Overclock.net › Forums › Cooling › Specialized Cooling › Peltiers / TEC › My math says you can't cool a reservoir with a sane # of TECs
New Posts  All Forums:Forum Nav:

My math says you can't cool a reservoir with a sane # of TECs

post #1 of 9
Thread Starter 
Spoiler alert***
From what I can tell, I would need 158 peltier coolers to cool the water 11C once the water is flowing.

I got all excited the first time I saw this

Then I realized the above had almost zero thermal load.
I have an extra PSU and a CORSAIR CAFA70 120mm Dual-Fan CPU cooler, plus some old heat syncs and was day dreaming about building something similar to this

Great build thread and experiment documentation here: http://forums.vr-zone.com/project-lo...r-worklog.html
The above was built by fusionmkx. His results with 600g of non-circulating water and two 138W@15V TECs powered by separate 12V PSUs. The fan he used to cool the air coolers was a monster 2.8 Amp fan that actually kept the hot sides cool.
WATER TEMP CHANGE OVER TIME

To be fair the results above were achieved without any insulation, or proper clamping load on the TECs. But it still cooled the water, and would be along the lines of what I was planning to make.
So I started doing the math to see which TEC I would need. Unfortunately my math says that it's not which one, but how many...

Please tell me I'm wrong.

I have a Thermaltake Bigwater which has a pump that flows 500 L/hr, but per http://www.overclockers.com/king-flowmeter/
Flow rates can quickly drop to 12.7% of free-flow in a system.

Free-flow, 12.7% of free-flow
500, 63.5 L/hr
132, 16.76 gal/hr
0.138, 0.0175 L/second
138, 17.5 g/second

I was planning to have the res filled by the stock radiator outlet where I'm guessing the coolant will generally be about 35C. And to have the pump draw directly from the TEC cooled res. So the question becomes how many watts will it take to cool 17.5 g of water/second?

I found a nice formula from an HVAC engineer:

"Here is a simple formula to calculate load when water is the medium:
Q = 500*GPM*(Delta T)

Q = load in BTUh*
500 = constant
GPM = flow in gallons per minute
Delta T = temperature difference in Fahrenheit between the supply water and the return water
*To get BTUh from Watts, multiply watts by 3.412"

And a nice dew point chart on Wikipedia. The red parallelogram represents the likely humidity and temps in my apartment/case.

On a bad day I could see condensation as warm as 24C!
Best case scenario is I wouldn't see it till 12C.

But lets only shoot for only 24 just to be conservative and safe from condensation.
With the water returning from my radiator at 35, the delta T would be 11C

Q(in BTHh) = 500(a constant) * 0.368(gallons/min) * deltaT F *1.8(convert to C) = Watts * 3.412(conversion factor)

Q = 500 * 0.368 * 1.8 * deltaT = Watts * 3.412
331.2 * dT = Watts * 3.412

dTF of 11C would take me from 35C to 24C and would require 1067 watts!
dTF of 23C would take me from 35C to 12 and would require 2233 watts!

I was thinking that prolonged idle times might be the saving grace of the design but, now I doubt it.
Once the water is no longer stagnant the heat load of the water gets crazy!
Looking at fusionmkx’s temp drop over time above you see he dropped from ~29.3 C to ~22C in the first 30 min of cooling. If you halve those results to represent a single pelt design you achieve a delta T of ~7.3C/60min/600g of water.
that's 0.12166 C/min and 0.0020277 C/sec/600g of water

With my middle of the road pump producing 17.5g/sec of water I could hope for a slightly better 0.0695 C steady-state reduction in temp.

For this to be worthwhile I would really want at least a 11C reduction.
11C / 0.0695 C = 158


Presuming that I had the same efficiency as fusionmkx, I would need 158 peltier coolers to cool the water 11C once the water is flowing. And an even more rediculous 331 TECs to drop the coolant 23 C, from 35 to 12C!

