Permabanned
Sub'd - Looking forward to this after the air-cooling exploits. Love to see pics on how you assemble the pelts and radiator setup.
eXtreme Liquid Cooling - Build Log + Overclocking
- Thread starter Pneumonic
- Start date
Update 7:
Test Set-Up Power Supply Prep
So whilst I wait for the TEC Units and Heat Exchangers to make their way through the post-Christmas mail I am starting to prepare my test power-supply unit..
Now for the time being I am going to run the TECs between 0v - 12v just to see what sort of cooling I get... the nice thing about the units I have is they are rated at 15.5 Amps @ 24v, so running them at 12v each unit is drawing a mere 7.75 Amps which multiplied by eight gives a total draw of 62 Amps @ 744 Watts which is within the range of most 'high-end' PC PSU.
For the test bed I am going to be using a Corsair 850 Watt PSU which is rated to draw 70 Amps / 840 Watts on the 12v Rail. The unit is redundant and has been previously 'modified' so I am not to worried if it doesn't make it
The nice thing about TEC's is the lower the power you run the unit the more 'efficient' the unit is... so even at 12v (50% power) the unit will give around 65-70% of its total 'cooling' capacity. Ideally you want to aim to run a TEC at around 35% power. In this instance I would get around 55% of the TECs total 'cooling' capacity at 35% power.
So running eight units at 12v still gives a total 'cooling' capacity of 1260 Watts... even running all eight at 3v will give around 400 watts of cooling which should be enough to hold the cold side liquid temperatures at load.. (well lets hope so anyway!)
With the ever increasing TDP of modern CPU this is one of the main reasons why it is better to use multiple TEC units to chill a medium rather than direct-die cool.
So the first stage of this job is to take the PSU to pieces and start the process of separating the various wires and then I will have to get the soldering iron out
The picture shows the wires separated out into 3v & 5v (which I am not interested in!), 12v and the ground wires. Although on 'one' rail the 12v is actually connected on four different 'contact points' within the PSU. The end goal is to have two contact points feeding a PWM unit which will then control four TEC units. The only other wire I am interested in is the green wire which is normally found on the 24pin motherboard plug.. you can see it grouped with a ground (black wire).. this wire is what tells the PSU its connected to the motherboard and if to turn on or not.. it will therefore be connected in the end to a switch on the front of the PC so I can switch the unit on and off.. The two PWM units will also have there control wire to one control knob so I can control all eight TEC units in one place.
I am currently toying with the idea of mounting the PSU PCB in a new housing... it is rather noisy with its standard fan control at 80% load + so I may just ignore that and set it up in a custom housing with adequate quiet cooling or I may place some spare mosfet water blocks I have on the heatsinks and water-cool the bugger not sure yet.. Although this build is 'eXtreme' I also intend it to be quiet!
Test Set-Up Power Supply Prep
So whilst I wait for the TEC Units and Heat Exchangers to make their way through the post-Christmas mail I am starting to prepare my test power-supply unit..
Now for the time being I am going to run the TECs between 0v - 12v just to see what sort of cooling I get... the nice thing about the units I have is they are rated at 15.5 Amps @ 24v, so running them at 12v each unit is drawing a mere 7.75 Amps which multiplied by eight gives a total draw of 62 Amps @ 744 Watts which is within the range of most 'high-end' PC PSU.
For the test bed I am going to be using a Corsair 850 Watt PSU which is rated to draw 70 Amps / 840 Watts on the 12v Rail. The unit is redundant and has been previously 'modified' so I am not to worried if it doesn't make it
The nice thing about TEC's is the lower the power you run the unit the more 'efficient' the unit is... so even at 12v (50% power) the unit will give around 65-70% of its total 'cooling' capacity. Ideally you want to aim to run a TEC at around 35% power. In this instance I would get around 55% of the TECs total 'cooling' capacity at 35% power.
So running eight units at 12v still gives a total 'cooling' capacity of 1260 Watts... even running all eight at 3v will give around 400 watts of cooling which should be enough to hold the cold side liquid temperatures at load.. (well lets hope so anyway!)
With the ever increasing TDP of modern CPU this is one of the main reasons why it is better to use multiple TEC units to chill a medium rather than direct-die cool.
