Dying AIO?

[QcRMoD]

Hi,

Just wanted to tap into better intelligence than mine.

i7 7700K / H115i

Symptom: sometimes CPU temp ramps up in BIOS. If I let it go it will eventually go into protect. But if I boot into Windows before that happens then everything is fine. Get 25-30*C on the CPU at idle.

Symptom: any time the system gets a modicus of load, sometimes just from background stuff happening, I hear the fans ramp up according to my profile in BIOS – meaning CPU easily hits 60*-70*C.

Symptom: running a bench the CPU goes straight to TJ max. About instantly. Doesn’t crash the PC - CPU throttles holding on to dear life. Of course I shut it down pretty fast.

Things done: swearing

Things done: touching tubes in and out block/radiator, seems to be a good flow all around, no warm spot per see anywhere along the lines, but I do admit the tubes are fairly lukewarm to touch. The block is also warmish but nothing like hot. I cannot remember how it was when brand new but considering what it does, I’m not entirely sure if it’s abnormal or pretty normal.

Things done: repasting/reinstalling

Things remarked: at one specific time in BIOS when the temp was ramping up, I just lightly “shook” the tubes and it went down about instantly. Yet the flow seemed normal as per tested above when I did that.



My 7 years old H115i is dying/essentially useless, isn’t it? Or must I consider another option/test?

Thanks all!

 

[wNotyarD]

First question: does your AIO make strange noises? Namely, does it sound like there's air trapped into the pump?
Second: if the first is a negative, how is the pump speed being reported in the BIOS?

 

[QcRMoD]

1- Not that I was ever aware of even with the panel off.
2- HWinfo says 820 minimum 1040 current/avg/max

Will go and check in BIOS.

 

[wNotyarD]

Can you set it to full speed while leaving the fans at any PWM curve you desire? If yes, check what speed the pump reports and compare it to the spec sheet.

 

[QcRMoD]

BIOS reports 1030 RPM.

Looking quickly seems it can do some 1890 RPM or something.

Unsure if I’ll be able to change it the Corsair link USB cable has long been disconnected but I’ll verify what I can do and possibly where that damn cable was put away.

EDIT:
Won't work through BIOS. God it has been so long that I can't properly remember the details, but I remember that I had set pump through Corsair iCue for Quiet profile. It made the PC virtually silent and kept temps kosher since I did not hold on to the overclock when I decided to "settle down". I think I remember that the profile once set was not software dependent, or that it could be set to be kept on the hardware but I'm not sure. Whatever it was, it won't go above 1040 or so.

So I'm about to turn this off and open the case. I would believe that I was smart enough to leave the USB link hanging behind the motherboard. So we'll see once I can get iCue running and report on the AIO.

But meanwhile I was perusing the build log I had made back when I configured the PC. I did not remember THAT point but the CPU was delidded. And IF it was, then it's 100% sure I used CoolLabs Liquid metal for the die. When I said I done repasting, I sure did not redo the die. So I guess in all good conscience I'll have to see about that too.

Shaking my own head at myself for basically forgetting stuff about my own build.

FURTHER EDIT:

Link reconnected.

So these things are reported divided by 2? The Quiet profile runs 2060-2080 RPM according to iCUE. Now set to Extreme, running 3090 RPM - HWinfo reporting half. A bit noisy but I remember that was why I used the Quiet preset - and it WAS hardware stored. It's just the pump noise as I recall it to be normal and a bit fatiguing at full RPM.

Coolant temperature follows CPU temperatures at idle, which I think is a good sign.

FURTHER FURTHER EDIT

Running OCCT as I write this on the Linpack test.

Coolant temps remains well below 35*C. CPU goes from 65-90*C which is a bit high considering it is not overclocked but nothing like before. iCUE does not report the fans but I knew I had bypassed the controller and plugged directly in Mobo. They run somewhere between 1250-1500 RPM. I reset the BIOS settings there so I think it's what they're meant to run at but will verify.

@wNotyarD I must thank you very really really much for awakening me to my own stupidity and memory lapses. I think I'm just overdue to redo the die paste and regain control of things I had let go. But as it looks like, the AIO functions as new.