Bogus
Edited by Gigapunk - 10/11/11 at 7:54am
    
CPUMotherboardGraphicsRAM
Pentium Dual-Core e6300 Wolfedale-2M PEGATRON IPIBL-LB (Benicia) Intel G33/G31 internal, but ordering a 6850 soon 4x2 
Hard DriveOptical DriveOSMonitor
750 and 80 gig 7200rpm ATAPI DVD A DH16A6L Windows 7 Ultimate 64bit DELL E173FP [Monitor] (17.1" 
PowerCase
350W from a Dell Dimensions 8400 HP Pavillion 96120f 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Pentium Dual-Core e6300 Wolfedale-2M PEGATRON IPIBL-LB (Benicia) Intel G33/G31 internal, but ordering a 6850 soon 4x2 
Hard DriveOptical DriveOSMonitor
750 and 80 gig 7200rpm ATAPI DVD A DH16A6L Windows 7 Ultimate 64bit DELL E173FP [Monitor] (17.1" 
PowerCase
350W from a Dell Dimensions 8400 HP Pavillion 96120f 
  hide details  
Reply
post #2 of 9
your math seems solid. That is why peltier coolers are extremely impractical. Though most people with TEC setups put the cooler directly on the chip and use a water block on the hot side to dissipate the heat. I used to have a very old celeron rig with a 45w TEC that ran sub ambient (about 10C) but that is only because the celeron has a TDP of 10w so for a sig rig TECs are pretty pointless because assuming a perfect world peltier effect has about a 5% efficiency so a 125w processor let say produces 120 watts of heat so to dissipate just that and run at room temp would require 2400 watts of electrical energy into the TEC. then the rest of that power in the tec goes to heat on the hot side which then is 234 watts of heat on the hot side. So unless you have a 3kW PSU peltier coolers are quite useless.
Sandy Thunder
(18 items)
 
QuickSilver X1
(15 items)
 
 
CPUMotherboardGraphicsRAM
i7-3930k C1 4.6GHz @1.344v Sabertooth X79 2x EVGA GTX 560 Ti SC SLI 900/2106 Dominator 1866 16GB 9-10-9-27 
Hard DriveHard DriveCoolingOS
Crucial M4 64GB 'Cuda 320GB XSPC Rasa 750 RX240 push/pull + Alphacool UT60 ... Win 7 Pro x64 
MonitorMonitorMonitorKeyboard
Asus VE198D 19" Samsung 920NW 19" Samsung 906BW 19" Logitech G510s 
PowerCaseMouseMouse Pad
Corsair HX850M Carbide 400R Cyborg M.M.O. 7 Corsair MM200 
AudioOther
Bose Music Monitor Nuforce Icon HDP 
CPUMotherboardGraphicsRAM
Phenom II x6 1090T BE Sabertooth 990FX Radeon HD 5770 1GB Corsair Veangence Blue 
Hard DriveOptical DriveCoolingOS
WD Caviar Black 1TB  Lite ON 16x DVD-R/RW ThermalRight Silver Arrow Windows 7 Professional 64 bit 
MonitorKeyboardPowerCase
Syncmaster 921BW 19" & Syncmaster 965 19" and S... Apple aluminum wired keyboard Corsair TX650 Silverstone TJ05 with window and thermal contro... 
MouseMouse PadAudio
logitech mx air None Monsoon MM-700 
  hide details  
Reply
Sandy Thunder
(18 items)
 
QuickSilver X1
(15 items)
 
 
CPUMotherboardGraphicsRAM
i7-3930k C1 4.6GHz @1.344v Sabertooth X79 2x EVGA GTX 560 Ti SC SLI 900/2106 Dominator 1866 16GB 9-10-9-27 
Hard DriveHard DriveCoolingOS
Crucial M4 64GB 'Cuda 320GB XSPC Rasa 750 RX240 push/pull + Alphacool UT60 ... Win 7 Pro x64 
MonitorMonitorMonitorKeyboard
Asus VE198D 19" Samsung 920NW 19" Samsung 906BW 19" Logitech G510s 
PowerCaseMouseMouse Pad
Corsair HX850M Carbide 400R Cyborg M.M.O. 7 Corsair MM200 
AudioOther
Bose Music Monitor Nuforce Icon HDP 
CPUMotherboardGraphicsRAM
Phenom II x6 1090T BE Sabertooth 990FX Radeon HD 5770 1GB Corsair Veangence Blue 
Hard DriveOptical DriveCoolingOS
WD Caviar Black 1TB  Lite ON 16x DVD-R/RW ThermalRight Silver Arrow Windows 7 Professional 64 bit 
MonitorKeyboardPowerCase
Syncmaster 921BW 19" & Syncmaster 965 19" and S... Apple aluminum wired keyboard Corsair TX650 Silverstone TJ05 with window and thermal contro... 
MouseMouse PadAudio
logitech mx air None Monsoon MM-700 
  hide details  
Reply
post #3 of 9
Quote:
Originally Posted by DSP1 View Post
your math seems solid. That is why peltier coolers are extremely impractical. Though most people with TEC setups put the cooler directly on the chip and use a water block on the hot side to dissipate the heat. I used to have a very old celeron rig with a 45w TEC that ran sub ambient (about 10C) but that is only because the celeron has a TDP of 10w so for a sig rig TECs are pretty pointless because assuming a perfect world peltier effect has about a 5% efficiency so a 125w processor let say produces 120 watts of heat so to dissipate just that and run at room temp would require 2400 watts of electrical energy into the TEC. then the rest of that power in the tec goes to heat on the hot side which then is 234 watts of heat on the hot side. So unless you have a 3kW PSU peltier coolers are quite useless.
I don't really know where your getting your information, but I can keep a 2500k @4.6ghz @ 15ºc or -10dt using only 150w power draw.

