|
|
Xtreme Cooling LN2, Dry Ice, Peltiers, etc... All the usual suspects |
Thread Tools |
03-18-2004, 12:25 PM | #1 |
Cooling Neophyte
Join Date: Mar 2004
Location: Southeastern USA
Posts: 5
|
How do you figure power requirements?
I'd like to use a peltier system to cool both my GPU and CPU; I was looking at the Swiftech units, and trying to reverse-engineer them so I can make it myself a little cheaper; if not that I'd just get their GPU and CPU waterblock/pelt combination units and do the rest myself. Anyway, they sell the meanwell S320-12 PSU; and looking at the specs it looks like it does 25a @ 12v? I'm assuming this would only power one of their 226w 25a pelts. First question: Shouldn't the PSU produce a little more amps than what the pelt is rated for, or should they be equal?
Now I'd like to run another 80w/8a pelt for my GPU; can I wire it in series with the CPU pelt? If so, how do I figure power requirements? Is it as simple as 8a+25a=33a, so I just need a PSU capable of 33a? or does running them in series make it a 24v circuit? |
03-18-2004, 01:39 PM | #2 | |
Cooling Savant
Join Date: May 2003
Location: Okotoks, A.B. Canada
Posts: 726
|
Quote:
in series you add up the voltage... in paralel you add up the amps. so to run in parallel and you had 8a & 25a u would need a 12volt rated @ 33a to run in series is a hole diffrent thing. i don't remeber the math anymore but i figure you would be getting from a 24 volt supply 18 volts across the 25a tec and 6 volts across the 8a tec... i think... if i'm wrong.. flame away guys. my 2¢ worth
__________________
"Great spirits have always encountered violent opposition from mediocre minds" - (Einstein) |
|
03-18-2004, 01:41 PM | #3 | |
Put up or Shut Up
Join Date: Dec 2001
Location: Spokane WA
Posts: 6,506
|
Quote:
If the specs on the TEC are 25A @ 12V then you want a PS at that rating. Being the 226W pelt is 25A @ 12V that means it is wasting 74watts and shows the inefficency of TEC's as 25A times 12V = 300watts. That will use up that meanwell PS all on it's own if you want to run that pelt at full power. You can splice in the 80W TEC into that supply aswell but it will lower the peformance of the 226watt TEC some. Probably best to run the 226W TEC off the power supply and that's it. The other pelt (80W @ 8A) you can probably run off your Computers power supply if it is a big enough PS. That isn't a lot of power. You also should check your cards wattage usage as some modern cards are running 80watts and that small TEC may not be enough and could cause a negative effect. This is all my thinking, anyone more electrically enlightened can shoot me down at anytime. I would wait for a second opinion. EDIT: May be cheaper and better just to buy the blocks from Swiftech aswell. TEC blocks are more complicated than regular one's. |
|
03-18-2004, 05:36 PM | #4 |
Thermophile
Join Date: May 2001
Location: UK
Posts: 1,064
|
In parallel everything has the same voltage. So for speccing you add the currents. Series is more complicated as everything has the same current, but the voltages are split. For ohmic devices the split follows the law of potential dividers. Unfortunately, TECs aren't ohmic - they're semiconductors. I wouldn't reccomend running them in series.
If you want to run both off a single supply you need a unit rated for 33A continuous (nb: make sure the rating on a supply you buy is continuous, not peak). Also be warned, if you put a load of 100W on a 12V, 25A TEC you'll need to sink 400W off the back - 300W from the TEC and 100W from the load.
__________________
Once upon a time, in a land far far away... |
03-18-2004, 06:02 PM | #5 |
Cooling Neophyte
Join Date: Jan 2003
Location: Montreal, Canada
Posts: 60
|
I have yet to see any 226W pelt rated at 25A@12V, most of the time, they are rated (I really do believe Swifty's are too) for 25A@15.4V. so for 12V, your 226W pelt will actually draw about 19A. Not sure about the 80W, but I think they are also rated for 15Vmax supply, so they wont require 8A, but more like 6A. So you might be able to run both pelts off one single meanwell psu, but i really urge you to contact Swiftech and ask them beofre you buy it and fry it for putting a too heavy load on it.