 

[wNotyarD]

Things remarked: at one specific time in BIOS when the temp was ramping up, I just lightly “shook” the tubes and it went down about instantly. Yet the flow seemed normal as per tested above when I did that.

Just noticed this, and I'll share a little experience:

I had an episode with the Thermaltake AIO I formerly used (had it for some 4 years) which made me go DEFCON 2.
After a routine maintenance, when I turned the PC on it made an unbearable and weird noise which I've never seen my PC do before and temps went sky high in BIOS already. So I deduced something was wrong with the pump.
I then turned off the PC, unmounted the rad and then gave it a hearty shake while keeping the tubing connections on the pump pointing up and the rad higher than it. Then proceeded to reinstall the rad.

It solved the immediate issue, but led me to lose confidence in my AIO. End result: I got my current air cooler.

 

[QcRMoD]

Just noticed this, and I'll share a little experience:

I had an episode with the Thermaltake AIO I formerly used (had it for some 4 years) which made me go DEFCON 2.
After a routine maintenance, when I turned the PC on it made an unbearable and weird noise which I've never seen my PC do before and temps went sky high in BIOS already. So I deduced something was wrong with the pump.
I then turned off the PC, unmounted the rad and then gave it a hearty shake while keeping the tubing connections on the pump pointing up and the rad higher than it. Then proceeded to reinstall the rad.

It solved the immediate issue, but led me to lose confidence in my AIO. End result: I got my current air cooler.

Hahaha... it's a good and enlightening anecdote the final sentiment of which I totally share.

You can be sure the next PC will dispense of an AIO. It's all cute with cleaner looks and everything but I cannot stomach that kind of shit anymore - the "constant" monitoring and tinkering especially after years when all I want is to not be bothered anymore.

Now while I have you around... do you know anything of using liquid metal on the die for a delidded CPU? I seem to remember it would basically go to reproduce a soldering or about but only with much less wasted material which in turn tends to lead to suboptimal dissipation. Which in turn means that after 7 years and lately some CPU baking, it will probably be a pain in the a$$ to just remove the lid from the die. Should I even bother or should I continue with tinkering with the system and fans profile first?

Or like, when I unclamp the CPU from the Intel locking bracket, is there some kind of wisdom that says that if the lid stays fix and seems unmovable I'm still good, but if the metal liquid has gone all dry/to hell the "link" is mostly likely to break and I'll be able to remove the lid easily?

 

[wNotyarD]

Can't help you there, don't have any experience with liquid metal and even less so with delidding CPU's.

 

[Dr. Dro]

Just noticed this, and I'll share a little experience:

I had an episode with the Thermaltake AIO I formerly used (had it for some 4 years) which made me go DEFCON 2.
After a routine maintenance, when I turned the PC on it made an unbearable and weird noise which I've never seen my PC do before and temps went sky high in BIOS already. So I deduced something was wrong with the pump.
I then turned off the PC, unmounted the rad and then gave it a hearty shake while keeping the tubing connections on the pump pointing up and the rad higher than it. Then proceeded to reinstall the rad.

It solved the immediate issue, but led me to lose confidence in my AIO. End result: I got my current air cooler.

Funny you mention I'm in about the same boat. I've noticed some major performance degradation on my ID Cooling AIO that i've had for the past few years. Gonna pickup an NH-D15S to replace it.

Hahaha... it's a good and enlightening anecdote the final sentiment of which I totally share.

You can be sure the next PC will dispense of an AIO. It's all cute with cleaner looks and everything but I cannot stomach that kind of shit anymore - the "constant" monitoring and tinkering especially after years when all I want is to not be bothered anymore.

Now while I have you around... do you know anything of using liquid metal on the die for a delidded CPU? I seem to remember it would basically go to reproduce a soldering or about but only with much less wasted material which in turn tends to lead to suboptimal dissipation. Which in turn means that after 7 years and lately some CPU baking, it will probably be a pain in the a$$ to just remove the lid from the die. Should I even bother or should I continue with tinkering with the system and fans profile first?