http://www.overclock.net/peltiers-te...thout-tec.html

With this setup I dissipate 175w of power at -20dt using only 150w of power with a cop of 1.16.
Edited by Mindchi|l - 10/11/11 at 9:12am
post #4 of 9
Quote:
Originally Posted by Gigapunk View Post
Spoiler alert***
From what I can tell, I would need 158 peltier coolers to cool the water 11C once the water is flowing.

I got all excited the first time I saw this

Then I realized the above had almost zero thermal load.
I have an extra PSU and a CORSAIR CAFA70 120mm Dual-Fan CPU cooler, plus some old heat syncs and was day dreaming about building something similar to this

Great build thread and experiment documentation here: http://forums.vr-zone.com/project-lo...r-worklog.html
The above was built by fusionmkx. His results with 600g of non-circulating water and two 138W@15V TECs powered by separate 12V PSUs. The fan he used to cool the air coolers was a monster 2.8 Amp fan that actually kept the hot sides cool.
WATER TEMP CHANGE OVER TIME

To be fair the results above were achieved without any insulation, or proper clamping load on the TECs. But it still cooled the water, and would be along the lines of what I was planning to make.
So I started doing the math to see which TEC I would need. Unfortunately my math says that it's not which one, but how many...

Please tell me I'm wrong.

I have a Thermaltake Bigwater which has a pump that flows 500 L/hr, but per http://www.overclockers.com/king-flowmeter/
Flow rates can quickly drop to 12.7% of free-flow in a system.

Free-flow, 12.7% of free-flow
500, 63.5 L/hr
132, 16.76 gal/hr
0.138, 0.0175 L/second
138, 17.5 g/second

I was planning to have the res filled by the stock radiator outlet where I'm guessing the coolant will generally be about 35C. And to have the pump draw directly from the TEC cooled res. So the question becomes how many watts will it take to cool 17.5 g of water/second?

I found a nice formula from an HVAC engineer:

"Here is a simple formula to calculate load when water is the medium:
Q = 500*GPM*(Delta T)

Q = load in BTUh*
500 = constant
GPM = flow in gallons per minute
Delta T = temperature difference in Fahrenheit between the supply water and the return water
*To get BTUh from Watts, multiply watts by 3.412"

And a nice dew point chart on Wikipedia. The red parallelogram represents the likely humidity and temps in my apartment/case.

On a bad day I could see condensation as warm as 24C!
Best case scenario is I wouldn't see it till 12C.

But lets only shoot for only 24 just to be conservative and safe from condensation.
With the water returning from my radiator at 35, the delta T would be 11C

Q(in BTHh) = 500(a constant) * 0.368(gallons/min) * deltaT F *1.8(convert to C) = Watts * 3.412(conversion factor)

Q = 500 * 0.368 * 1.8 * deltaT = Watts * 3.412
331.2 * dT = Watts * 3.412

dTF of 11C would take me from 35C to 24C and would require 1067 watts!
dTF of 23C would take me from 35C to 12 and would require 2233 watts!

I was thinking that prolonged idle times might be the saving grace of the design but, now I doubt it.
Once the water is no longer stagnant the heat load of the water gets crazy!
Looking at fusionmkx’s temp drop over time above you see he dropped from ~29.3 C to ~22C in the first 30 min of cooling. If you halve those results to represent a single pelt design you achieve a delta T of ~7.3C/60min/600g of water.
that's 0.12166 C/min and 0.0020277 C/sec/600g of water

With my middle of the road pump producing 17.5g/sec of water I could hope for a slightly better 0.0695 C steady-state reduction in temp.

