Overclock.net › Forums › Cooling › Water Cooling › PC Coolant System Fluid Flow
New Posts  All Forums:Forum Nav:

PC Coolant System Fluid Flow

post #1 of 12
Thread Starter 
There is a great article on PC water coolant chemistry at http://www.overclockers.com/pc-water...mistry-part-i/.
Does anyone know of an article at the same level entitled something along the lines of "PC coolant fluid dynamics"? Here is why I am looking for this.

I would like to start building a liquid PC cooling system that for a number of reasons will be external. I would like to bring 1/2" to/ from the PC case and then split it into 4 1/4" parallel flows inside the case to the cooling blocks.
Why? Because:
1) 1/4" tubing is neater, easier to work with inside
2) 1/4" tubing, if handled properly, offers less wind resistance for any residual air cooling.
3) Serial is a single point of failure system.
4) In a GPU-GPU-CPU serial I could end up heating (not cooling) the CPU (perhaps an exaggeration but you get the idea).
5) In parallel it is easier to control issues of high vs low resistance blocks (except possibly at the splitter but I think that will be okay too).

Yes, I know, any of these points can be argued but I think they are more right than wrong. And that good article that I am looking for would discuss this.

I would also love the article to discuss issues of turbulence at junctions and components. (Is it a major issue for the pump and how do I quantify this?) And since I can wish for anything I would ask for thermodynamic equations to optimize coolant flow rate for a given coolant-component temperature differential. (This is the #1 issue right?)

Thanks for any thoughts and help.
post #2 of 12
it has been done before. I tried it but it gains nothing. half inch tube is the best way and in serial. That is the best way to do it
post #3 of 12
each time u branch to a parallel, your flow decreases on the parallel loop.

Each time your flow decreases u lose holding capacity of water.

Each time you lose holding capacity of water, it heats up faster.

A true parallel requires flow reducers and flow monitors to ensure all pathways are getting equal flow.

True parallel should only be attempted at the gpu sector where things are closer to be exact, then branching off in a node. Or the parallel loop should be fed by its own pump at the start of parallel.

Your model of parallel is old thinking, which we grew out of, because we realized in most cases serial is the way to go.
Nadeshiko
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 990X eVGA Classified E759 eVGA GTX 580 x 2 + GTX 550Ti Corsair Dominators 
Hard DriveMonitorPowerCase
Intel G2 x 3 in R0 + C300 256GB - Gaming Gateway 2485W x 2 + Asus 27 LED center AX1200 MM + Pedistool 
  hide details  
Reply
Nadeshiko
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 990X eVGA Classified E759 eVGA GTX 580 x 2 + GTX 550Ti Corsair Dominators 
Hard DriveMonitorPowerCase
Intel G2 x 3 in R0 + C300 256GB - Gaming Gateway 2485W x 2 + Asus 27 LED center AX1200 MM + Pedistool 
  hide details  
Reply
post #4 of 12
equal flow is a huge issue with parallel. in most cases the water would (in one of the blocks) be moving slow enough to have time to boil... which happens pretty quick.

unless you have basically individual loops for every block, fed by a single res.
Unnamed Build
(23 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i7 3770k MAXIMUS V GENE NVIDIA GeForce GTX 780 Hydro Copper Vulcan 
Hard DriveCoolingCoolingCooling
Samsung 1TB 840 SSD Koolance CPU-380i XSPC Dual Bay Reservoir Phobya Xtreme 200mm Radiator 
CoolingCoolingCoolingOS
XSPC EX120 120mm Radiator Silverstone AP181 Fan Gentle Typhoon AP15 Windows 8 Professional x64 
MonitorKeyboardPowerCase
HP LP3065c DasKeyboard Ultimate Model S Seasonic X650 Silverstone TJ08B-E 
MouseMouse PadAudioAudio
Logitech G700 Mionix Propus 380 Audioengine D1 USB DAC Parasound 2125 Amplifier 
AudioAudioAudio
Beyerdynamic DT 770 Pro-80 Closed Studio Headph... Audio-Technica ATH-M50 Definitive Technology Studiomonitor 450 
  hide details  
Reply
Unnamed Build
(23 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i7 3770k MAXIMUS V GENE NVIDIA GeForce GTX 780 Hydro Copper Vulcan 
Hard DriveCoolingCoolingCooling
Samsung 1TB 840 SSD Koolance CPU-380i XSPC Dual Bay Reservoir Phobya Xtreme 200mm Radiator 
CoolingCoolingCoolingOS
XSPC EX120 120mm Radiator Silverstone AP181 Fan Gentle Typhoon AP15 Windows 8 Professional x64 
MonitorKeyboardPowerCase
HP LP3065c DasKeyboard Ultimate Model S Seasonic X650 Silverstone TJ08B-E 
MouseMouse PadAudioAudio
Logitech G700 Mionix Propus 380 Audioengine D1 USB DAC Parasound 2125 Amplifier 
AudioAudioAudio
Beyerdynamic DT 770 Pro-80 Closed Studio Headph... Audio-Technica ATH-M50 Definitive Technology Studiomonitor 450 
  hide details  
Reply
post #5 of 12
Thread Starter 
Thanks all. Good thoughts.
post #6 of 12
High level thermodynamics and fluid modeling really isn't required in computer cooling, there are so damn many inefficiencies...