Or like, when I unclamp the CPU from the Intel locking bracket, is there some kind of wisdom that says that if the lid stays fix and seems unmovable I'm still good, but if the metal liquid has gone all dry/to hell the "link" is mostly likely to break and I'll be able to remove the lid easily?

Replacing TIM under IHS with liquid metal was done on some previous generation Intel CPUs that used regular thermal paste there. You'd have to delid it, apply the liquid metal and then close the CPU back up, preferably with thermal epoxy. This isn't needed on newer CPUs. The 7700K may benefit from it, but it's honestly not worth bothering with IMO

 

[QcRMoD]

Funny you mention I'm in about the same boat. I've noticed some major performance degradation on my ID Cooling AIO that i've had for the past few years. Gonna pickup an NH-D15S to replace it.



Replacing TIM under IHS with liquid metal was done on some previous generation Intel CPUs that used regular thermal paste there. You'd have to delid it, apply the liquid metal and then close the CPU back up, preferably with thermal epoxy. This isn't needed on newer CPUs. The 7700K may benefit from it, but it's honestly not worth bothering with IMO

You didn’t get the whole thing.

I am already delidded from 7 years ago when I used liquid metal for the die/ihs.

I’m asking if I should REDO that die pasting thing or not or just unclamp and see if the IHS gets off easily which I surmise would mean the liquid metal is all dried/fucked or if not.

 

[Dr. Dro]

You didn’t get the whole thing.

I am already delidded from 7 years ago when I used liquid metal for the die/ihs.

I’m asking if I should REDO that die pasting thing or not or just unclamp and see if the IHS gets off easily which I surmise would mean the liquid metal is all dried/fucked or if not.

I see. It is not necessary. If you already did this once, then there's no point in removing it and doing it all over again. LM won't expire like a polymer based TIM.

 

[QcRMoD]

Yeah I seemed to remember as much and the final application was perfected from testing an Haswell 4770K with six different kinds of TIM and LM (confounded) and finding the most reliable setting was using the Cool Laboratories initial liquid metal (can't remember the marketing naming difference) on the die AND none other than Noctua regular TIM on the IHS (seemed to garner slightly better temps in ALL settings but die application).

I wonder... if I've found my old building log for the Kaby, I might still have the comparison grids of all tests carried on the Haswell 4800K.

By my memory the Haswell had been tested with an OC of 4.8GHz for these applications.

I was much more into IT back then than now. Could barely decipher most of my OC/build tests logs for the Kaby. In my mind I had never reached such a deep understanding and it's only coming back to me now that I started participating here. I mean, coming back to me that I was once that deep into it. Not coming back to me so much is understanding what I could do and understand back then.

EDIT:

Haha can't find the comparison grid cause it was done on Excel and erased since I would surmise.

What I DO find is all tests done and TIMs used in pictures of benching, and from my file naming scheme, I can probably extrapolate what was used and why.

Oops also sorry it wasn't 4800K it was 4770K. Corrected.

 

[LabRat 891]

Just noticed this, and I'll share a little experience:

I had an episode with the Thermaltake AIO I formerly used (had it for some 4 years) which made me go DEFCON 2.
After a routine maintenance, when I turned the PC on it made an unbearable and weird noise which I've never seen my PC do before and temps went sky high in BIOS already. So I deduced something was wrong with the pump.
I then turned off the PC, unmounted the rad and then gave it a hearty shake while keeping the tubing connections on the pump pointing up and the rad higher than it. Then proceeded to reinstall the rad.

It solved the immediate issue, but led me to lose confidence in my AIO. End result: I got my current air cooler.

I actually had this just happen to me. (except shaking was not sufficient)

Grabbed the H100i V2 off my shelved 3770k to mount to a GPU, instead. I 'bench tested' the AIO, even plugged in the USB to monitor the pump: worked fine, only a brief 'gurgle' @ 1st start.
After mounting the cooler to the GPU, 1st test (no matter orientation) the pump is just groaning like mad.

Same thing happened w/ an old Antech Kuhler 620. Mounted to GPU, 1st tests worked great. Removed, shelved for a week, come back, and the AIO is groaning and whirring w/ the GPU quickly overheating.