For this to be worthwhile I would really want at least a 11C reduction.
11C / 0.0695 C = 158


Presuming that I had the same efficiency as fusionmkx, I would need 158 peltier coolers to cool the water 11C once the water is flowing. And an even more rediculous 331 TECs to drop the coolant 23 C, from 35 to 12C!

Bogus
You don't need anywhere near that many tecs. Generelly we aren't trying to instantly drop our water temps with one water turnover. your better off taking the radiator out of your cold loop and then doing multiple water turns to drop your temps.
post #5 of 9
Thread Starter 
Quote:
Originally Posted by Mïndçhì|l View Post
I don't really know where your getting your information, but I can keep a 2500k @4.6ghz @ 15ºc or -10dt using only 150w power draw.

http://www.overclock.net/peltiers-te...thout-tec.html

With this setup I dissipate 175w of power at -20dt using only 150w of power with a cop of 1.16.
That Arqtic Cooler is pretty cool. It has a high quality 62mm TEC that can supposedly cool 400W. 400W is way more heat than any chip can produce, that's why it works so well. But 400W is also WAY less than the heat load of even a small amount of flowing water.
    
CPUMotherboardGraphicsRAM
Pentium Dual-Core e6300 Wolfedale-2M PEGATRON IPIBL-LB (Benicia) Intel G33/G31 internal, but ordering a 6850 soon 4x2 
Hard DriveOptical DriveOSMonitor
750 and 80 gig 7200rpm ATAPI DVD A DH16A6L Windows 7 Ultimate 64bit DELL E173FP [Monitor] (17.1" 
PowerCase
350W from a Dell Dimensions 8400 HP Pavillion 96120f 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Pentium Dual-Core e6300 Wolfedale-2M PEGATRON IPIBL-LB (Benicia) Intel G33/G31 internal, but ordering a 6850 soon 4x2 
Hard DriveOptical DriveOSMonitor
750 and 80 gig 7200rpm ATAPI DVD A DH16A6L Windows 7 Ultimate 64bit DELL E173FP [Monitor] (17.1" 
PowerCase
350W from a Dell Dimensions 8400 HP Pavillion 96120f 
  hide details  
Reply
post #6 of 9
Quote:
Originally Posted by Gigapunk View Post
That Arqtic Cooler is pretty cool. It has a high quality 62mm TEC that can supposedly cool 400W. 400W is way more heat than any chip can produce, that's why it works so well. But 400W is also WAY less than the heat load of even a small amount of flowing water.
The thing about trying to cool water with tecs is that it take multiple passes of a chiller to cool the water. Once you have the water cooled, you only need enough qmax to remove the heat produced by your cpu. You only need 1 or 2 400watt qmax tecs to cool any modern cpu whether it is used in a direct die or chiller.

For example if you had a 400w qmax tec and you did 10 passes of your water through the chiller, you potentially just removed 4000watts of heat from the water. More passes of water through your chiller the more heat you can remove with a single tec. This is the main reason you see tec chillers that resemble heat exchangers instead of fins sitting in a bowl of water.
Edited by Mindchi|l - 10/11/11 at 10:43am
post #7 of 9
Thread Starter 
Quote:
Originally Posted by Mïndçhì|l View Post
The thing about trying to cool water with tecs is that it take multiple passes of a chiller to cool the water. Once you have the water cooled, you only need enough qmax to remove the heat produced by your cpu. You only need 1 or 2 400watt qmax tecs to cool any modern cpu whether it is used in a direct die or chiller.

For example if you had a 400w qmax tec and you did 10 passes of your water through the chiller, you potentially just removed 4000watts of heat from the water. More passes of water through your chiller the more heat you can remove with a single tec.
That makes sense. 800w (400 x 2) isn't too far off from my quoted 1167w to get a decent drop in a single pass.

The problem is that the contraption above is only getting a fraction of its rated watts into the water.

From the quoted build thread:

"Now this is a little bit disappointing. The voltages of the two pelts were measured to be 11.25v and 11.63v, which is rather low, the PSUs just cannot provide enough power to the pelts, being non-ohmic devices, it's hard to measure the current drawn without a proper galvanometer. But based on the ohmic-resistance without power, it's about 5-8ohms for the two pelts used. So current drawn is estimated to be about 5-6A. That works out to a disappointing power output of 16watts."

