in. What type of bleed mechanism did you employ and what procedure did you employ to remove air and "off gasses" ? Without doing so, there will be air trapped in the system which will inhibit flow. In a custom loop, the best method is to install a rider on one of the spare rad ports on the top side of the top rad. If the rad has no spre ports, one mist install a tee at the highest point in the loop with the branch side pointing up. Here one can attach a valve and a rigid acrylic tube with valve on top. Tilt the system as necessary to have the tee at the highest point, fill the tube 90% of the way and close the valve . Typically as the fluid traverses the loop, air / gas will come out of solution or any bubbles that would have formed will collect on the tube and you will have a visible drop in the water level in the tube. Shut down the system, open the valve and top off as necessary, rinse and repeat. Not that any high points in the system may have trapped air.... you will need to work these out by bending tubing or tilting the system.

3.b Not using the 2nd pump presents a significant flow restrictiion as pump 1 is being forced to spin the 2nd's pump and motor

4. While not optimal, there's nothing terribly wrong with your application method. However, introducing finger oils to the TIM is a bad idea. You can use a plastic glove ot some TIMs (i.e. Gelid) come with an application "spatula", However once spread around (smoothly spread method came in 2nd place in 2nd link) , use a plastic credit or hotel lock card to scrape off the excess, leaving behind only the stuff that is left in nooks and crannies. This method is often used on HDT type coolers after the two line application method is used. What method you use depends upon the cooler / block type.

https://archive.benchmarkreviews.com/index.php?option=com_content&task=view&id=170&Itemid=4
https://www.pugetsystems.com/labs/articles/Thermal-Paste-Application-Techniques-170/

5. Note that many blocks are designed to prevent overtightening by including a "stop" mechanism ... you screw them down till you can't go any more w/o abnormal pressure. Try and go beyond this point and you will strip or break the screw. It also allows for absolutely even contact pressure.

6. Other than Ivy Bridge, I have not encountered an Intel CPU that requires delidding to reach the highest OC using custom water cooling or OLC type AIOs.

I don't see this being resolved w/o addressing the problems above . I know money is tight, but ultimately there are a whole lot of deficiencies to overcome. While I applaud AC for producing a AIO w/ a copper rad, going cheap on the pump was abad idea. Yes it saves money but that is immediately offset when you have to add a pump for the GFX card. AFAIK, Swiftech and EK are making the only AIOs that both provide all copper systems (not sure about the GFX system) and provide a pump with the necessary flow capacity to adequately do water cooling. The EK system is the only one that doesn't (theoretically) require bleeding.

 

[Wet_Paint]

1. You are experiencing a problem common to all CLCs in that the rated pump flow is on;y a fcartion of a gpm.... for custom loops, the minimum is considered 1,0 to 1.25 ... more doesn't help but less does. And that rated 0.3 is w/o all the extra tubing, fittings for the GPU so it's far less than the rated flow.

2. Marketing types love to play with numbers. When a package reads 0.30 gpm / 2.0 psi that doesn't mean what it says ... it means it can produce 0.30 gpm @ 0 plsi and 0 gpm at 2 psi in marketing speak. So the reality is you are likely down around 0.1 to 0.15 gpm at best.

3.a If I understood correctly, you have a Open Loop type AIO with a GFX card cooler designed to be connected together and using 2 pumps. I don't understand where / how the NH-D15 fits in. What type of bleed mechanism did you employ and what procedure did you employ to remove air and "off gasses" ? Without doing so, there will be air trapped in the system which will inhibit flow. In a custom loop, the best method is to install a rider on one of the spare rad ports on the top side of the top rad. If the rad has no spre ports, one mist install a tee at the highest point in the loop with the branch side pointing up. Here one can attach a valve and a rigid acrylic tube with valve on top. Tilt the system as necessary to have the tee at the highest point, fill the tube 90% of the way and close the valve . Typically as the fluid traverses the loop, air / gas will come out of solution or any bubbles that would have formed will collect on the tube and you will have a visible drop in the water level in the tube. Shut down the system, open the valve and top off as necessary, rinse and repeat. Not that any high points in the system may have trapped air.... you will need to work these out by bending tubing or tilting the system.

