High GPU and water temperatures in new custom loop

Aug 5, 2018
10
0
10
0
Hi everyone.

Im new to this custom watercooling. Been wanting to make a loop in my computer for a long time, and now i finally did. After a lot of hassel, with a bad monoblock i finally got my CPU temps to get steady.

Now my problem is that my water temps are rising kinda high (i think they are high?), and my GPU is getting a lot hotter then what im seeing other people report.

When idle my temps are as follows:
CPU: 35C
GPU: 30C
Water: 34C
Ambient: Somewhere around 25C

When i start gaming, my temps go up tho (ofc they go up):
CPU: Sits around 63C, spikes up to 72C on hottest cores.
GPU: Have seen it go as high as 61C. This is in both games and heaven bench.
Water: Settles in at around 44C (fans are running a lot louder then i would like at this point)
Ambient: Guess it is still around 25 C.

Are these temps normal? Im a bit dissapointed atm, as my custom loop is making more noice then my CLC + air cool gfx.

Specs:
I7-8700k @ 5.0GHz, 1.33vcore, LLC level 6(of 7), delidded
Asus Z370-f gaming motherboard
MSI 1080 TI gaming X, small OC of +50 clock and +100 mem speed
Case is Lian-Li o11 dynamic

Watercooling:
2x 360 EK radiatior. The one with medium thickness.
EK EVO supremacy CPU block
Phanteks GPU waterblock
EK combo res/pump
Using 6 noctua fans at 85% speed. 3 as intake in the side, 3 exhaust in the top.


Flow seemed alright when filling my loop. Currently the pump sits at 45% speed at all times. Rads are getting warm when under load.
Pic of the loop, no water in: The loop
 


to make it short, I'll quote one of my previous messages:
1. already mentioned it previously - change the side fans on the rad to be exhaust but keep them on - you want it for a number of reasons:
1.1 the top rad will be not reusing hot air from the side rad when the case is closed
1.2 it's much easier to clean "pull" config. in "push", the dust/dirt buildup is between the fan and rad and it requires disassemble to clean. I'd do the same change from push to pull to the top rad, but that might require change in fittings used to connect the tube going from rad to res.
2. remove the dust filters on the panels where radiators are installed - they are hurting performance without bringing something useful (once all rads are exhausting air)
test the change with the case open, than closed with fans set to 100%.
good result - the liquid should be 7-10C over the ambient and the GPU 10-15C above liquid. so you are looking at 17-25C difference between ambient air and GPU temp.
Just do the:
1. make the side radiator fans to exhaust the air out of the case. also true for the top rad
2. remove the dust filters where radiators installed.
3. add a fan or two at the bottom of the case (optionally rear too ) to push the air into the case.
 
Your temps are high based on my own experience. I'm cooling the same CPU and same GPU series. My CPU idles around 30C and floats around 60 or so in gaming, though I've haven't checked in awhile. I'm running a single loop with CPU and GPU. However, my water temp and GPU are much lower. I have thermal probes for both water and air temps. As I'm typing this, the air temp is around 26C and water is 27C. During gaming runs for a few hours, highest I've seen water temp is about 35C. My GPU also idles around 30C, but even full load when OCing gets to low 40C range. Have to look into this one. Nice build, BTW.
 
Aug 5, 2018
10
0
10
0
Im also running both CPU and GPU in a single loop. Order being Res>Pump>Rad1>GPU>CPU>Rad2.

I get my water temps from a monsoon stop plug with a temp probe, placed in the inlet of my GPU block. So these should be accurate enough. The ambient temp i have no way to read, but from the thermometer in my car. But cant be THAT far off.

Have tried running my pump at both lower and max speed, but makes no difference at all. No flowmeter tho, but as i said before, the fluid flowed no probs, when filling the loop.

I got no idea how to troubleshoot this :(
 
OK. Firstly, what was wrong with original monoblock? Flowrate would be one of first things to check. You should have plenty of cooling capacity for your setup. Fans should be able to be ran much lower speed I would think. I have mine set to about 1200RPM. Sure the GPU block is mounted well? Can you hear air bubbles in the system? Possible you have an air pocket that needs worked out. Not seeing anywhere with a restriction in pic.
 
Aug 5, 2018
10
0
10
0
The monoblock i used to start off, didnt make proper contact with my CPU. When i took it off, the TIM hadnt even spread out. Tried several times, but no luck. Temps would be like 60C idle and 95C load. Mounting pressure is just fine with the normal CPU block tho.

Just tried to "shake" the system a bit. Got a little air out of the top rad, and sounds like there could be a little more. But i imagine there should be quite a large air pocket in the rad, to increase temps like this? Didnt help anything on the temps btw.