__________________
OCS CS clan |
03-18-2004, 06:09 PM | #6 | |
Cooling Savant
Join Date: May 2003
Location: Okotoks, A.B. Canada
Posts: 726
|
Quote:
__________________
"Great spirits have always encountered violent opposition from mediocre minds" - (Einstein) |
|
03-19-2004, 03:25 AM | #7 |
Cooling Neophyte
Join Date: Jan 2004
Location: Phoenix
Posts: 6
|
f1lite,
Check out some of power supplies (Bulldog Series) from Rivergate Distributing http://www.rivergatedist.com/Power_supply.htm . The RDC-30 is about $147.00 clams. To figure out your power consumption for your TEC, Click-Here . Later, monkey_1 Last edited by monkey_1; 03-06-2005 at 11:41 AM. |
03-19-2004, 07:17 AM | #8 |
Thermophile
Join Date: May 2001
Location: UK
Posts: 1,064
|
Power consumption for TECs doesn't vary linearly with voltage (though it's quite close), and it varies with load... I'd say go for a 30A supply for both at 12V (though note a lot of high current supplies are 13.8V, not 12V).
__________________
Once upon a time, in a land far far away... |
03-19-2004, 07:19 AM | #9 |
Cooling Neophyte
Join Date: Mar 2004
Location: Southeastern USA
Posts: 5
|
Thanks everyone. Lots to think about; I've read a couple different results others have had with different pelts; seems most folks are using the 156w pelts for CPU? Is that because of the waste heat you have to deal with?
Also read that the 80w pelt of the GPU is no more effective that H2O; so what I'm going to do is get the Swifty waterblock, coldplate and gaskets and run only H2O on the GPU first, and if I'm not satisfied I'll buy a 130 pelt for it. The reason I went with the 226w pelt for CPU is because thats what the Swifty uses, and I got a good deal on a MCW462 so I was just going to buy the TEC conversion for it. Do you guys think thats too much pelt? What can I expect the difference to be between a 156w and 226w pelt? Looks like running in parallel is what I need to do; thanks prandtl, I think I read somewhere that a guy called Swifty about running both on that meanwell and they said no problem, so you are probably right. I also went and checked the specs on the pelt, and here's the specs: Couples U max I max DT max Q max L ×W×H(mm) R(ohms) 127 15.2 VDC 24 A >67 ° C 226.1 W 50 × 50 × 3.10 0.48 Then this little note: "Minimum requirements for a dedicated power supply: 25A at +12 volts. If adjustable voltage is available: setting the voltage higher than +12V is not recommended. Lower voltage, can be safely used, down to 9 volts." What about using a seperate ATX power supply? I've got a stack of old HP servers here I could snag the PS out of. I just have to make sure it will run 25A? Any sites that show you how to activate the 12v rail on a PSU without actually hooking it up? Anyway, I found the recommended Meanwell S320-12 for $94, so if I can't rig a PSU than I'll just buy it. Maybe I can run both a 156w and a 130w off it; but I'm still trying to figure out what I need to get the job done. I'd like to get down to 0c on my P42.8c@3.5GHz. Also, I found a Meanwell S350-12 that runs at 29A; would this be a better choice? EDIT: also, if I use a variable-wattage PS, does that mean I can run the pelt at different temps? Sorry for all the questions, I'm a noob with this stuff. Last edited by f1lite; 03-19-2004 at 07:59 AM. |
03-19-2004, 09:23 AM | #10 |
Thermophile
Join Date: May 2001
Location: UK
Posts: 1,064
|
Well using dT = (1 - Q/Qmax)*dTmax (which is a rough measure of dT for TECs) and assuming a 100W heat load gives:
dT = (1 - 100/156)*67 = 24.05C for the 156W dT = (1 - 100/226)*67 = 37.35C for the 226W So on the face of it the 226 looks much better. However you have to sink more power. Assuming 75% efficiency the load is: Qtot = 100 + 156/0.75 = 308W for the 156W Qtot = 100 + 226/0.75 = 401W for the 226W If we then assume a 0.08C/W system and 25C ambient the hot side temps are: Thot = 25 + 308*0.08 = 49.64C for the 156W Thot = 25 + 401*0.08 = 57.08C for the 226W This gives a cold side temp of: Tcold = 49.64 - 24.05 = 25.59C Tcold = 57.08 - 37.35 = 19.73C Now, unfortunately I have no idea what the actual C/W rating for your system is so that is fairly rough - if you cooling system is worse then the 156W is going to be better, if the cooling system is better than the 226W is better. also if the heat load is much greater than 100W then the 156W will be inadequate. For reference a P4 2.8 puts out 75W stock, overclocked to 3.5G without adjusting the voltage means approximately 94W of heat load. To get to 0C on your P4 with a 156W would require a perfect cooler (or very low ambient), so I don't think it's going to be adequate for your needs. With the 226W you'll need to keep the hot side at 37C or less, which will require a cooling system capable of 0.03C/W, which is no easy feat! ATX supplies can work - you'll be hard put to find one with enough 12V I suspect. To get one to run you can connect the grey and green wires (I think) and it'll power on. You'll probably need at least 1A of 5V load to get proper operation out of it.