Considering this happened to 2 diff coolers, with 2 diff age/use profiles prior to storage...
I'm gonna go out on a limb and guess:
Probably sludge/gunk, that got 'mixed into' the coolant @ 1st start after storage, and then 'settled in' and blocked up the pumps.

Sadly, all my experiences (3x 120mm models, 2x 240mm) w/ AIOs points towards them being single-installation consumables.

Like you, I think I'll stick w/ air cooling (unless passive+pumpless phase-change loops, become 'a thing')



@Dr. Dro

Copper IHS + Liquid Metal Alloy/Amalgam

After ~2.5 years since I de-lidded, added liquid metal, and replaced the IHS of my 8700K, core #1 suddenly experienced a massive spike in temperature in usage (around 24-25 deg. higher) while the other 5 cores only experienced a small rise in temps (2~3 deg.) Suspected the liquid metal needed...

www.techpowerup.com

LM TIM will 'expire' through metal amalgamation/alloying.
However, there are ways to mitigate that
Notably though, *I* did not go through those procedures when I de/re-lidded my 3770k (5+ years ago)

I'd bet my 3770k needs a re-do of Galinstan/LM TIM, before restoring to use.
-and I'd bet OP's does (or eventually will) as well.
Nothing against OP, I've learned much in those years-between, myself.

 

[QcRMoD]

I've not had any problem of the sort with the AIO mounting or reciprocating bad temps in benches - when new or a couple years later when I ditched the OC and regulated the AIO to Quiet and then unplugged the link. This done probably in 2019 when I moved and decided to simplify my life.

But I appreciate that I might be due for a total dismounting and remounting of everything from grounds up.

However do I WISH to be really forced to do so?

I think the answer would be readily decipherable to anyone reading this: FUCK no!

I originally had thought that LM TIM was a rather permanent solution. It is not.
(The 'low-melt' metallic pads might be; being an Indium-heavy non-liquid soft alloy)

From what I've since-learned, one of the primary elements in the eutectic alloy, preferentially alloys to Copper (and to a lesser degree, nickel). Over time, your LM TIM application is slowly turning into a rather 'bumpy and dry' amalgam of Gallium and Copper(and/or Nickel).

In the past, When I have used LiquidMetal TIMs, I use clear nail lacquer to mask around the die. (so, you're on-point there).

I still have a small quantity of Galinstan on-hand but, I've stopped using it and switched to PTM7950 for 'high performance' TIM applications. Long-term reliability has become a much more important consideration to me, in the last few years.

More memories coming back: I did also nail-lacquered these.

And also I think that ALMOST-BUT-NOT theory of LMs was already more or well known even back then.

Thank you so much @LabRat 891 for your participation in my queries and forays.

 

[wNotyarD]

I actually had this just happen to me. (except shaking was not sufficient)

[...]

Considering this happened to 2 diff coolers, with 2 diff age/use profiles prior to storage...
I'm gonna go out on a limb and guess:
Probably sludge/gunk, that got 'mixed into' the coolant @ 1st start after storage, and then 'settled in' and blocked up the pumps.

Sadly, all my experiences (3x 120mm models, 2x 240mm) w/ AIOs points towards them being single-installation consumables.

Like you, I think I'll stick w/ air cooling (unless passive+pumpless phase-change loops, become 'a thing')

Mine was certainly just trapped air. But it scared the shit out of me. Luckily the shaking sent the air back to rad and freed the pump, but I at least had my former 5600X's Wraith Spire at hand if push came to shove.

And while I am sticking with air (AK620D) for the time being, I wish to take a new step when I build my next rig and install a custom water cooling loop. Where I live is just way too hot and while temps aren't critical with my current hardware (at most GPU hotspot in the low 90C's and CPU at 75C's) I'd like for a cooler and even more powerful system in the future.

 

[QcRMoD]

Mine was certainly just trapped air. But it scared the shit out of me. Luckily the shaking sent the air back to rad and freed the pump, but I at least had my former 5600X's Wraith Spire at hand if push came to shove.

