Cooling Mining Rigs with Liquid Chiller

Mar 15, 2018
3
0
10
Hi All,

I am new in water cooling, I have set a normal water cooling loop for 5 GPUs and here what I have used;

1. 5x Gigabyte 1080 Ti Waterforce WB Xtreme Edition 11G
2. 1x Thermaltake RL480 Radiator
3. 4x Arctic F12 PWM Fan
4. 1x Thermaltake PR22-D5 300ml Pump/Reservoir Combo
5. XSPC fittings
6. Clear Vinyl Tubing 3/8ID X 1/2OD

The setup was as follow;

Pump/Reservoir >>> 5 GPUs (Connected in serial) >>> Rad (with 4 fans attached)

I have used normal distilled water
The ambient temperature was around 20C (dropped by A/C)

The setup worked fine but the GPUs temperature was around 60+C (the lowest) to 70+C (the highest)

It was winter but in summer the outside temperature will go up to 50C, all temps will rise few degrees and thus I will run into overheat.

However, all pipes seems to be stressed and I didn't like that.

and, I wanted to overclock and add more GPUs.


So, I upgraded the water loop and now I am running the following;

1. Added another Rad (in serial)
2. Replace the 4 fans with stronger ones (old 74cfm/fan, new 250cfm/fan)
3. Added circulating chiller/pump (17L/m at 2.9psi) (in serial)

The new setup is;

Chiller >>> GPUs >> Pump/Reservoir >> Rads

So the theory here, The Chiller will chill the water to 18C and will cutoff till it rises to 25C, the water will go to the GPU and will reach the radiator at very hot temp and low pressure, the pump before the rad will make sure the water get more pressure to pass through the radiators and back to the chiller and nearly room temp max, so the chiller will make sure it is between 18C to 25C, and so on.

This worked very well so I added another GPU (making total 6) and now GPUs temps are 47 50 54 56 62 67.

I run into a problem where reservoir build up air so I have to open the cap from time to time to have the air out, as soon as I open the cap the reservoir refills so I have no chance to add water if I should, any one knows why ?

However, I plan to make this bigger as follow;


|--- 5 GPUs ---|
|--- 5 GPUs ---|
|--- 5 GPUs ---|
|--- 5 GPUs ---|
Chiller >| |>-- Pump >-- Rads
|--- 5 GPUs ---|
|--- 5 GPUs ---|
|--- 5 GPUs ---|
|--- 5 GPUs ---|


I think this will work because the chiller 17L/m at 2.9PSI, means each branch will have 2L/m at 2.9Psi which I think it is enough for the water to go through all GPUs, the temps and the pressure will leave each branch at the same temp and pressure as it was only one branch, do you think this will work ?
 
Solution
What do you mean by 'pipes seem stressed'?

Sounds like your reservoir issue might be due to flow restriction...are you running all of these in serial or in parallel? I see you are using a single D5 pump, but even that can have issues with excessive restriction and GPU blocks tend to be a bit more restrictive than most watercooling components.

If you're going to be adding that many GPUs to this loop, you're going to need to re-plan how this will function. One of these pumps will likely not handle the restriction in a design like this, even if it is all parallel. In fact, you might have to work to configure independent, smaller 'loops' like the one you currently have for each of the 5-GPU blocks you show above. Using 1 pump for 40...


I would guess that your mystery air is due to either dissolved gases in the water or a leak.

I just thought of something that you ma use. Since the temperature gets very hot there. How about plumbing a large reservoir and placing it in a small refrigerator . That would significantly reduce the temperature of your cooling water.
 
What do you mean by 'pipes seem stressed'?

Sounds like your reservoir issue might be due to flow restriction...are you running all of these in serial or in parallel? I see you are using a single D5 pump, but even that can have issues with excessive restriction and GPU blocks tend to be a bit more restrictive than most watercooling components.

If you're going to be adding that many GPUs to this loop, you're going to need to re-plan how this will function. One of these pumps will likely not handle the restriction in a design like this, even if it is all parallel. In fact, you might have to work to configure independent, smaller 'loops' like the one you currently have for each of the 5-GPU blocks you show above. Using 1 pump for 40 GPUs isn't going to work. Even in parallel, 1 of those D5 pumps is probably limited to 5 GPUs each, and that's really pushing it. If you are wanting this entire setup to be a single, large parallel setup, you are going to need to consider something along the lines of a more commercial or industrial pump and then branching your tubing from larger sizes down into smaller manifolds. You also have to consider that high-pressure pumps can also generate too much force which flow restriction might cause fitting or tubing failure.

I just thought of something that you ma use. Since the temperature gets very hot there. How about plumbing a large reservoir and placing it in a small refrigerator . That would significantly reduce the temperature of your cooling water.

This isn't a good option. Commercial refrigerators and freezers are only meant to remove thermal load once, and then maintain temperature. A computer is continually adding thermal load, meaning the compressor would continuously run...causing it to fail. You then have an insulated box where thermal load continues to build.
 
Solution
FWIW in automotive cooling systems older cars would lose water when the water got hot and expanded, when they cooled air would get sucked back in. The cure was a cap that sent the extra water to a catch can with the end of the hose submerged so the water would be drawn back in.
 
It sounds like in this instance, there is either a leak causing air to be introduced, or there is enough restriction in the loop that the pump is causing a displacement of coolant by pressurizing the air, hence what is seen when the lid is opened and the coolant level normalizes.
 
I mean the pipes are stretched like they will explode and really hot.
Also the chiller is industrial type have a built in pump that gives 17L/m, so I am using two pumps.

And thanks for all your answer
 
That sounds like the end result of excessive restriction and poor flow. Tubing should never visibly swell, collapse or otherwise become mishapen. Also, if tubing is very, very warm like that, that means you are likely not getting proper flow of coolant.
 
I know what I did wrong. The position of the pump was wrong. As the loop has two pump I shouldn’t put them in different part of the loop, this create difference in pressure, exactly as some said here, the reason why I have water in reservoir reduces.

I added another GPU and it working fine.

So now I make it like this,
2 Pumps >>> 7 GPUs >>> 2 Rads
And my temps are all in 40s C expect one in mid 50s C

When I overclock them the max I get 64C.

So the solution is, if you want to have a loop in series you have to put pumps together, blocks together, rads together.

I will split the loop into two parallel 5+5 GPUs and let you know what happen