After that, i let a little liquid out of the loop, so i could see the air move in the system. Water seems to be flowing at a very decent rate.
 
nice looking loop.
the idle temps are fine. i'm very curious where is the liquid temp sensor as i can't find it on the pic.
the liquid temp goes too high for the amount of rads and the ambient of 25C.
so the questions are:
1. how are the fans configured on the vertical rad (intake or exhaust)
2. have you tried to increase the pump speed to 100%
3. did you had the case side panels removed during the testing.
If the answer to q 2 and/or 3 is no, try to do it and share the results.

As a side note, I hope you aware that current CPU block orientation is not optimal. outer cores will be hotter than inner cores due to liquid channeling withing the block in this orientation.

 

rubix_1011

Contributing Writer
Moderator
I completely missed that there was a link to an image - good looking build, but all those 90 degree angles can add up to some flow restriction and slowdown. (there should be a gradual, 'arcing' 90 bend fitting, in my opinion).

D5's do have a bit more trouble getting air out of restrictive loops due to lower head pressure than say, most DDC pumps do, so you could still be fighting some air pockets, likely inside the radiators themselves.
 
Aug 5, 2018
10
0
10
0
First, im really glad you like the build. Spend a whole lot of time on it, and like how it turned out myself :) The temp sensor is located at the spare intake port, on the back of the GPU. The vertical rad/fans are intake (there are vent holes on the back cover), top rad/fans are exhaust. Pump is already running at 100%, tho it made no difference compared to 45%.
I was not aware that the CPU block orientation made any difference tho. The core to core difference is very low tho, so dont think that will be my main concern atm.

But i just tried some stuff:
- I see no air bubbles in my GPU block. I do see some bubbles on the edges of the CPU block insert. I tried getting these to move, but they are just stuck in there for good. No odd sounds from the pump at all.
- Removing a cap from the reservoir made no difference to flow.

Now for the thing that kinda worked a bit.
- I gave it yet another good old shake, turned it around, tapped the radiators a bit. And there actually was some air in the top rad. As i worked the air out of the system, i got a look at the flow again(indicated by the air bubbles) and it seems to be moving rather fast around the loop.

After getting rid of some air in the system, i ran a GPU stresstest again(no time for CPU today)


First test without any panels on panels:
GPU: 47
Liquid: 40
Ambient: 27,5
Fans: 1000RMP (around 75%)


Second test, with all panels on:
GPU: 50
Liquid: 43,6
Ambient: Still 27,5
Fans: 1200 RPM

So yeah, panels makes a +4C difference in water temp.

Will give it a good shake and turnaround again tomorrow, to see if i can get more air of the system, but i highly doubt there is anymore. But worth a shot.


Thanks for all the helpfull answers so far!


 

rubix_1011

Contributing Writer
Moderator
4C difference is likely due to case airflow, or not enough ambient room air making it into the chassis with the panels closed, which does mean your core temps will report slightly higher, although I'm surprised at almost 1:1 change by comparison - I would have expected maybe 2C difference on reported core temps.

Good news is you are reporting what appears to be healthy flow rates, which helps eliminate that concern. Air pockets still may exist, but might work their way out more and more if you tilt your case while running (top off your reservoir first to prevent air from making it's way back around) Once you see your reservoir level begin to drop, this is a sign that air is beginning to collect there instead of elsewhere in your loop. Simply uncap and fill up to displace this air and prevent it from getting pulled back into the loop.
 
Aug 5, 2018
10
0
10
0
Just tried to run the system tipped to all sides, upside down, face down, shakeing it a bit. Only got a tiny tiny bubble of air from the rads. Cant hear any air moving aswell. I really dont think there is anymore air in the rads. Only air i see, is some in the CPU, which really dont wanna move, and some tiny bubbles the the side of the reservoir.
Problem remains the same. After around 10 mins in heaven bench, water temp is at 43.2C, and slowly rising when i stopped the test. Both rads are getting equally warm to the touch.
 
Still watching this thread. See the mystery is still there. Radiator capacity is enough that neither should really get "warm" as mentioned. Is the pump unusually hot for some reason? Not sure where this excess heat could be coming from. What are your idle water temps? What are GPU temps getting to with current testing?
 

rubix_1011

Contributing Writer
Moderator
Also, what are CPU and GPU core temps being reported as? Curious as to what those are also.