3.b Not using the 2nd pump presents a significant flow restrictiion as pump 1 is being forced to spin the 2nd's pump and motor

4. While not optimal, there's nothing terribly wrong with your application method. However, introducing finger oils to the TIM is a bad idea. You can use a plastic glove ot some TIMs (i.e. Gelid) come with an application "spatula", However once spread around (smoothly spread method came in 2nd place in 2nd link) , use a plastic credit or hotel lock card to scrape off the excess, leaving behind only the stuff that is left in nooks and crannies. This method is often used on HDT type coolers after the two line application method is used. What method you use depends upon the cooler / block type.

https://archive.benchmarkreviews.com/index.php?option=com_content&task=view&id=170&Itemid=4
https://www.pugetsystems.com/labs/articles/Thermal-Paste-Application-Techniques-170/

5. Note that many blocks are designed to prevent overtightening by including a "stop" mechanism ... you screw them down till you can't go any more w/o abnormal pressure. Try and go beyond this point and you will strip or break the screw. It also allows for absolutely even contact pressure.

6. Other than Ivy Bridge, I have not encountered an Intel CPU that requires delidding to reach the highest OC using custom water cooling or OLC type AIOs.

I don't see this being resolved w/o addressing the problems above . I know money is tight, but ultimately there are a whole lot of deficiencies to overcome. While I applaud AC for producing a AIO w/ a copper rad, going cheap on the pump was abad idea. Yes it saves money but that is immediately offset when you have to add a pump for the GFX card. AFAIK, Swiftech and EK are making the only AIOs that both provide all copper systems (not sure about the GFX system) and provide a pump with the necessary flow capacity to adequately do water cooling. The EK system is the only one that doesn't (theoretically) require bleeding.

3.a The NH-D15 is in completely different system. Its really difficult to bleed the system, so i usually do it in parts. I open the "plug" on CPU waterblock and only other "hole" which is on GPU Rad. This causes most of the liquid to shoot out of the lower positioned "hole". Then i just fill the rest with destilled water, liquid seems clean but there are occasionaly "oils spots" on top of the liquid if that makes any sense. When refilling the only way i can make sure its fully filled is when I run the whole system (in BIOS), lay it on its side, open the CPU waterblock hole and pour destilled water into it. There is no better way to refill this thing, but i dont mind. I sometimes do this multiple times, changing the position of the radiators and pump until there is no "watery" sound when wiggling them.

3.b Like i said, i tested it, tried all combinations CPU/GPU pump, X/GPU, CPU/X and all were completely identical.

4. interesting articles, ill definetly try the X method if get new thermal paste.

5. yes there are springs in each mounting screw which prevent overtightening

6. Given the right cooling the CPU can hit 5Ghz @ 1.32V without throttling (without delid) but i had to drop my ambient temperature to -5°C to achieve that.
I didnt buy this as "cheap" option, at the time there wasnt anything else avaible, only this or CLCs. Even now the avaibility of such parts is close to none. Literally only one shop has EKWB radiators and they are all sold out, and they have only two variants avaible, 280mm and 360mm. And SwiftTech doesnt sell anything at all in my country.

To be honest i guess ill just start saving up for new 280mm rad and pump while ill keep the waterblocks as they are.

 

[lexluthermiester]

Like i said the LEDs run for some time after the PSU is unplugged, I can even attempt to boot without PSU plugged in, fans start to spin up but quickly drain the PSU capacitors and turn off in about 1 second before POST completes.

Yeah saw that, you posted it right before I posted. Still after unplugging the PSU, press the system power button to drain all the capacitors in the system. You don't want that voltage leaking/shorting to other parts potentially causing damage.

 

[Durvelle27]

Your coverage is pretty pad in the center of the IHS which is what generates the most heat due to the CPU die being in the center
If pea size isn’t working for you, you could do a X with the TIM and ally even pressure on the cooler to spread the TIM

 

[Wet_Paint]

Your coverage is pretty pad in the center of the IHS which is what generates the most heat due to the CPU die being in the center
If pea size isn’t working for you, you could do a X with the TIM and ally even pressure on the cooler to spread the TIM

Did some research, Skylake-X chips arent "flat", the middle of the CPU is bit taller, kinda explains why the thermal paste just slides to the less tall part of the CPU.

Source (der8auer vid)


As for my problem there doesnt seem to be any other solution than just saving some money for better pump and more radiators to improve the temperature.

 

[sepheronx]

This is one of many reasons I avoid liquid cooling.

Please update us when/if you get a new pump and radiator please. Would like to know if it fixes your issue.

 

[eidairaman1]

This is one of many reasons I avoid liquid cooling.

Please update us when/if you get a new pump and radiator please. Would like to know if it fixes your issue.

Aios are what the pain is.

 

[Aquinus]

Aios are what the pain is.

Speak for yourself, my H100i v2 works fine.

 

[Wet_Paint]

UPDATE: So i managed to get official "expansion kit" for eisbaer for really cheap. It was pre-filled 280mm (30mm thick) radiator with fittings and quick-disconnect connectors. After i connected in into the loop the temperatures improved but only for the GPU (decrease by approx 5-8°C) but the CPU is about the same. I tried having only the CPU block and radiators in the loop but the results were identical.