So the first stage of this job is to take the PSU to pieces and start the process of separating the various wires and then I will have to get the soldering iron out
The picture shows the wires separated out into 3v & 5v (which I am not interested in!), 12v and the ground wires. Although on 'one' rail the 12v is actually connected on four different 'contact points' within the PSU. The end goal is to have two contact points feeding a PWM unit which will then control four TEC units. The only other wire I am interested in is the green wire which is normally found on the 24pin motherboard plug.. you can see it grouped with a ground (black wire).. this wire is what tells the PSU its connected to the motherboard and if to turn on or not.. it will therefore be connected in the end to a switch on the front of the PC so I can switch the unit on and off.. The two PWM units will also have there control wire to one control knob so I can control all eight TEC units in one place.
I am currently toying with the idea of mounting the PSU PCB in a new housing... it is rather noisy with its standard fan control at 80% load + so I may just ignore that and set it up in a custom housing with adequate quiet cooling or I may place some spare mosfet water blocks I have on the heatsinks and water-cool the bugger not sure yet.. Although this build is 'eXtreme' I also intend it to be quiet!
Last edited:
Soldato
If you do water cool the psu - have a separate block for each HS - I used a copper spreader plate over an earthwatts 380.
Due to slight height differances I pulled a few components off the pcb when tighten up - Was a pita to fix.
Can be seen in this old project mock up below, used an old Maze cpu block. Then I got the Yesico
WARNING - if you don't know what you are doing DO NOT mess around with PSU electrics. (/dad rant)
Due to slight height differances I pulled a few components off the pcb when tighten up - Was a pita to fix.
Can be seen in this old project mock up below, used an old Maze cpu block. Then I got the Yesico
WARNING - if you don't know what you are doing DO NOT mess around with PSU electrics. (/dad rant)
Soldato
I wish you luck, but every TEC loop I've seen that's not directly cooling a component has never worked partcularly well. With a case that big, would a chiller not have been a better option?
Associate
"15.5 Amps @ 24v, so running them at 12v each unit is drawing a mere 7.75 Amps "
Not sure about the above, 15.5 amps at 24 v, would be 31A at 12V, or am I missing something ?
Not sure about the above, 15.5 amps at 24 v, would be 31A at 12V, or am I missing something ?
Don
Halve the voltage, you'll halve the current for a fixed resistance. Not quite that simple with TECs though as their non linear devices.
I've added you to msn according to your profile page on here as I think I want to do something similar to this. I've been planning it for a while but now feel motivated to get some things cut. If that fails, I'm <username>@hotmail.co.uk
Based on the handbook here it's very important to avoid shear stresses in the ceramic when mounting. Since the mounting pressure is high and the plates inflexible, the tecs have to be exactly the same thickness. Melcor offer them lapped in pairs if desired to achieve this. I think you're using 8 pelts and two cold plates, so how are you planning on arranging these? I really like the idea of fitting a large number in the 5.25" form factor but am struggling to work out how to achieve this.
Buying cold plates is starting to look like a far, far superior idea to making them. I've registered with their site, the CP15 you mention appears to have dimensions of 95 by 150 or 300mm. This explains 40mm tecs instead of the larger ones, but I don't see where you're fitting eight of them. Two broad in the 95mm direction, fitting the rest lengthwise gives either six or twelve (forteen if you're optimistic) total?
The most common 40mm tec I've found is the 12710. A datasheet suggests with a hot side of 25 degrees it'll move 50W at a 25 degrees delta for a cost of 96W. Alternatively it'll move 20W at a 20 degree delta for a cost of 25W. If using six of these you'd be looking at 300W for 600W used, or 120W for 150W, so I can't see a way for this to hold true?
I hope your maths is right and mine isn't.
To clarify, this is roughly what I think you're considering. Have I got the wrong end of the stick? 93 x 20 x 188mm
Based on the handbook here it's very important to avoid shear stresses in the ceramic when mounting. Since the mounting pressure is high and the plates inflexible, the tecs have to be exactly the same thickness. Melcor offer them lapped in pairs if desired to achieve this. I think you're using 8 pelts and two cold plates, so how are you planning on arranging these? I really like the idea of fitting a large number in the 5.25" form factor but am struggling to work out how to achieve this.
Buying cold plates is starting to look like a far, far superior idea to making them. I've registered with their site, the CP15 you mention appears to have dimensions of 95 by 150 or 300mm. This explains 40mm tecs instead of the larger ones, but I don't see where you're fitting eight of them. Two broad in the 95mm direction, fitting the rest lengthwise gives either six or twelve (forteen if you're optimistic) total?
