Overheating Watercooling loop? - Please Help

Associate
Joined
4 Jan 2020
Posts
19
Hi All,

First time poster on the forums but long time lurker.

I have recently decided to take the plunge and build my own watercooling loop after many years of building pcs on air and having a room that heats up like an oven. The issue is my GPU is currently running what i believe is way too hot for a water cooling loop.

My PC specs
CPU - AMD 3950x - PBO on no other overclock

GPU - Asus Strix 2080ti OC Edition
Overclock settings for GPU via MSI Afterburner
Core Voltage - +0
Power Limit - +125
Temp Limit - 88C
Core Clock - +175
Memory Clock - +1000

RAM - Corsair Vengance - 3600MHZ

Motherboard - Gigabyte X570 Aorus Elite

PSU - Seasonic SSR-850FX FOCUS Plus Gold 850W

I have currently in my loop (im pretty sure this is overkill but wanted to be sure)
2x EK-CoolStream Classic SE 360
6x EK-Vardar EVO 120S BB (700-1150rpm) Running at 100%
1x EK-XRES 100 Revo D5 PWM (incl. pump) Running at 100%
1x EK-Velocity - AMD Nickel + Plexi
1x EK-Vector Strix RTX 2080 Ti RGB - Nickel + Plexi and EK-Vector Strix RTX 2080 Ti Backplate - Black

I have my loop order of Pump/res > GPU > Radiator > CPU > Radiator > Pump/res

Before the loop on air my temps were
CPU ran 75-77C under full load for 15 mins with an NHD15 with fans fixed to 70%.
GPU ran 75-78C under full load for 15 mins with stock cooler set to 70% overclock has been stable for around 3 months with not issues or changes to the above temperature.
First Loop attempt
CPU ran at 70-72C under full load for 15 mins
GPU ran at 87C constant and thermal throttled I stopped the test after 5 mins
Currently with my loop my temps are
CPU runs 70-72C under full load for 15 mins
GPU runs 78-81C under full load for 15 mins
When both ran together under full load for 15 mins the GPU runs at around 83-85C and the CPU runs between 78-81C

Initially I thought the issue was poor paste spread on the GPU, so i removed the water block and confirmed that this was not the issue but i believe that i hadn't had enough mounting pressure as the thermal pads looked untouched on the GPU.
I removed the old paste and reapplied some more in the same method as previously done and made sure to tighten the screws harder this time using the provided screws and plastic washers.
I did the same for the CPU and removed the old paste and reapplied new and re-tightened.
Both paste spreads fully covered both the IHS on the CPU and the entire chip on the GPU.

Tested again and as you can see above the temps are better but still are running awfully hot I believe.
Is there any suggestions as to what may be the cause of the issues and if so any advise would be amazing, im really out of ideas as to what could be the issue.
Or am i just expecting to much out of the loop and it should be that hot?
Loop picture below

JPEG_20200101_222830.jpg


Thanks all,
Josh
 
Soldato
Joined
18 Jul 2009
Posts
2,784
First thought would be the pump not moving water around effectively. Quick check to feel the radiators during load. If they feel hot then you know the pump is moving water around otherwise could be a dodgy pump.
 
Associate
Joined
5 Mar 2017
Posts
2,216
Location
Cambridge
GPU only I would say wrong screws used to fix the block. May not few loose, but not enough contact to work properly.
CPU can be the protective film still there (yes, happens), but harder to get a poor contact, opposed to the GPU.
I would first make sure you're using the correct ports as in/out. Not much of a difference, but may cause some performance impact.
Fittings: I decided to go with the ones from EK, as I was able to see that any fitting with thread longer than 5mm would stick out of the designated threaded area and, again, restrict waterflow.
Running the pump at 100%, only when trying to get rid of air bubbles. I've noticed a marginal loss in performance, as the coolant pass too fast. Mine stays at 3 out of 5. No noise, more than enough flow.
 
Soldato
Joined
18 Oct 2002
Posts
8,270
Location
Aranyaprathet, Thailand
Basically heat emitters staying hot means either:

a) poor transfer to the blocks
b) poor transfer to water (caused by low flowrates)
c) poor transfer to the air (low fan airflow)

You need to isolate the possible issues.

1) use a cpu load tester (e.g. Prime95) to test the cpu cooling.
2) use a gpu load tester (e.g GPU-Z) to test the gpu cooling.

If one overheats but the other does not, then you may have a poor mount on the overheating component. Check it again.
If both are overheating it is unlikely that both are poorly mounted.This suggests either pump or fans. If you can feel air through the radiators, then it is more likely to be the flow.

Flow isn't simply caused by the pump. Does it have manual speed control? Check it is on the highest setting.