__________________
Once upon a time, in a land far far away... |
03-19-2004, 09:56 AM | #11 |
Cooling Neophyte
Join Date: Mar 2004
Location: Southeastern USA
Posts: 5
|
Butcher, are the results you gave to reach a goal of 0c, or is this just to run the thing? If my C/W is higher (probably more realistic) will I still be able to get 226w significantly below ambiant? Then I can fine-tune my H2O system to try and get closer to .03C/W?
|
03-19-2004, 11:32 AM | #12 |
Thermophile
Join Date: May 2001
Location: UK
Posts: 1,064
|
Those are to run the thing - your final cpu temp would be roughly Tcold. You'll need at least a 0.08C/W system just to get below ambient.
Recalculating for a C/W of 0.15 you'd be at: Thot = 85.15C Tcold = 47.8C Comparatively, a straight water system of 0.15C/W would give: Tcpu = 25 + 100*0.15 = 40C TECs are really only worth it with a decent cooling system. Just to note, TECs are only rated to Thot = 85C in general, so if your system can't manage 0.15C/w don't bother. Something to note however is that C/W is not constant for a varying load - my water system has much better C/W with a high load than low. I assume it's because convection through the radiator works better with a higher heat delta.
__________________
Once upon a time, in a land far far away... |
03-19-2004, 11:54 AM | #13 | |
Cooling Neophyte
Join Date: Mar 2004
Location: Southeastern USA
Posts: 5
|
Quote:
EDIT:Of course, I don't know what your average MCW462/Eheim1250/ChevetteHC+120mmHSF water cooler runs as far as C/W goes. Last edited by f1lite; 03-19-2004 at 12:02 PM. |
|
03-19-2004, 11:57 AM | #14 | |
Cooling Savant
Join Date: May 2003
Location: Okotoks, A.B. Canada
Posts: 726
|
Quote:
without a load you should get frost no problem on a water system that isn't very good. add 85-100 watts of heat from the CPU to disipate and i think you will get a difrent picture. my 2¢
__________________
"Great spirits have always encountered violent opposition from mediocre minds" - (Einstein) |
|
03-19-2004, 12:57 PM | #15 |
Cooling Neophyte
Join Date: Jan 2003
Location: Montreal, Canada
Posts: 60
|
Butcher
Your Qtot are way overestimated, using Since87's pelt calculation excel sheet, i got a Qtot (using 100W from cpu) of 315W, instead of 401W. I also have the feeling that the estimated Q of the cpu is also overestimate, but I have no idea by how much. Tho even with this lower Qtot, seeing temp around 0°C under load probably wont be possible unless you use chilled water. As far as I know, "ambient temperature" in the file is a reference to water temperature (or air in the case of air-cooled pelt). So the best you can do is play with the file and see how a pelt will work under different conditions. You will soon notice, that the better the wb, the higher the voltage can be put. On the other hand, you will get better results at low voltage in the case of high C/W wb, so a variable psu really can help you find the best voltage for your setup. p.s. when I said Swifty I was refering to Swiftech p.p.s. I dont know how accurate Since87's calculator results are, but since he probably based it on this article I think we can trust it.
__________________
OCS CS clan |
03-19-2004, 12:57 PM | #16 | |
Cooling Savant
Join Date: Mar 2002
Location: FL
Posts: 787
|
Quote:
__________________
When you do things right, people won't be sure youv'e done anything at all. Looking to buy/trade for used Deep Fryer and Vacume Pack Sealer. |
|
03-19-2004, 01:03 PM | #17 |
Cooling Neophyte
Join Date: Jan 2003
Location: Montreal, Canada
Posts: 60
|
In a way they are, but you also have to make them so that applying the 200-300 lbs pressure on the TEC wont make the wb base bend and hence make bad contact with the TEC.
Personnally, I believe that a #rotor style block with a relatively thick base will do a good job at cooling a TEC. Also, using more than 4 screws to tight the wb/TEC/coldplate assembly together might help having a more uniform pressure/contact.