The most common 40mm tec I've found is the 12710. A datasheet suggests with a hot side of 25 degrees it'll move 50W at a 25 degrees delta for a cost of 96W. Alternatively it'll move 20W at a 20 degree delta for a cost of 25W. If using six of these you'd be looking at 300W for 600W used, or 120W for 150W, so I can't see a way for this to hold true?
I hope your maths is right and mine isn't.
To clarify, this is roughly what I think you're considering. Have I got the wrong end of the stick? 93 x 20 x 188mm
Last edited:
Soldato
This looks awesome, subbed. Can I please ask one fav? For each update, could you please put a link in the first thread?
Just helps when your trying to read any epic multi page log (which i assume this will become)
EDIT: See that you have done this - many thanks.
Just helps when your trying to read any epic multi page log (which i assume this will become)
EDIT: See that you have done this - many thanks.
Last edited:
Associate
Jon: I built something similar, but smaller using 2 437W (but probably 360 in reality) QMAX Tecs running in parallel at 14 volts ( total current draw 25 amps) so the CoP was >1. The difference in thickness between those 2 (standard) TECs was < 0.01 mm so I don't think it will be a problem. Best to have the clamping screws at the midpoints of the edges of the TECs rather than the corners to avoid cracking the ceramic. I have quite a lot more info if you are interested, although you've obviously done a lot of homework!
HNY
Soldato
Surge - Do you have a build log or any information about TEC Chillers? I have only read about using direct contact TEC's.
Associate
Here's my build log (I have a different username at XS)
http://www.xtremesystems.org/forums/showthread.php?t=230755
and here's someone else's (lots of TEC stuff on this site)- 40 pages
http://www.overclock.net/peltiers-tec/395259-quad-tec-water-chiller-project-ultrasonic2s.html
Soldato
Thanks mate. Sorry for being OT.
Wow, what an interesting read. Thought this post might be of interest. It appears that for every TEC added you dramtically increase the heat on the hot side!
OMG! I just read Ultrasonics build log!!! Part way through the thread he actually confirms your design somewhat in that to get the best temps perhaps 8 TECs would be required.
Wow, what an interesting read. Thought this post might be of interest. It appears that for every TEC added you dramtically increase the heat on the hot side!
OMG! I just read Ultrasonics build log!!! Part way through the thread he actually confirms your design somewhat in that to get the best temps perhaps 8 TECs would be required.
Last edited:
Jon, I was looking at using the 30cm (ish) length unit and starting with eight adding more if required..
The QMax on the the TEC you have linked is 80 watt (ish) at 25c where as I plan to use units that have a QMax of 225 watt at 25c which is where I get the additional cooling power.. I am using the data sheets from the web-site where I ordered the TECs to calculate out the values..
I will provide a load more info tomorrow just got home and am very tired.. but will update in the morning
225 watt at 25c, with dimensions of 40mm? This sounds almost too good to be true, I'll look forward to your update. Cheers
Best found so far (feels like days of searching) 12711-5M31-15CQ at $26 a unit before shipping
At 4V, cop = 2, moves 20W for delta20
At 6V, cop = 0.66, moves 20W at delta30
At 6V, cop = 1.3, moves 43W at delta20
It's also a 40mm one, this appears to be the optimum for 6V. Any larger and cop falls too low as power drawn is excessive, any smaller and currents fall too far to move significant wattage. This has a qmax of 141, much lower than your 225, so I'm very interested in your find.
Further research suggests the 19915 is worth considering. I'm now sufficiently tired that I can't read the graphs though.
Assuming 2mm thick aluminium sheet folded to form the casing, it's possible to fit two of the things shown above ^ into the space occupied by a single optical drive. That's 12 40mm tecs in rather little space even using four cold plates rather than the required 3. Not much space left over for circuitry, but enough for plumbing I think.
Unfortunately the recommended source for these charges $186 a throw, with an email to ask for european pricing. I'm not sure my student loan will stretch to this, I may have to make the damned things after all.
Best found so far (feels like days of searching) 12711-5M31-15CQ at $26 a unit before shipping
At 4V, cop = 2, moves 20W for delta20
At 6V, cop = 0.66, moves 20W at delta30
At 6V, cop = 1.3, moves 43W at delta20
It's also a 40mm one, this appears to be the optimum for 6V. Any larger and cop falls too low as power drawn is excessive, any smaller and currents fall too far to move significant wattage. This has a qmax of 141, much lower than your 225, so I'm very interested in your find.