Assuming the pump is on full flow, you may have a blockage or pinchpoint somewhere.
 
Soldato
Joined
28 Aug 2017
Posts
2,755
Location
United Kingdom
pure and simple your ek vardar fans arn't running fast enough, you should be aiming for at least 1500 rpm, for optimal performance you should be looking at fans around 1750-2200rpm, the heatoutput from you components at stock is around 550w of heat, the se rads you have are able to cool around 420w of heat each but thats with fan speeds in access of 1850rpm.

you should be able to feel airflow from the rads a good few inches away, if you dont the fans arnt pushing enough airflow through resuting in you loop getting really hot to the point of shutdown or the pipes could blow off whilst the pc is on, if that happens it'll be a very bad day.

i'd say it's your fans that are the problem as loop order isnt really going to affect temsp that much, have a look at faster rpm fans
 
Soldato
Joined
21 Dec 2019
Posts
6,439
Location
Planet Thanet
Got same pump/res combo
That pumps definitely up to the job even if it was at 50% should be ok
Are you getting an actual RPM reading for the pump?
My gtx1070 is only 45c to 50c under heavy gaming load as a comparison point
 
Associate
OP
Joined
4 Jan 2020
Posts
19
GPU only I would say wrong screws used to fix the block. May not few loose, but not enough contact to work properly.
CPU can be the protective film still there (yes, happens), but harder to get a poor contact, opposed to the GPU.
I would first make sure you're using the correct ports as in/out. Not much of a difference, but may cause some performance impact.
Fittings: I decided to go with the ones from EK, as I was able to see that any fitting with thread longer than 5mm would stick out of the designated threaded area and, again, restrict waterflow.
Running the pump at 100%, only when trying to get rid of air bubbles. I've noticed a marginal loss in performance, as the coolant pass too fast. Mine stays at 3 out of 5. No noise, more than enough flow.

I thought it was the same when I first fitted the gpu block so checked and double checked and made sure i used the correct provided screws for the GPU.
The CPU i removed the film on the block when originally fitted.
I have tried running the pump from 60% to 100% and the difference in performance is very marginal indeed for me too.

Basically heat emitters staying hot means either:

a) poor transfer to the blocks
b) poor transfer to water (caused by low flowrates)
c) poor transfer to the air (low fan airflow)

You need to isolate the possible issues.

1) use a cpu load tester (e.g. Prime95) to test the cpu cooling.
2) use a gpu load tester (e.g GPU-Z) to test the gpu cooling.

If one overheats but the other does not, then you may have a poor mount on the overheating component. Check it again.
If both are overheating it is unlikely that both are poorly mounted.This suggests either pump or fans. If you can feel air through the radiators, then it is more likely to be the flow.

Flow isn't simply caused by the pump. Does it have manual speed control? Check it is on the highest setting.

Assuming the pump is on full flow, you may have a blockage or pinchpoint somewhere.

I have tried CPUZ as i cannot get Prime95 to run for some reason and this after 15 mins has topped at 72C
I have run GPUZ after this and after 15 mins has topped at 63C
I then ran both together straight after these tests and the GPU went to 65C and CPU 75C

Not sure why these results are so different for the GPU, To test load before for GPU I ran conan exiles in an area i know is basically maxing out my GPU before and the temps are so different, could this mean its an issue with memory maybe as im guessing a game would put more load to the memory compared to a GPUZ?

When it comes to airflow i can barley feel the airflow of the fans behind the radiator at the front of my case and the top the airflow is rather slow too, so is it possible the fans are not high enough RPM for the radiator thickness?

The pump is on 100% via the BIOS as its PWM controlled and see no other way to change this, I have also checked for pinches on the pipes and nothing can be seen and the water looks clean when watching it move around the system and inside of the water blocks.

When trying again with conan after these tests the GPU went to 78C and CPU was around 62C
 
Associate
OP
Joined
4 Jan 2020
Posts
19
Got same pump/res combo
That pumps definitely up to the job even if it was at 50% should be ok
Are you getting an actual RPM reading for the pump?
My gtx1070 is only 45c to 50c under heavy gaming load as a comparison point

My RPM shows at 4750 on the BIOS but this does change up and down slightly

pure and simple your ek vardar fans arn't running fast enough, you should be aiming for at least 1500 rpm, for optimal performance you should be looking at fans around 1750-2200rpm, the heatoutput from you components at stock is around 550w of heat, the se rads you have are able to cool around 420w of heat each but thats with fan speeds in access of 1850rpm.

you should be able to feel airflow from the rads a good few inches away, if you dont the fans arnt pushing enough airflow through resuting in you loop getting really hot to the point of shutdown or the pipes could blow off whilst the pc is on, if that happens it'll be a very bad day.

i'd say it's your fans that are the problem as loop order isnt really going to affect temsp that much, have a look at faster rpm fans

After a few messages from you amazing guys i think that sounds like the issue to me also, the airflow i feel is so low that its not right IMO and i should probably go for faster fans as a test at least?
 
Soldato
Joined
21 Dec 2019
Posts
6,439
Location
Planet Thanet
pure and simple your ek vardar fans arn't running fast enough, you should be aiming for at least 1500 rpm, for optimal performance you should be looking at fans around 1750-2200rpm, the heatoutput from you components at stock is around 550w of heat, the se rads you have are able to cool around 420w of heat each but thats with fan speeds in access of 1850rpm.

you should be able to feel airflow from the rads a good few inches away, if you dont the fans arnt pushing enough airflow through resuting in you loop getting really hot to the point of shutdown or the pipes could blow off whilst the pc is on, if that happens it'll be a very bad day.

i'd say it's your fans that are the problem as loop order isnt really going to affect temsp that much, have a look at faster rpm fans

Sorry to disagree
Yes 1150rpm is on the lower side fan wise
But I only run mine 900 to 1200 max
1850 to 2200rpm would be severely loud
Ideally maybe they should have got 1500rpm fans so had a bit in reserve
But I will be very surprised if most people in here with custom loops are running 1850~2200rpms

Edit forgot the make of my fans lol
Noise blocker eloops I think
 
Soldato
Joined
28 Aug 2017
Posts
2,755
Location
United Kingdom
Sorry to disagree
Yes 1150rpm is on the lower side fan wise
But I only run mine 900 to 1200 max
1850 to 2200rpm would be severely loud
Ideally maybe they should have got 1500rpm fans so had a bit in reserve
But I will be very surprised if most people in here with custom loops are running 1850~2200rpms


i used to run noctua af-12 fans with the low noise adaptors at 1750 rpm and the fans were silent and cooling on my pe rads was amazing, i had a 3900x and a 2080 ti running sub 70 for the cpu and sub 45 for the gpu both had bigish overclocks on too 4.45ghz on the cpu and 2050mhz on the gpu and 8400mhz for the gpu ram

specs for your build?
 
Soldato
Joined
21 Dec 2019
Posts
6,439
Location
Planet Thanet
i used to run noctua af-12 fans with the low noise adaptor at 1750 rpm and the fans were silent and cooling on my pe rads was amazing, i have a 3900x and a 2080 ti running sub 70 for the cpu and sub 45 for the gpu both had bigish overclocks on too 4.45ghz on the cpu and 2050mhz on the gpu
Will take your word on the noctuas never tried those
Though to be fair those are probably quieter than most would be at 1750rpm

Edit they did 1750rpm at 7v presumably is what the noise adaptor does?
 
Soldato
Joined
28 Aug 2017
Posts
2,755
Location
United Kingdom
Will take your word on the noctuas never tried those
Though to be fair those are probably quieter than most would be at 1750rpm

noctuas are pretty much the best fans you can buy yes the colour is naff but for sheer cooling vs noise they are unbeatable, expensive too but totaly worth it

yeah i used the 7v adaptor cables that come in the box, tbh i did run at the max 2200rpm and they wre still bloody quiet :)
 
Last edited:
Soldato
Joined
21 Dec 2019
Posts
6,439
Location
Planet Thanet
What you use to get the 70c~72c load on the cpu?
Admittedly I haven't really pushed my new 3900x properly yet but it hasn't hit 60c yet doing a load of every day stuff all at once

And yes loop order makes no difference it will equalise/reach equilibrium regardless of order
As long as res is feeding pump

noctuas are pretty much the best fans you can buy yes the colour is naff but for sheer cooling vs noise they are unbeatable, expensive too but totaly worth it

That beige/brown I mean what are their marketing people thinking
Though now can get all black wow not much improvement lol

yeah i used the 7v adaptor cables that come in the box, tbh i did run at the max 2200rpm and they wre still bloody quiet :)
 
Last edited by a moderator:
Soldato
Joined
28 Aug 2017
Posts
2,755
Location
United Kingdom
What you use to get the 70c~72c load on the cpu?
Admittedly I haven't really pushed my new 3900x properly yet but it hasn't hit 60c yet doing a load of every day stuff all at once

And yes loop order makes no difference it will equalise/reach equilibrium regardless of order
As long as res is feeding pump


mainly cinebench r20 and a couple of games would make my chip hit that but normally it would sit in the mid 30's idle that was at 4.45ghz on all cores and setting a 1.345v vcore.