Depending on your ambient room temp, you are hitting a much higher delta-T than someone with as much radiator space as you do.
 
with the amount of rads you have, over 40C liquid temp is a bit too much for just heaven @27C.
though the GPU temp is fine.
I compare it to what I get with 33% less rad surface (2x240), higher ambient ~33C, slower running fans up to 1350RPM (guess we use the same notctua PPC), though a less power hungry card (heavily overclocked and power modded so its a bit lower power consumption). And yet I never see the coolant at even 40C.
that is about 9-10C difference (ambient + coolant).
A few degrees can be attributed to different sensor, but not 10.
My guess is that there is some issue with the flow in the loop. it might be not related to the air trapped in the loop.
What is the exact model of the temp sensor (link to it would be awesome) - some of them are too long to be mounted on the GPU terminal - the depth of the sensor including thread should be under 5mm.

As a side note, I'd recommend to make the vertical rad exhaust as well - it will make the performance more similar to "removed panels" as the top rad will not be re-using hot air from the vertical rad.
 
Aug 5, 2018
10
0
10
0
"Also, what are CPU and GPU core temps being reported as? Curious as to what those are also."
Not sure what you mean with this? The temps im writing to you guys, are just the readout from HW monitor.

My rads are not getting hot. But i can easily tell that they heat up a bit under load.
The fans im using is the https://www.overclockers.co.uk/noctua-nf-f12-pwm-chromax-premium-grade-fan-120mm-fg-06n-nc.html with a max RPM of 1500, when my water hits 44c, they are running like 1200 RPM

But yeah, im not really THAT unhappy with GPU temps at 55C. Though i had expected more like in the high 40's tbh. If they would just keep the temps with a lot lower fan speed. But will turn around the side fans again, and run only exhausts. Had it like that to start, but switched it, just to try.

The temp pulg im using is: https://www.overclockers.co.uk/monsoon-premium-g-1-4-temp-plug-chrome-wc-137-mo.html
It is connected to my Corsair Commander Pro.
Might order another temp sensor, and stuff that in the Reservoir, and see if i get a different reading. I posted this thread on reddit aswell, and some guys there got confused, that my system was above ambient when i turned it on. So just to explain, just when i boot up the PC, the temp sensor report the water temp equal to ambient temp. After like 10 mins sitting on desktop, water temp gets to around 34c.

Is there any way to test the flow through the loop? Several people have suggested this to me by now. I could just let in some air, and let the air circulate, while i get a video of it for you guys. Would you be able to say good / bad looking at that?
 
Aug 5, 2018
10
0
10
0
(dont know how to make a Quote)
"Is the pump unusually hot for some reason? Not sure where this excess heat could be coming from. What are your idle water temps? What are GPU temps getting to with current testing?"

Pump is not getting hot at all. All i feel is some minor vibrations, which seems normal. Right from boot, water temp is equal ambient (reading ambient on a fridge thermometer. The best i got). After 10 mins, water temp has gone up from 29c(today) to 35c, with fans running slowish(40%). Idle GPU temps today is 31c (cooler then the water, seems odd to me).

As soon as i open Heaven, the GPU sits at 47c. From there, the GPU temp slowly increases with water temp. After 10 mins in Heaven(Thats what she said:p ), GPu temps is 55c, water temp 41c, fans at 1200RPM. (Running with panels off today).
 

rubix_1011

Contributing Writer
Moderator
I thought you had seen different load temperatures once a few adjustments had been made. Based on your liquid temps, you are seeing a higher delta-T than I would definitely expect for a system running as much radiator space as you are seeing.

Watercooling delta-T is dependent upon several things: coolant flow rate, airflow of fans and radiator surface area. You could also consider thermal conductivity of the blocks and radiators, as well.
 
oooh ... that's the chroma ... at 1200 RPM it makes more sense. still high, but not terribly wrong :)
when you turn on the computer it's normal to have the liquid at the same temp as the ambient. at idle it should climb ~5-10C depending on the fans speed.
There is nothing wrong with the sensor too.

IMHO, you should see mid-high forties for the GPU. But i'm out of ideas for non-invasive troubleshooting. Going forward I can only suggest things that will require some temporary or even permanent (like adding flow sensor) modifications.
Just say yes and I'll start typing :)
 
Aug 5, 2018
10
0
10
0
"Going forward I can only suggest things that will require some temporary or even permanent (like adding flow sensor) modifications.
Just say yes and I'll start typing :) "

I want this build to run flawless, so tell me what to do, and i will try it. Tho if it involves adding new stuff, like a flowmeter, to the loop, i wont do it in the next 14 days. Tho im kinda oldish, there is still a LAN and a new WoW exp comming up :D
 
