Hello, and welcome to this experiment to get as cool temps in my rig to get the highest overclocks as possible First of all, I would like to give a big thanks to Antec for making this TEC chiller experiment possible And Monsoon for some fittings: As said, the point of this TEC experiment is to cool down the water temperature in my rig to get some nice overclocks. The rig is not final yet, but it's coming along really nice and you can se more here; http://forums.bit-tech.net/showthread.php?t=256575&page=2 But here a sneak peak; But back to the chiller; As the title say this is a TEC chiller. In short a TEC/Peltier can transport heat from one side to the other, and by that make one of the side cold, but the other warm. To know more look here; http://en.wikipedia.org/wiki/Thermoelectric_effect The TEC does this just by using electricity which is why a TEC is a genius invention. However TEC's are not that efficient just yet, which is why they are not that common use in things like a fridge (but they will perhaps become in a few years). But their inefficiency isn't that big a problem for this experiment since we just bump up the numbers of TECs and power for them should be no problem (except if the electric system in my apartment can't handle that amount of Watt's). So the idea (for a TEC chiller in general) is to block the TECs in between two water blocks. The "cold-side" water block will be incorporated in the rigs loop (no radiator in this loop when the chiller is in work) so that the water will get cooled down and the hardware should get some low temperatures and then high overclocks The "hot-side" of the "TEC sandwich" should of course be cooled (or it will most likely melt). But the lower the temperature of the "hot-side" the lower the temperature of the "cold-side" we can get. And since the experiment is to get as low temperature as possible, the cooling for the hot-side is also in the extreme category (more on this later ) TEC's Right now it looks to be these TECs; http://customthermoelectric.com/tecs/pdf/12711-5M31-24CZ_spec_sht.pdf How many is still a bit uncertain, but it will be at least 12 of them (perhaps 16). But if you have any other ideas for some better TECs, please speak up. They will be run with 12V, so have this in mind Water blocks The water blocks for the "TEC sandwich's" will be custom made in copper and plexi top. The internal design for the water block is not still decided, but so far I'm looking toward something like this (Design from Skyrip at Overclock.net); But perhaps with smaller and more of the squares to create some turbulence and high surface area that should make the blocks efficient to transfer heat to the water ;-) But other designs are more then welcome as well So there you have my idea for an extreme TEC chiller. I don't know if I will make subzero temps when bench with just CPU or one GPU or the whole system, but the hardware will be isolated to cope with the possible condensation ;-)
haha, i was thinking: nice somebody doing tek stuf again and the first thing i noticed was that those blocks really look a lot like mine(not read the tekst yet) my advice would be to go for more teks with lower wattage, on ebay you are able to get 65W qmax teks for around 2,30 dollar a piece. this would also yield the advantage that you will get a higher surface area.(always buy a couple extra because some tend to break during shipping) personally i would not reccomend the use of 2 copper plates since that method is quite expencive. especially when you would want more pins. this would need a smaller cutter so it would take much longer to make and the risk of the cutter breaking is also much bigger. personally what i find the most interesting tec technology is the direct touch type(no interface between tek plate and water) i made a prototype of it: http://forum.highflow.nl/f34/super-cool-14427/ if i made a new one i would use thicker plexiglass(10mm or higher) and other silicone gel(by accident i used acylic based and that didnt cure)
Last weak a heavy package came (guess from who ): And within that a few more packages; Can you guess what is inside those packages? If you thought fans, then you were right And lots of them (56 of them to be precise) Some nice fans with Double ball-bearing and able to move a good amount of air with 2000 rpm. Properly a bit noisy, but that really doesn't matter for me, they just need to move a lot of air. But if silence is needed they can always be slowed down to 1200 rpm. But for this project I'm also in need of quiet a lot of thermal compound, and it have to operate at both high and low temperature along having good thermal conductivity. Which Antec also have supplied Some nice Formula 7 with diamond particles (not sure if they really do the trick to getting lower temps, but it seems like a good thermal compound): But there is plenty of it, which is needed for all the TEC's and the SG09 build's hardware Antec did also through in a bit extra for my three U2711 screens (could use some cleaning after some 2 years of use ) So that's it for now. Next step would be to get the water cooling for the "hot-side" loop home (guess what all those fans are for ), and to decided which and how many TEC's I should get and order them (so the water block also can begin to take shape/design them).