__________________
OCS CS clan |
03-19-2004, 01:43 PM | #18 |
Cooling Neophyte
Join Date: Mar 2004
Location: Southeastern USA
Posts: 5
|
I already bought the block; swifty MCW462-U. Got a good deal, $30; I can get the TEC upgrade from Swifty for $56, and then I just need the rest of the H2O setup and a PSU for the TEC.
If I had a great H2O setup, and decided to add a 130w pelt to the GPU, would a 350W/30A PSU like the Meanwell S350-12 support them both in a parallel configuration? |
03-19-2004, 01:50 PM | #19 | |
Cooling Savant
Join Date: Mar 2002
Location: FL
Posts: 787
|
Quote:
right, so whats complicated about making a #rotor style block. If you wanted to, you could even go so far as to put a screw into every one of the pins, that would make for a sturdy enough block, right?
__________________
When you do things right, people won't be sure youv'e done anything at all. Looking to buy/trade for used Deep Fryer and Vacume Pack Sealer. |
|
03-19-2004, 02:28 PM | #20 | |
Put up or Shut Up
Join Date: Dec 2001
Location: Spokane WA
Posts: 6,506
|
Quote:
If your the DIY type and have usable tools and the knowledge to use them you don't care about money and you will do it yourself. If you want a good quality end results, don't have tools, and don't know the first thing about block design then buy premade and save yourself a shit load of time and money. I am in the middle of designing a TEC block and it is not nearly as easy as a regular water block. Lot's more stuff to consider. I am trying to design it to hold a 50mm TEC but the AMD mounting holes are not in a good spot. After the O-ring groove is milled it doesn't leave much room for a proper bolt pattern to clamp the cold plate/tec/block together. What I got so far in attached pic. It is a Rotor variant. The holes will be connected through the middle. Just have to figure out the hole pattern for the coldplate. Being I know nothing about the math of figuring out proper thicknesses to use I will have to make 4 different sized cold plates and try a few different base thicknesses before I have something I would be happy with. If I am going to go through all the BS involved in TEC cooling then I am going to do it right and to the best of my abilities or it isn't worth it. After years of designing and building regular blocks this is much moreof a challenge to do right If your goingto do it yourself at least to it better than if you buy a pre-made. It will not be cheaper if done right. |
|
03-19-2004, 03:24 PM | #21 | |
Cooling Neophyte
Join Date: Jan 2003
Location: Montreal, Canada
Posts: 60
|
Quote:
As for cold plate thickness, BillA has made some recommendations for the thickness. Also, I thought the recommended pressure was 300ish pound, not psi, wich is the same, I think, as about 75psi (with 50x50 TEC)?? The americain's unit system is killing me! Superart making the block is not complicated, what is complicated, as Jaydee pointed out, is making a proper or good design.
__________________
OCS CS clan |
|
03-19-2004, 04:00 PM | #22 | |
Put up or Shut Up
Join Date: Dec 2001
Location: Spokane WA
Posts: 6,506
|
Quote:
|
|
03-19-2004, 04:42 PM | #23 | |
Thermophile
Join Date: May 2001
Location: UK
Posts: 1,064
|
Quote:
Q of the cpu is an estimate, however the stated power output for a P4c at 2.8GHz is 70W (sorry, misread the data sheet slightly and got 75, still not far off even so). (source: intel processor data sheet). I adjusted using the forumula Q = Qspec * (f/fspec). So Q = 75 * (3500/2800) = 93.75W - should be 70 * (3500/2800) = 87.5. Note however that this assumes stock voltage - increasing voltage causes a very large power rise. Also note that the rated power for a P4c 3.4GHz chip from intel is 89W so my estimate might be a little low. I'm not sure what Since87's sheet is based upon, but without knowing I'm not likely to trust his results over ohm's law and intel specifications.
__________________
Once upon a time, in a land far far away... |
|
03-19-2004, 04:45 PM | #24 | |
Thermophile
Join Date: May 2001
Location: UK
Posts: 1,064
|
Quote:
__________________
Once upon a time, in a land far far away... |
|
03-19-2004, 05:32 PM | #25 | ||||
Cooling Neophyte
Join Date: Jan 2003
Location: Montreal, Canada
Posts: 60
|
Quote:
At the "Major features" enumeration, they write: Quote:
Quote:
Quote:
__________________
OCS CS clan |
||||
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
Thread Tools | |
|
|