I even tried two different orders of components in the loop:
CPU → 120mm → 280mm → GPU → 360mm →
and
CPU → 120mm → GPU → 280mm → 360mm →
but the temperatures were identical.

This is still with the "bad" hydronaut, there was more than i predicted so i was able to do full "X" on the CPU.

Also im considering getting XSPC Ion Pump/Reservoir in the future. Would that be fine? Im concerned that it might cause some issues since there would be two pumps in the loop with different flow rate (70 L/h and 420 L/h).

Do you guys have any other ideas what could cause this problem other than the thermal paste?

 

[micropage7]

have you checked the contact area between processor and the heatsink?

 

[Komshija]

Bad mount, defected cooler base, greasy CPU lid or greasy cooler base.

 

[Wet_Paint]

UPDATE: After replacing the thermal paste the temperatures didnt improve much. Later I flushed my loop and refilled it with distilled water, the temperatures were really good right after flushing, about 15°C drop (low 30s) at idle and 20-25°C drop (mid 80s) during load, the temperatures werent as good as when the loop was new but it wasnt that far off.

The next day my temperatures went to 50-60°C idle and +105°C under load. During that time i didnt even move the PC physically or made any alterations, just turned the PC off.

Ive also noticed abnormal CPU Pump RPM. According to the manufacturer the pump is supposed to run at 2600RPM. My pump is running at 3250-3350 RPM, before the latest flushing it was running at +- 2900RPM. My friend has a similiar loop with the same pump and he is getting about 2700RPM.
Is this a sign of dying pump?

 

[Vya Domus]

At his point I would try another cooler. Anything, doesn't matter just to see if the temperatures modify over time and then you can sort out that the cooler is indeed the problem and not the paste or whatever.

 

[Aquinus]

Are you certain that there isn't anything gunking up the blocks or radiator? It sounds like flow rate starts good but, gradually [quickly] gets worse.

The thing with it getting better immediately after flushing the liquid makes me think that there might be something obstructing liquid flow. If the RPM of the pump is increasing, it's possible that the impeller could be damaged but, that doesn't explain why it starts good and gets bad over time unless it's a combination of something gunked up and a damaged impeller. It's possible that flushing the loop might be un-gunking enough to operate properly but, only for it to get gunked back up after a little while.

I guess a question I would have is, what kind of coolant have you run in this loop? Has it only been filled with distilled water plus maybe some additive or have you used something like a colored coolant before? I've heard horror stories of colored coolant gunking up loops.

 

[cucker tarlson]

try disabling the specter and meltdown patches. seriously. download inspecter,disable then restart (not shut down-power on,restart).

my temps dropped by 10 degrees.

 

[Wet_Paint]

try disabling the specter and meltdown patches. seriously. download inspecter,disable then restart (not shut down-power on,restart).

my temps dropped by 10 degrees.

Ive disabled those months ago. (still disabled, just checked)

Are you certain that there isn't anything gunking up the blocks or radiator? It sounds like flow rate starts good but, gradually [quickly] gets worse.

The thing with it getting better immediately after flushing the liquid makes me think that there might be something obstructing liquid flow. If the RPM of the pump is increasing, it's possible that the impeller could be damaged but, that doesn't explain why it starts good and gets bad over time unless it's a combination of something gunked up and a damaged impeller. It's possible that flushing the loop might be un-gunking enough to operate properly but, only for it to get gunked back up after a little while.

I guess a question I would have is, what kind of coolant have you run in this loop? Has it only been filled with distilled water plus maybe some additive or have you used something like a colored coolant before? I've heard horror stories of colored coolant gunking up loops.

Yesterday I spent about 4 hours total flushing each radiator 3 times (the 360mm one 5times). The liquid always came out clear with nothing in it except for the initial bleed of the loop, which was little bit white and with smaller-than-sand-sized copper "bits".
I should also mention that some of the "quick disconnects" are unable to be disconnected anymore, no idea why, i hope it doesnt mean that they are corroded by algae.

For first year the loop was used only with the pre-filled unknown liquid. After it started making wierd noises (similar to car being unable to start) and the liquid level decreasing i just kept topping it off with distilled water without any additives. Currently its on only distilled water with nothing in it.

When playing the temperatured seemed stable between 60-70°C for first two hours but over the span of next hour started going to 70s and even peaking in 80s. Taking the CPU waterblock/reservoir/pump combo and turning it upside down seems to have reduced the temperatures by 7-10°C but i expect it to rise again eventually and even now cinebench run will cause the CPU to throttle even at 4.5ghz.