Further research suggests the 19915 is worth considering. I'm now sufficiently tired that I can't read the graphs though.
Assuming 2mm thick aluminium sheet folded to form the casing, it's possible to fit two of the things shown above ^ into the space occupied by a single optical drive. That's 12 40mm tecs in rather little space even using four cold plates rather than the required 3. Not much space left over for circuitry, but enough for plumbing I think.
Unfortunately the recommended source for these charges $186 a throw, with an email to ask for european pricing. I'm not sure my student loan will stretch to this, I may have to make the damned things after all.
Last edited:
Associate
At first i thought the bath WAS the extreme watercooling, will be keeping an eye on this thread.
Should make for an interesting build.
that would be fun
Associate
This is a good place to buy TECs (according to the folks at XS and Overclock.net- seemingly they arrive in 3 or 4 days). The 50mm x 50mm one at the bottom looks good, but appears not to be sealed- but that is easily rectified with silicone sealeant
http://www.hebeiltd.com.cn/?p=z.peltier.pricelist
When I was building mine, I e-mailed Custom Thermoelectric about prices and shipping to the UK a couple of times, but never received a reply......
Eventually bought the QMAX 437s from elsewhere (since I didn't know about the Chinese place at the time).
JonJ,
I have sourced mine from a company called AMS Technologies Ltd, who are charging £64 / unit.. this includes the drilling of M4 holes in the units to allow mounting. ..
The TEC units I have got from custom thermoelectric module number: 19911-5L31-15CQ
Surge,
Thanks for all the info I will go take a look although I think I must have read every post on TEC's at XS now..
I have sourced mine from a company called AMS Technologies Ltd, who are charging £64 / unit.. this includes the drilling of M4 holes in the units to allow mounting. ..
The TEC units I have got from custom thermoelectric module number: 19911-5L31-15CQ
Surge,
Thanks for all the info I will go take a look although I think I must have read every post on TEC's at XS now..
Last edited:
Ah, so you are going with the 19915s Jimbo recommends. What voltage are you planning on running them at for the larger exchanger?
From their graphs I get 6.5A at 12V for 78W used, moving 50W at d30 or 85W at d20, so the cop at 12V is rather good really. Still can't see where your numbers are coming from though. What temperatures are you hoping for? You mention running them at 4V, for the above tec the best you'll see is 2A used, moving 25W at a 10 degree delta so I'm not sure severe undervolting will work out all that well, perhaps run a few of them at 6V and turn the others off?
I shall contact ams tech. Trying to source 5mm copper to see if it works out appreciably cheaper. Cheers for the thread, it's great to see someone doing this
Need to check some things, but I suspect using half as many of the ones you're looking at at 12V works out better than my original idea of loads at 6V. Performance is definitely improved at higher deltas using a small number of 12V. However insulation becomes more of a problem. I've asked for a quote for a small number of the 150mm long plates, thanks for the recommendation. It's up against £50 per square foot of 3mm copper at present, which will do four plates comfortably, with about £30 of delrin needed. Would be rather bigger than your solution (about 45mm thick sandwich), but might be significantly cheaper.
From their graphs I get 6.5A at 12V for 78W used, moving 50W at d30 or 85W at d20, so the cop at 12V is rather good really. Still can't see where your numbers are coming from though. What temperatures are you hoping for? You mention running them at 4V, for the above tec the best you'll see is 2A used, moving 25W at a 10 degree delta so I'm not sure severe undervolting will work out all that well, perhaps run a few of them at 6V and turn the others off?
I shall contact ams tech. Trying to source 5mm copper to see if it works out appreciably cheaper. Cheers for the thread, it's great to see someone doing this
Need to check some things, but I suspect using half as many of the ones you're looking at at 12V works out better than my original idea of loads at 6V. Performance is definitely improved at higher deltas using a small number of 12V. However insulation becomes more of a problem. I've asked for a quote for a small number of the 150mm long plates, thanks for the recommendation. It's up against £50 per square foot of 3mm copper at present, which will do four plates comfortably, with about £30 of delrin needed. Would be rather bigger than your solution (about 45mm thick sandwich), but might be significantly cheaper.
Last edited: