Custom Water Cooling Loop not Cooling?

Poozle

Reputable
May 9, 2015
328
0
4,810
So I just finished my new water cooling loop (first time) and it seems to be failing to cool my GTX 980 TI and I7 5820k. Here is the order
Res>Pump>360mm RAD>240mm RAD> CPU> GPU> Res. My temps while idle are at 50 and will crash the computer if stressed at all. Help as to why?? thanks!
Edit: its only the GTX 980 TI that is over heating. I was able to overclock my I7 5820k by an additional 1.5 GHZ (4.8) at 1.270V while staying at 60 Degrees during prime 95
 
Solution
1. I think we are talking a "D5 pump top" rather than a pump. A PWM controlled pump must have a cable (usually 2 wires) connected to the 4 pin CPU header to provide speed control. You should be able to control pump speed curves via the utility provided with your MoBo. 1450 rpm is quite low.

As you can see here, 1440 rpm is about 25% PWM

ex-pmp-214a.jpg


At 20% PWM, we see here that the pump delivers 0 gpm...

swiftechmcp655pwm-t16.png


.. you'll need to operate between 60% (2800 rpm) and 100% of full speed (4700 rpm) to accomplish anything...
Why do you have it pumping into a Rad then into a second Rad?
You should try using the below:
Res -> pump -> cpu -> radiator -> GPU -> radiator #2 -> res

That way cool liquid goes to the CPU, then gets cooled, then to the GPU, then cooled once more back to the res.
You also may wish to check if your pump is capable of the amount of loop you are sending it through.
 
There is no order requirement other than having the res upstream of the pump. Even at low flow rates, the temp thru the loop won't change by more than 3C.

Lets say the GPU adds 3.5C and CPU 1.5C

Res>Pump>360mm RAD>240mm RAD> CPU> GPU> Res

30C + 0 -3 C - 2C + 1.5 C + 3.5C = 30C again

1. What is the rpm of the pump ? At that rpm, what does the flow chart say it's pumping ?

2. What are the components in use ? Do you have a sketch of tube routing ?

3. What steps did you take to clean the rads before filling ?

4. Did you install the water block yourself ? Was TIM, and thermal pads where applicable, applied to all GPU, memory and VRM surfaces ?

5. Did you use clear blocks ? Can you see any debris lodged in the micro channels ?


Res -> pump -> cpu -> radiator -> GPU -> radiator #2 -> res

30C + 0 + 1.5 - 3.0 + 3.5 - 2.0 = 30C again

 

Poozle

Reputable
May 9, 2015
328
0
4,810


Because of aesthetics. It doesn't look good doing that. Besides I have heard loop order doesn't matter?
 
The impact of loop order is minimal ... as I said above, have never actually measured more than 3C or so and that shrinks as flow rate goes up. Theoretically, having the radiators before your CPU presents the best situation possible as this will be the lowest temperature location. Your CPU is the critical component here as if your CPU is 78C you might say OK, but at 80C you might be unhappy.

Splitting them to put one before GPU and one before CPU isn't really advantageous as the GPU will be ridiculously low in temp. Typical CPU temps are in the mid to hi 70s and typical GPU temps are in the high 30s because those GPU blocks have 7 or 8 times the mass and contact area to work with so heat transfer is much more efficient.

So if your GPU temps go from 39C to 41C do you really care ?

For efficacy of plumbing, I find this to be the most efficient:

Reservoir => Pump => Bottom Rad => Tee => split Flow to GPU 1 and 2 => Tee => Top rad => CPU Block => MoBo Block => Reservoir

It works out nicely as biggest rad is right before CPU and that's where my only temperature concern lies.
 

Poozle

Reputable
May 9, 2015
328
0
4,810

I do not know the RPM of the pump. its EKWB D5 pump listed on their website I guess. I did not clean out the rads and i am using clear everything. I see no clogged GPU, CPU, or tubing. I am using MX-4 thermal paste. I just disabled the loop and GPU block and put it back together. new order is Res>pump>Rad 360mm>CPU>240mm rad> GPU> res. Now i can not even find out GPU temps as Afterburner is not showing anything and NZXT CAM will not launch. The computer is working though. I know this because if the GPU wasnt, i could not even go to bios as I7 Extreme edition CPU have no Intel iGPU. A

Edit: yes I did apply all thermal pads where said to be applied. The pump RPM is 1450 to start and is PWM. capable of reaching 4800 RPM. is there any way to control the pump rpm?
 
1. I think we are talking a "D5 pump top" rather than a pump. A PWM controlled pump must have a cable (usually 2 wires) connected to the 4 pin CPU header to provide speed control. You should be able to control pump speed curves via the utility provided with your MoBo. 1450 rpm is quite low.

As you can see here, 1440 rpm is about 25% PWM

ex-pmp-214a.jpg


At 20% PWM, we see here that the pump delivers 0 gpm...

swiftechmcp655pwm-t16.png


.. you'll need to operate between 60% (2800 rpm) and 100% of full speed (4700 rpm) to accomplish anything.

https://martinsliquidlab.wordpress.com/2012/12/25/swiftech-mcp-655-pwm-drive-pump-review/4/

2. I am unfamiliar with much or the wording you are using regarding your components. It is rare to see a 5820k paired with a 980 Ti as normally one wouldn't choose a 5820k unless they were using 3 GPUs.

3. Rads should be cleaned prior to first use.... this involves flushing with water and a bot of shaking to remove solid particles as well is cleaning with acidic and basic solutions to get rid of all the chemicals used in the fabrication process.

4. What are you using as coolant ?

5. The component order is not your problem.

6. You can view GPU TDP, temps, load % etc in Furmark as well as GPU_z. Do not use twp monitoring utilities at the same time and of one isn't opening, make sure it's still not running via Task manager.
 
Solution

Poozle

Reputable
May 9, 2015
328
0
4,810
OK I FIXED IT! Here is what I did.
1. I re-did the order of the pump and it DID make a small difference, but still very noticeable. My idles went from 32-36 (OC) to 26-28 (OC) on my cpu. The stress test made it stay at about 59 average (Aida 64) vs 64-65 average. The gpu I had to re install Nvidia drivers and now it runs at 20-28 idle and 39-41 under load (OC of course to 1520MHZ core and 3800 Mem) Thanks for the help.
Here is what else I did.
2. Found some old TIM on the GPU chip sides and wiped it off with Q-tips
3. took apart loop? (not sure if this or re ordering the loop truly helped, im just listing.
4.New Drivers including deleting all nvidia old drivers and re installing which is why it wasnt reconized by MSI afterburner, GPU-Z ETC. along with a re install of Java.