Will be keeping an eye on this as I have been wondering about TEC's for a while but never really looked in to it properly. Will be interesting to see your results.
So I have come a bit further in with this project The following wasn't meant for this chiller, but for my SG09 build that was planned long ago. That's also why they have the colors that they got (not Photoshop, but spray painted). And yes, that is a Phobya External Radiator "Bench edition" stand, along with a D5 pump in a EKWB X-RES 140mm housing that have been spray painted white. The stand only comes in black, and I ran out of paint on a Saturday (this summer), and haven't come much further since. BUT it will get painted just as white as all the rest. Then pump will also get a slap of paint. Not sure if I should go green on it or white like the rest. On the other side we got the 9x120mm radiator that is 45mm thick, along the 120mm fans from Antec I think that the plexi fans are looking really well with the rest, and I will look much better when the radiator stand will get all white. Another package came this Friday including some of the best looking fittings right now (I do think that acrylic tubing and fittings looks nice, but for me Monsoon fittings and their colors are just the right thing ) So thanks to Monsoon for sponsoring some fittings for this chiller project If we take a closer look at the pump/reservoir, you can see that it's not the normal water that is running round in the little loop. I have always like the way that some colorful crystals was floating around in a liquid up in chemistry lap. But most of those crystals was either expensive or very toxic. So when I saw that Mayhems had made a similar product for watercooling, I new I had to have it. The liquid is Mayhems Green Aurora. (pump settings was "1" = the lowest) But the Aurora is not ideally for this chiller experiment, so when that time comes, I will have found something more suitable. The Aurora was for the looks in the SG09 build Doesn't those beautiful Monsoon fittings deserve a picture in focus?, I do believe so Now before we end this update, soe of you may think; "that painting will get you higher temps". And yes, that is correct, BUT again this was intended for the SG09 build, which really doesn't require a radiator like that, but I really like silence (or my girl does), so I got the radiator. As for the chiller, one of these is clearly not enough for the amount of heat I will put into that loop. But that's also why there is coming two more of these in (just 60mm thick) and all three of them will have push pull (that's way the 56 fans ) I do think that should do the trick Just as a closing picture (don't stare to long at it), the structure of the radiator stand and radiator in background:
Hi, well the blocks are yours Problem with more low wattage TECs is also a much higher surface area that needs to be cooled. Now I like your idea about direct watercooling of the TEC's surface with a plexiglass. But there is a problem with it, first of with to those cheap TECs often isn't near the same hight (which is important when they are between the same block). Secondly, have you tried the direct cooling? I don't think it will work very well, do to the little resistant and the very little contact area for the water. I simply doesn't believe that the water will take up that much heat by this method. (I may be wrong that would be cool, but doubt it). The thing with the copper, is that there you can make a much larger cooling surface for the water and more resistance. And the thermal uptake to the copper is also efficient. Back to your blocks, can you tell me the dimensions of the squares?
hey you can find the build of the direct cooling chiller here: http://forum.highflow.nl/f34/super-cool-14427/ personally i would not reccomend building a direct cooling system unless you know exactly what you are doing. the biggest issue is that your teks cannot have a fracture in the ceramic or else that tec will create a shortcut and burn down. also it gets significantly more difficult when you start using more teks. cooling performance whas the same or even better than the copper cooling system. for the copper block i used a 3mm spacing between the pins and the pinns were 1.5x1.5mm. problem with copper is ofcourse that it is expensive and not the easiest material to mill especially with small cutters (below 3mm) personally i think the direct cooling and the copper cooling are about the same considering temp and difficulty to make. direct cooling is more advanced though(not noob proof ) in my experience with those cheaper teks is that they have a tolerance of about 0.1mm, and they probably come from the same factory as the ones you can find in a local shop. and i found that there is almost no difference between the local shops teks and the chinese ones.(all made in china)
Hi all Now with Christmas over and exams, I'm once again back to hopefully finish this build and the TEC cooler. Also Happy New Year This update isn't a big one, but I am in need of your help I'm not at big water flow engineer, so if you got any ideas for at better water block for the TECs, please let me know Currently I got this designed; This will be in copper and it is 6mm high (with the squares 4mm(height)x2mmx2mm) and will get a acrylic top (where the inlet and oult will be and at each end). Holes in the sides are for screws and springs to press the two blocks around the TEC tight together There will then be one at each side of the TECs. The question is if this block is good enough for the amount of heat they need to take up and give to the water (4 or 5 TECs + some of the hardware load = approx. 1000w)? Also Skyrip have suggested that I could use direct cooling of the TEC's instead (where the water runs directly over the TECs). It would make it a lot cheaper and easier to make. But I'm in doubt as to how well it would work, due to there is limited restriction (just a plain surface). So Skyrip; Did you ever get to test it (temperature uptake from the cold and hot side and so forth?). So if you got any suggesting for the water blocks please let me know
direct cooling would be cheaper but it is much harder to make. copper plate with plexiglass would be much easier to get watertight. i would personally go for a copper block with straight slots and using a back forward situation: (c is where the inlet and outlet are) C===== ] C===== ] this will give you a smaller path so less chance for a deadspot of the water. i would reccomend making the slots as small as possible like ek blocks etc. to make slots smaller than 1mm a sawmill must be used because an endmill will die very fast. with slots of approx 0.5mm and around 30 to 60 on each side that would yield sufficient thermal exchange power. your design could work but is not optimum certainly when used with 1 in and output port in the middel. that would create a large deadspot on either side of the input.
Yea, have been getting quite the respond on the blocks as to choose straight lines (which I will do). I do like the idea of the direct cooling, but I think I will stick to the classic cobber blocks. Will ask my "block maker" what dimensions the lines can be made in Thanks for the replays skyrip
So that last of the watercooling gear have arrived I would like to thanks aquatuning.de for there sponsorship of these parts along the parts for the SG09 build But back to the good stuff; As I said, I do no believe that radiator cooling will be my bottleneck, and here's why: One big, fat 9x120mm radiator (60mm thick) And when you place that in a Phobya radiator stand along 18 Antec fans it looks like this: That is a heavy setup, and there is one more of those along the white/green radiator (45mm) posted above And they cool really well. Right now the green/white radiator + D5 pump is cooling a mining rig with a HD 7990 + 2x HD 7970 down to the cores are at only 44 degrees and water around 37 degrees (and that is a 24/7 setup) And that is with only four 180mm fans at 700 rpm. So the three of these radiators with push pull configuration with the 2000 rpm should be more than enough for this chiller I also got some extra stuff, like an extra D5 pump to help in the cold loop and four temperature sensors (green LCD of course) and a lot of screws for the fans: Some of the Monsoon fittings also arrived, and quite a lot of them; And when taken out and assembled it looks like this; Beautiful But this is not all of them, actually these are just around half of what I need for this chiller and SG09 build (around 30 pcs. both fittings and angle adaptors). Some of the are in the mining rig, others are not yet bought due to uncertainties of how many TEC waterblocks that will be made (how many fittings is needed). And once again, thanks to Monsoon for sponsoring some of their fine fittings, and I must say you are right; "Bling is good" As for the TEC waterblocks I will go with the long thin lines, but have had time yet to design/draw them. But should happen soon and hopefully I will find some cheap (but good) copper at that time. (if you got a link to a seller in EU (that ships to Denmark) please post it ) I don't have that much more, except that I got a new pump for the SG09 (other one burst into flames) and I swapped out the GTX 670 4gb cards for a new; A nice Gigabyte GTX 780 TI thanks to my friend at Coolgamers.dk It came as rev 1.0, so thought It was the reference PCB design, but luckily it was the custom more powerfull design I tried the card, and I must say it performs well and is really silent. Even in game after many hours of play, I cant hear it even with the case open (bot only at stock clocks for now and 1.5 meters away from case). So that will be the card that will get chilled (maybe another one will come along as well if my wallet is up for it). And EKWB is making a full cover waterblock for it (yea), bot it will first get released in late February :-/ But when under the chiller it will be with a core block only (either custom one or a EKWB block), and with heatsinks on Vram and VRM. So that's it for this time, any questions is welcome, and link to a good but cheap copper retailer is appreciated