I'm not in a hurry, and so you aren't too - for now the loop is working well enough.
an addition of a flow meter is useful regardless of the current "issue". It helps to identify future possible issues like partial blockage due to some gunk buildup or whatever else. you see flow drop - time to do serious maintenance to the loop :)
here it gets a little tricky - you need some controller to connect the sensor. I like the ones from aquacomputer.
aquaero is great, but expensive. let me know how far you want to go with it in terms of budget, controlling options and space occupied.


now to the problem: let's start with "easy" thing.
1. already mentioned it previously - change the side fans on the rad to be exhaust but keep them on - you want it for a number of reasons:
1.1 the top rad will be not reusing hot air from the side rad when the case is closed
1.2 it's much easier to clean "pull" config. in "push", the dust/dirt buildup is between the fan and rad and it requires disassemble to clean. I'd do the same change from push to pull to the top rad, but that might require change in fittings used to connect the tube going from rad to res.
2. remove the dust filters on the panels where radiators are installed - they are hurting performance without bringing something useful (once all rads are exhausting air)
test the change with the case open, than closed with fans set to 100%.
good result - the liquid should be 7-10C over the ambient and the GPU 10-15C above liquid. so you are looking at 17-25C difference between ambient air and GPU temp.

another optional change - https://noctua.at/en/nf-a12x25-pwm
they push much more air at much lower noise. let's say at 1500RPM they will produce about the same noise as your current fans at 1000RPM while pushing way more air through the rad. but they are expensive and ugly - shouldn't be a problem if you manage to hide them on top rad.

the other thing to test the the flow - for that a flow meter or a 1.5-2 meter (with some barb fittings) flex tube will be required.
I'm less inclined to test it because (if i understood you correctly) the temp sensor is located on the back of the GPU inlet port - which means it is in the coldest point in the loop after all rads and before any heat emitting component.
 
Aug 5, 2018
10
0
10
0
So i just tried to turn the fans in the side, and remove the dust filters. Then ran heaven again. Both runs is with the panels on. Fan speed did change the results more then i did expect tbh. But the system is just waaay too loud then.

Fans 850 RPM
Ambient: 26
Water: 43,6
GPU: 53

Fans 1350 RPM
Water: 38
GPU: 47


I kinda dont wanna change the fans, as i just brought them for this build, and as you say, they are expencive as hell. This model should be good enough aswell, i hope. Have been running the 140mm and 120mm ugly versions on mine and GF's AIO for couple of years. Here with max speed around 750 RPM and 550 when browsing.
That said, if they are a limiting factor/problem, i will change them.
Other then that, budget is pretty open, anything within reason goes. Dont want to clutter up the build more tho, so if stuff is gonna be permanent, it needs to be stealth :)

 
The delta between the coolant and GPU is great.
So, i'll insist on another go - top and both side panels off. that will tell if the problem is airflow, or radiator performance.
in case of the former (5C or more improvement in liquid temp) - you can either leave it as is (totally fine option) or to consider the fans change since they push more air at much lower noise (even at higher RPM) - the cost is important, but on the other hand they have 6 years of warranty. Another reason to change the fans, is because you have a bit thicker rad which requires a bit more static pressure from the fans. Though adding another set of fans exactly like yours (push/pull config) should improve the temp by 2-3C while keeping the noise at the same level.
in case of the later, it's either significant amount of air in the rad or may be some path (or more) is blocked.

The reason I think that your loop performance can be improved is because I just did the test on my system with following params:
2x240 30mm rads - thinner than yours (less important) and have 33% less surface area (very important)
Heavily overclocked GTX 1070 + couple of prime95 threads to compensate power consumption difference with your GTX 1080Ti
4xNF-F12 @1350RPM
Ambient of ~30-31C
more restricted airflow - intake is on the rear ~120mm fan size opening and very restricted bottom.
and yet i the liquid didn't cross 41C (removing the side panel would drop the liquid temp by over 3C to bellow 38C)
Given that in your loop there is 50% more rad surface and ambient is lower by 4-5C it should do much better.

I'll post the links to "gadgets" later.
 
Making some progress I see. I'm starting to wonder if it's from the RPMs being so low causing your temps higher than expected What FPI are your radiators? Higher FPI rads need higher RPM fans for the "sweet spot" of their cooling capability due to fin density, whereas lower FPI are better suited for lower speed ones than others. That said, the higher the RPMs, the better the performance regardless of radiator type. I run mine (Gentle Typhoon AP-15s) at about 1200 RPM. Although audible, they are quiet enough for me. For running sub 1000 RPMs, it's best to have tons of extra cooling capacity to be near silent and have very low temps.
 

ASK THE COMMUNITY

TRENDING THREADS