I don't get 16 watts, but I still get a fraction of the original 138watts of the individual pelts.

watts = amps x volts = 6 A * 11.63 V = 69.78 watts
    
CPUMotherboardGraphicsRAM
Pentium Dual-Core e6300 Wolfedale-2M PEGATRON IPIBL-LB (Benicia) Intel G33/G31 internal, but ordering a 6850 soon 4x2 
Hard DriveOptical DriveOSMonitor
750 and 80 gig 7200rpm ATAPI DVD A DH16A6L Windows 7 Ultimate 64bit DELL E173FP [Monitor] (17.1" 
PowerCase
350W from a Dell Dimensions 8400 HP Pavillion 96120f 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Pentium Dual-Core e6300 Wolfedale-2M PEGATRON IPIBL-LB (Benicia) Intel G33/G31 internal, but ordering a 6850 soon 4x2 
Hard DriveOptical DriveOSMonitor
750 and 80 gig 7200rpm ATAPI DVD A DH16A6L Windows 7 Ultimate 64bit DELL E173FP [Monitor] (17.1" 
PowerCase
350W from a Dell Dimensions 8400 HP Pavillion 96120f 
  hide details  
Reply
post #8 of 9
Quote:
Originally Posted by Gigapunk View Post
That makes sense. 800w (400 x 2) isn't too far off from my quoted 1167w to get a decent drop in a single pass.

The problem is that the contraption above is only getting a fraction of its rated watts into the water.

From the quoted build thread:

"Now this is a little bit disappointing. The voltages of the two pelts were measured to be 11.25v and 11.63v, which is rather low, the PSUs just cannot provide enough power to the pelts, being non-ohmic devices, it's hard to measure the current drawn without a proper galvanometer. But based on the ohmic-resistance without power, it's about 5-8ohms for the two pelts used. So current drawn is estimated to be about 5-6A. That works out to a disappointing power output of 16watts."

I don't get 16 watts, but I still get a fraction of the original 138watts of the individual pelts.

watts = amps x volts = 6 A * 11.63 V = 69.78 watts
I found that less water and more turns in the cold loop returns much better results. Here are a couple threads of chiller builds I did for some friends

http://www.overclock.net/peltiers-te...er-24-7-a.html

http://www.overclock.net/peltiers-te...ontroller.html

As you can see one used 3 tecs, and the other used 2 tecs and cooled both cpu's adequately. both kept the water at 10ºc or less at full cpu utilization.

I would say thats a lot less than the 158 peltier coolers you came up with
Edited by Mindchi|l - 10/11/11 at 11:03am
post #9 of 9
Thread Starter 
Quote:
Originally Posted by Mïndçhì|l View Post
I found that less water and more turns in the cold loop returns much better results. Here are a couple threads of chiller builds I did for some friends

http://www.overclock.net/peltiers-te...er-24-7-a.html

http://www.overclock.net/peltiers-te...ontroller.html

As you can see one used 3 tecs, and the other used 2 tecs and cooled both cpu's adequately. both kept the water at 10ºc or less at full cpu utilization.

I would say thats a lot less than the 158 peltier coolers you came up with
That's really slick. I think that the waterblocks must be a whole bunch more efficient than my soaking radiator idea.
    
CPUMotherboardGraphicsRAM
Pentium Dual-Core e6300 Wolfedale-2M PEGATRON IPIBL-LB (Benicia) Intel G33/G31 internal, but ordering a 6850 soon 4x2 
Hard DriveOptical DriveOSMonitor
750 and 80 gig 7200rpm ATAPI DVD A DH16A6L Windows 7 Ultimate 64bit DELL E173FP [Monitor] (17.1" 
PowerCase
350W from a Dell Dimensions 8400 HP Pavillion 96120f 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Pentium Dual-Core e6300 Wolfedale-2M PEGATRON IPIBL-LB (Benicia) Intel G33/G31 internal, but ordering a 6850 soon 4x2 
Hard DriveOptical DriveOSMonitor
750 and 80 gig 7200rpm ATAPI DVD A DH16A6L Windows 7 Ultimate 64bit DELL E173FP [Monitor] (17.1" 
PowerCase
350W from a Dell Dimensions 8400 HP Pavillion 96120f 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Peltiers / TEC
Overclock.net › Forums › Cooling › Specialized Cooling › Peltiers / TEC › My math says you can't cool a reservoir with a sane # of TECs