So, long time no see. I'm sorry for this interruption in this build log/experiment but my studies and personal situation needed some attention. And sadly I haven't been able to locate anyone with in Denmark that could and would help me making the waterblocks (had one for a few months until he backed out...) But the good news is that I have found a solution that I think/hope will work out just as well. Did cost a little extra, but what the heck So small package (though quite heavy) from aquatuning came this week: Which contains the rest of the stuff for cooling these TEC's. Some 6 meters of tubing, a reservoir and some fansplitters for the 54 fans As for the solution to cool the TEC's I went ahead and bought 20 Alphacool CPU waterblocks Some nice blocks, though I have seen better surfaces on other blocks out there. But I believe these will do just well when sandwiched around a TEC Since there is 20 waterblocks that means I would also need some fittings: Some 50 pcs of 90 degrees fittings and 10 straight ones. Also with all these CPU blocks and the three big radiators I needed some more pump power so I bought two more D5 pumps (four in total now). The plan is to get some of the 550W TEC's that is 62*62mm in size (found one from UK that would sell 10 of those to me): http://www.ebay.com/itm/TEC-Thermoe...D&orig_cvip=true&rt=nc&_trksid=p2047675.l2557 They are a little bigger than the cooling surface area of the CPU coolers, so I will also get some 6mm thick 70*70mm copper plates, so the whole area of the TEC is in contact with the cooling surface. I am open for suggestions if you believe there is a better TEC out there for this setup (will be running at 12V) That was all for this time, the TEC's and copper plates will be bought in the upcoming week And then a Classified 980 or similar when they arrive
Okay, so some good news (and a bit of bad news). The PSU's arrived a little time ago from Antec (thanks again ). Three lovely Platinum PSU, two 1300W and a 1000W: Modular and OC link so two PSU's can run together And enough power for the TEC's and stable power for the GFX when OC. So the TEC's also arrived. Bought from a German seller over eBay (we have some high tax here in DK, so it was affordable to get these compared to china or US). However they only had 6 pieces left of the 15Amp's TEC's (15 Volt) However the had four 4 of the 26amps TEC's Bit more expensive but lets see how well they perform. They are a little bit bigger than the 15 amps: And here on the CPU waterblock: The 26 amps' corner is not covered by the block due to the screws hole in the copper plate. But I hope and think it will work out anyway Now the bad news is that I will first have time to play with all these things until the weekend. I did also buy four more of the 15 amps from china. So I will be doing some testing how different setups will cool (TEC and radiator setup). However they will first arrive in minimum 14 days. Perhaps just four of the 26 amps is enough for a GTX 980
This... is quite silly. I am very much looking forward to it. Question: What benefit is the TEC stage adding, if you still loads of massive raidators with multi fans? Is it just because you can? Will it allow a greater OC than water+rads alone as it can theoretically get below ambient?
So putting all of the TEC's together takes a little more time than expected. First apply a thin layer of Antec's Formula 7 Thermal paste to get the best transfer of heat And then make a sandwich of CPU coolers around it: And then connecting them all: My fingers was quite sore after some 60 fittings But all done and installed (takes up most of the table): I decided to start with just the two thick radiator since the TEC wattage have been lowered a bit, and as I expected the pumps had a hard time pumping the water around with two radiators and 10 CPU block. To night I started out with installing the chilled water on a older GTS 250 card: A bit overkill I think, but it just a test card With a little overclock I believe the card is pulling out 130-150W. A little less than the stock GTX 980. However the TEC setup had no problem in cooling the card. With heaven bench running the temperature of the cold loop began going towards 0 degrees C. It took a few minutes since it had to cool the ~1.5 liter of water that is running around in the cold loop. The core temperature of the GTS 250 followed the temperature of the water nicely with around +5 degrees. I have no anti freeze in the system yet so the plan was to stop a 0.5 degrees. But right at 0.6 the core temp shot up to ~35. Some water properly froze in one of the CPU coolers and blocked the water from running around. The above shows how the core temp fell from ~36 to 6 degrees in 15 seconds when the water began flowing around again So I should properly get some anti freeze in the systems. Any ideas what to get? Also, how much does I need to insulate the GFX? Perhaps someone got a good guide? Since it does get some condensation: The hot loop was at 32 degrees when I stopped, and the room temperature was 22. So with a load of 250-300W I should be able to get the cold loop down to ~-10 degrees
It is With just radiators and fans you can never go below the temperature that the air going through the radiator has (no matter how many radiators or fans). With TEC's I can go below that like showed above