And while I am sticking with air (AK620D) for the time being, I wish to take a new step when I build my next rig and install a custom water cooling loop. Where I live is just way too hot and while temps aren't critical with my current hardware (at most GPU hotspot in the low 90C's and CPU at 75C's) I'd like for a cooler and even more powerful system in the future.

Really wish you a custom loop will work for you better. A nice project to carry anyway.

 

[wNotyarD]

Really wish you a custom loop will work for you better. A nice project to carry anyway.

What good is a PC enthusiast if not planning the next great upgrade?

 

[Chrispy_]

Sounds to me like the AIO is gunked up. Even if the pump is running with decent flow, the microfins under the jet plate might be full of either organic goop or galvanic corrosion and so the water isn't actually cooling the cold plate directly at the die - the only reason any cooling is happening at all is because the outer edges of the CPU block are being cooled by high flow and there's a long/slow thermal conduction path to the centre of the coldplate.

If you're confident that the pump is running at or near it's optimum speed, and that the block is making good contact with the CPU, then you should unscrew the 8-16 bolts around the coldplate, and inspect/clean the microfins. If they're gunked up you should be able to scrub them clean with a toothbrush. If they're corroded it's time to scrap the AIO and buy a new cooler. Drain the loop, flush it under running water, drain it again and refill it with fresh off-the-shelf premixed glycol-based coolant.



A couple of examples of what I suspect you'll find...

 

[QcRMoD]

Obstination shall prove rewarding: I found that chart when I tested TIMs and LMs. Not indicated in the chart but the cooler was NH-D14 or some monster Noctua of that kind.

Sounds to me like the AIO is gunked up. Even if the pump is running with decent flow, the microfins under the jet plate might be full of either organic goop or galvanic corrosion and so the water isn't actually cooling the cold plate directly at the die - the only reason any cooling is happening at all is because the outer edges of the CPU block are being cooled by high flow and there's a long/slow thermal conduction path to the centre of the coldplate.

If you're confident that the pump is running at or near it's optimum speed, and that the block is making good contact with the CPU, then you should unscrew the 8-16 bolts around the coldplate, and inspect/clean the microfins. If they're gunked up you should be able to scrub them clean with a toothbrush. If they're corroded it's time to scrap the AIO and buy a new cooler. Drain the loop, flush it under running water, drain it again and refill it with fresh off-the-shelf premixed glycol-based coolant.

View attachment 337718View attachment 337723

A couple of examples of what I suspect you'll find...

Yeah man I won't even go there. This will hit the trash can when I'll buy myself a new PC. Will install an air cooler on the current one then and transfer it to media server duty, the one I have has started having random BSOD from time to time. But thanks for the cue and trying to help.


EDIT:

Did some work tonight.

First, cleaned my HTPC and bypassed it into Win 11. The case is an old Silverstone roughly ten years in service. Not the smartest insides but back then it was cheap and it fits in my A/V setup/shelves. Exterior is black obviously. In there an also old - more than ten years - 2.0 Soundblaster Titanium HD card that was of the rare kind to have a real RCA output and proper DAC back then.

But of course the real work was with the AIO and delidded CPU on my PC - which I took some time to clean out first, however there wasn’t much there.

When I unclamped the IHS came right off. The LM application had indeed aglomered and was just about clumping.

Immediately cleaned the most of the beaded clumps from the die and silicon.

A long cleaning process ensued.

Redid the die application with CLP.

Ready to install.

Used the Noctua NT-H1 on the IHS as usual.

As we speak the LM is getting settled down in light use. Temps are alright there. Will let it cool off overnight and tomorrow will stress the CPU.

 

[QcRMoD]

FINAL EDIT:

Did not go well at first in stress test. 70-95*C in OCCT and then nearing 100*C and not going down.

So I went for a final attempt and turned the block 45* anti-clockwise. Not the most gracious look with the Corsair logo sideways (and pardon the mess of wires, but no point to tidy up if it ain't gonna work)...

But now I've got something that's about right.

Still running atm, not going above 75*C.

So that's gonna be it until I trash that thing and buy an air cooler for EOL of that PC. I don't think that without an OC these temps are fine (perhaps they are) but it'll have to do for a while.