You don't have to use 1/2" tubing at all, you'll get the same results as 3/8", not sure about 1/4" probably wont make a difference but you'll get a larger pressure drop ofc.

Unfortunately there's no real resource for pressure drop of components at their flowrates or any real consideration for tubing/fitting losses.

The most I know of that has been done is by Martin of http://martinsliquidlab.org/

There's a lot of info there and a spreadsheet for calculating flowrates in loops, but it's not something I would rely on.
    
CPUMotherboardGraphicsRAM
e5645 @ 4.2Ghz P6X58D-E 5450 x2 6Gb Corsair Dominator 1600 
CoolingOSPowerCase
ALPENFOHN K2 Windows 7 Pro Antec Truepower Trio 650W Dell Optiplex 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
e5645 @ 4.2Ghz P6X58D-E 5450 x2 6Gb Corsair Dominator 1600 
CoolingOSPowerCase
ALPENFOHN K2 Windows 7 Pro Antec Truepower Trio 650W Dell Optiplex 
  hide details  
Reply
post #7 of 12
Thread Starter 
Crabid: thanks for pointing me back to Martins. Yes his spreadsheet is not exact but it is a good lead. The point that is coming out here is maintaining a minimum flow rate and balancing the resistance among the parallel branches. The other point is that in the end serial is just simpler.
post #8 of 12
serial is easier because we dont typically have high enough heat wattage for it us to have to distribute fresh water.

As i have said in other threads... its roughly 300W @ water moving 1gpm for it to go up 1C.

300W is more then what a gaming system will bring during a gaming session, unless ur scaled up on GPU's.
Nadeshiko
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 990X eVGA Classified E759 eVGA GTX 580 x 2 + GTX 550Ti Corsair Dominators 
Hard DriveMonitorPowerCase
Intel G2 x 3 in R0 + C300 256GB - Gaming Gateway 2485W x 2 + Asus 27 LED center AX1200 MM + Pedistool 
  hide details  
Reply
Nadeshiko
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 990X eVGA Classified E759 eVGA GTX 580 x 2 + GTX 550Ti Corsair Dominators 
Hard DriveMonitorPowerCase
Intel G2 x 3 in R0 + C300 256GB - Gaming Gateway 2485W x 2 + Asus 27 LED center AX1200 MM + Pedistool 
  hide details  
Reply
post #9 of 12
Quote:
Originally Posted by ThreeColt View Post
Crabid: thanks for pointing me back to Martins. Yes his spreadsheet is not exact but it is a good lead. The point that is coming out here is maintaining a minimum flow rate and balancing the resistance among the parallel branches. The other point is that in the end serial is just simpler.
Yea, that's pretty much it, essentially, a cooling loop doesn't change it's water temp more than 1-2oC over the whole loop, so assuming you have enough radiator to avoid excessive heat buildup in the loop (coolant temps are generally between 5-15oC above ambient depending on the loop) the rest just doesn't really matter.


As mentioned, the best place for parallel is graphics cards in SLI/Crossfire. Graphics cards a hell of a lot easier than CPUs to cool, if for no other reason than they have a hell of a lot more surface area on a full cover block compared to a CPU being something like 40x40mm pushing through 60-70% the wattage of a full GPU card.


So the flow doesn't matter as much (IMO), so splitting it 2 or 3 ways there actually helps the loop by lowering the restriction across those parts by a huge amount compared to series (4 or 9 fold depending on the number of cards), though essentially that just lowers your pump requirements.
    
CPUMotherboardGraphicsRAM
e5645 @ 4.2Ghz P6X58D-E 5450 x2 6Gb Corsair Dominator 1600 
CoolingOSPowerCase
ALPENFOHN K2 Windows 7 Pro Antec Truepower Trio 650W Dell Optiplex 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
e5645 @ 4.2Ghz P6X58D-E 5450 x2 6Gb Corsair Dominator 1600 
CoolingOSPowerCase
ALPENFOHN K2 Windows 7 Pro Antec Truepower Trio 650W Dell Optiplex 
  hide details  
Reply
post #10 of 12
Quote:
Originally Posted by Crabid View Post
So the flow doesn't matter as much (IMO), so splitting it 2 or 3 ways there actually helps the loop by lowering the restriction across those parts by a huge amount compared to series (4 or 9 fold depending on the number of cards), though essentially that just lowers your pump requirements.
actually flow doesnt matter once u've hit around .85-1.2gpm.

Anything over you typically wont notice... and anything under, you will notice.
Nadeshiko
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 990X eVGA Classified E759 eVGA GTX 580 x 2 + GTX 550Ti Corsair Dominators 
Hard DriveMonitorPowerCase
Intel G2 x 3 in R0 + C300 256GB - Gaming Gateway 2485W x 2 + Asus 27 LED center AX1200 MM + Pedistool 
  hide details  
Reply
Nadeshiko
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 990X eVGA Classified E759 eVGA GTX 580 x 2 + GTX 550Ti Corsair Dominators 
Hard DriveMonitorPowerCase
Intel G2 x 3 in R0 + C300 256GB - Gaming Gateway 2485W x 2 + Asus 27 LED center AX1200 MM + Pedistool 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Water Cooling
Overclock.net › Forums › Cooling › Water Cooling › PC Coolant System Fluid Flow