if you leave the cpu to its own devices pbo wont be enabled out of the box so the cpu will run anyware between 3.8 and 4.1ghz, if you go into the bios and enable it then the cpu will clock up to 4.5-4.6ghz (single core) and thats when it runs a bit warmer but nothing too bad tbh
 
Last edited:
Soldato
Joined
21 Dec 2019
Posts
6,439
Location
Planet Thanet
mainly cinebench r20 and a couple of games would make my chip hit that but normally it would sit in the mid 30's idle that was at 4.45ghz on all cores and setting a 1.345v vcore
Think I ran cine bench though may have been r15
The idle should be around 30c plus or minus a bit depending on ambient temp
Though obviously it's load that's important
That's a really good pump/res combo far as I am concerned since got one
Myself so can't see that being an issue
As mentioned 1150rpm is a little low to some but again ek vardars are good I can't see it being those after all point of watercooling is to be cooler but at less noise
Main thing is your gpu seems to be more of an issue than the cpu
Did you definitely bleed all the air out?
No flow meter?
Or didn't try bit of dye dripped into reservoir to see if it shot round the loop properly?
 
Associate
OP
Joined
4 Jan 2020
Posts
19
Think I ran cine bench though may have been r15
The idle should be around 30c plus or minus a bit depending on ambient temp
Though obviously it's load that's important
That's a really good pump/res combo far as I am concerned since got one
Myself so can't see that being an issue
As mentioned 1150rpm is a little low to some but again ek vardars are good I can't see it being those after all point of watercooling is to be cooler but at less noise
Main thing is your gpu seems to be more of an issue than the cpu
Did you definitely bleed all the air out?
No flow meter?
Or didn't try bit of dye dripped into reservoir to see if it shot round the loop properly?

I made sure all the air has been drained from the loop, i have moved it around and ran it on a separate PSU when fitting to ensure it was running with no leaks before powering on so no bubbles in any blocks and the res is almost completely full and this level has not changed since fitting.
I do not have a flow meter but i did try the dye trick and zipped down the tube with no issues at all.
Its defo the GPU that is of the main concern, this is connected to the rad at the top of my case, my thoughts after reading all of your guys comments is that as the airflow is too low, the fans mounted to the front rad is not able to blow through enough air to get fresh air to the top rad to blow through and cool the GPU? Is that the right line of thinking or am i talking out my rear end?
 
Soldato
Joined
28 Aug 2017
Posts
2,755
Location
United Kingdom
Think I ran cine bench though may have been r15
The idle should be around 30c plus or minus a bit depending on ambient temp
Though obviously it's load that's important
That's a really good pump/res combo far as I am concerned since got one
Myself so can't see that being an issue
As mentioned 1150rpm is a little low to some but again ek vardars are good I can't see it being those after all point of watercooling is to be cooler but at less noise
Main thing is your gpu seems to be more of an issue than the cpu
Did you definitely bleed all the air out?
No flow meter?
Or didn't try bit of dye dripped into reservoir to see if it shot round the loop properly?


problem is the 3950x is a higher binned cpu and if you look at reviews it pulls less power than the 3900x so in theory it should run a few degrees cooler than a 3900x, no way under water it should hit over 80 degrees, unless it had a massive overclock on it which it woundn't as pbo is enabled so only one core will go to 4.7ghz, if all cores hit that then yes the cpu should hit over 80 degrees but at that point its close to shutdown anyway
 
Soldato
Joined
28 Aug 2017
Posts
2,755
Location
United Kingdom
I made sure all the air has been drained from the loop, i have moved it around and ran it on a separate PSU when fitting to ensure it was running with no leaks before powering on so no bubbles in any blocks and the res is almost completely full and this level has not changed since fitting.
I do not have a flow meter but i did try the dye trick and zipped down the tube with no issues at all.
Its defo the GPU that is of the main concern, this is connected to the rad at the top of my case, my thoughts after reading all of your guys comments is that as the airflow is too low, the fans mounted to the front rad is not able to blow through enough air to get fresh air to the top rad to blow through and cool the GPU? Is that the right line of thinking or am i talking out my rear end?


if you have no airflow going through the case to push out the heat then the gpu and cpu are going to rocket up temp wise, dont forget at full chat a 2080 ti has 250watts minimum heat output if that heat inst expelled from the rad it will quickly heat up the loop and thus the cpu will run hot
 
Associate
OP
Joined
4 Jan 2020
Posts
19
if you have no airflow going through the case to push out the heat then the gpu and cpu are going to rocket up temp wise, dont forget at full chat a 2080 ti has 250watts minimum heat output if that heat inst expelled from the rad it will quickly heat up the loop and thus the cpu will run hot

Sounds like the low RPM fans may not be up to snuff then possibly?
 
Back
Top Bottom