At his point I would try another cooler. Anything, doesn't matter just to see if the temperatures modify over time and then you can sort out that the cooler is indeed the problem and not the paste or whatever.

The only other cooler that i could try is the NH-D15. Not sure if such a cooler would be able to handle 7820X at reasonable clocks but if the temperature "degrades" even more then i guess ill have to try it.

 

[cucker tarlson]

d15 will do fine

 

[Wet_Paint]

My temps when gaming went up to 96°C most of them time around 89°C, so i changed the cooler to NH-D15, 95°C in cinebench but 50s when gaming with no temperature spikes and the temperature "variance" was extremely small, about 6°C at most, with the watercooling it was >15°C.

Single GPU was still in the loop with pump running and all three radiators connected and it still reached 50°C after couple minutes of gaming.

Im fairly certain that its probably the pump, i just hope i can re-use the CPU waterblock, but what do you guys think?

 

[zenlaserman]

Greetings fellow Bohemian (my family came to US in the 1890s)! Having read this thread up to this point, here's my observations:

1) You were using too much TIM. It appears you may have that fixed now, but just for reference, roughly 1x15-20mm is a fine line.
-Less is more!

2) You are using way too many radiators. I'm no expert in hydrodynamics, but I've worked on automotive cooling systems often.
-That funny car-starting sound you heard was all the air finally bleeding out of your system. Now you have a lot more resistance.
-Your pump likely aint up to all the superfluous 180 degree turns. Try removing the 120 and 240. Or 120 and 360. Run just 1 radiator.

Snad to pomůže!

 

[Vya Domus]

My temps when gaming went up to 96°C most of them time around 89°C, so i changed the cooler to NH-D15, 95°C in cinebench but 50s when gaming with no temperature spikes and the temperature "variance" was extremely small, about 6°C at most, with the watercooling it was >15°C.

So there you have it, something is wrong with your watercooling setup, probably something to do with the pump as you are thinking, I am not an expert. But we've concluded there is nothing wrong with the paste and there is nothing weird with the CPU itself. Should have done that from the beginning, would have saved a lot time.

 

[Wet_Paint]

Greetings fellow Bohemian (my family came to US in the 1890s)! Having read this thread up to this point, here's my observations:

1) You were using too much TIM. It appears you may have that fixed now, but just for reference, roughly 1x15-20mm is a fine line.
-Less is more!

2) You are using way too many radiators. I'm no expert in hydrodynamics, but I've worked on automotive cooling systems often.
-That funny car-starting sound you heard was all the air finally bleeding out of your system. Now you have a lot more resistance.
-Your pump likely aint up to all the superfluous 180 degree turns. Try removing the 120 and 240. Or 120 and 360. Run just 1 radiator.

Snad to pomůže!

Thanks (or should i say Děkuji) for the advices. I've started using a lot less thermal paste after this "experience" so i think i got that covered now and that "car sound" was basically constant, the only way to get rid of it was opening the reservoir with running system and "overfilling" it as the surface level of the liquid would always go down when the system was on. Even this solution worked only for couple days, requiring another overfilling to fix it again. Ive actually tried running with less components in the loop but the results were identical, i think the damage was already done.

So there you have it, something is wrong with your watercooling setup, probably something to do with the pump as you are thinking, I am not an expert. But we've concluded there is nothing wrong with the paste and there is nothing weird with the CPU itself. Should have done that from the beginning, would have saved a lot time.

Now that i look back at it i wouldve done stuff differently, this was definetly learning exprience for me as this is my first proper "custom loop".


Conclusion: I would like to thank to all of you for your advices. I'll save up some money and buy EK-XRES 100 SPC60 and I'll pray that it will solve this issue once and for all.



*I guess the thread can be closed, no idea how it works.... maybe some moderator will notice?

 

[lexluthermiester]

Thanks (or should i say Děkuji) for the advices. I've started using a lot less thermal paste after this "experience" so i think i got that covered now and that "car sound" was basically constant, the only way to get rid of it was opening the reservoir with running system and "overfilling" it as the surface level of the liquid would always go down when the system was on. Even this solution worked only for couple days, requiring another overfilling to fix it again. Ive actually tried running with less components in the loop but the results were identical, i think the damage was already done.


Now that i look back at it i wouldve done stuff differently, this was definitely learning experience for me as this is my first proper "custom loop".


Conclusion: I would like to thank to all of you for your advices. I'll save up some money and buy EK-XRES 100 SPC60 and I'll pray that it will solve this issue once and for all.



*I guess the thread can be closed, no idea how it works.... maybe some moderator will notice?

I would say buy a better/more powerful pump rather than removing a radiator.