Water Cooling 2!
EXTREME
OVERCLOCKING
Experment 2 code name "wc2"
System:
Supercooled Celeron 400 @ 744 mhz
About 8 months ago , was the last time I updated this site .. Its a very good example of how good intentions go wrong. I had just been able to get my p2 300 to 600 mhz with the help of some water/TEC cooling. It has been a while since that, and through constant improvements I am back with my much more capable cooler and a new cpu to boot ! . Hopefully for beginners I can shed some light on some issues .. for those guys without machine shops at their disposal this is the site for u .. homemade coolers rock !
In search of higher mhz (as always) I decided that there wasn't much point sticking with the p2 300 . At 600 mhz it was already forcing pc100 ram to run 133 mhz .. e.g. that's 33% over spec for pc 100. Also the p2's and p3's are kind of clunky to super cool as they have cache chips which are external to the cpu core. So you end up needing to cool wider amounts of area. After some thought I saw that some celeron 366's ppga were overclocking to 550 with standard cooling. They have integrated L2 cache, and therefore are quite easily cooled with the minimum amount of cooling surface.
Celeron 400 ready to run
As you can see any super cooled cpu needs some serious insulation. This is too stop any condensation from causing a short of the cpu or slocket pcb. The chip is also been lapped to full copper finish for that added flat finish. The cpu is coated with polyurethane, around the corners some silicone still holds from a previous cooler experiment. You can see the 80x80 mm square incut . that's to allow the 80x80 mm copper plate to fit snugly onto the cpu while been nicely protected from the warm air outside.
The copper plate touches the cpu. It has 4 60watt peltiers, in a 2x2 fashion , 2 high, 2 wide, cooling the copper plate. Needless to say the copper plate becomes so cold it freezes in seconds. So what cools the 4x60watts + cpu heat ~ 60watts = 300 watts of heat ?? , after trying 2 different heat exchangers I found a drilled solid aluminum block , 25 mm thick to be the best by a country mile, some 10 C better than even a copper back plate connected to a aluminum water cooled heat sink .. see previous coolers for a shot of that exchanger.
So here is the cooling system so far:
CPU -> Copper cold plate (Flat 2.5mm lapped) -> 4 peltiers -> 25 mm Aluminum water block.
So potentially the cpu can be kept below 0 C at 744 mhz 2.4v with those 4 peltiers if the hot side can be maintained at less than 20 C. Those of you who saw my initial cooler which used the aqua chill cold water cooler will know that that water cooler copped well with one peltier's heat load and the p2 600's , holding the water at about 4 C.
Aqua Chiller unit in current cooler
From this picture you can see the water cooler is connected to a 50 litre polystyrene water reservoir, the reason for this is the pure water cooler cannot cope with 300watts, it cannot hold 3 litres of water below 30 C while coping with that amount of heat. After some thought about possible solutions, the following was the cheapest . Use more water and therefore have that heat spread more , therefore lower water temperature. Inside the 50 litre chilli box I cycle 10 ice packs every 6 hrs. This maintains the water temperature below 5 C. which in turn maintains the Aluminum block at 10 C ! .
Its temporary until stage 2 comes online, but that requires some serious $$$ so it won't happen till Christmas !! Water is pumped from the chillibox to the cooler and back again under pressure. If no ice is added to the chillibox, the pure water cooler slows down the rise of the water temperature , but the water still hits 15 C after a few hours and would rise further, just more slowly as there is more water than with just the pure water cooler alone. Here I am using the waters heat storage ability and ICE to maintain that low temperature, the same amount of heat energy in 3 litres would be more like 30 C .
Chillibox in action
Now that the watercooling issue has been sorted out I needed to develop a heat exchanger for the cpu that could support the 4x60 watt TEC units. To Start of with, here are the chemicals/solvents/sealants/insulation i used when making the heat exchanger.
Water Cooling essentials.
To the back we have 5 and 10 mm polystyrene, this is for insulation .. nothing beats polystyrene !! This is used for my outer insulation on my cpu heat exchanger. The pump is a NOVA 400 litre per hr filter fountain pump (I use it without the filter), I have two of these for the two pipe loops. Mineral turpentine, great for removing grease and silicon and generally most forms of scum. It also I just found out dissolves polystyrene .. so be careful with this stuff. Methylated spirits I use to clean up after the turpentine, meths is much less harsh and removes less grime but evaporates and leaves no residue unlike turpentine. Silicone comes in handy for gluing polystyrene and sealing peltiers. The Plastic stuff is a PCB sealant, and is used over my cpu and slot converter to seal out condensation effects from shorting the CPU. Polyurethane is also handy for sealing electronics, just the layer is not as thick as the plastic stuff but its cheaper and non conductive . Last on the right is some closed cel foam , this is excellent for insulating uneven surfaces as I can't use polystyrene for those surfaces. There's some Armaflex 10 mm thermoplastic pipe insulation there too, this stuff is plain awesome.. saved me 5 C and dripping pipes . There's some 16 mm hydraulic hose there as well, this stuff is much more flexible than PVC piping and does not crystallize after long lengths of time but is about 4x the price of PVC pipes.
Parts used in the CPU heat exchanger.
To the top right we have the water block , with brass tap fittings. See "waterblock" under articles for full dimensions of this aluminium block. To the right, we have the 80x80 mm , 2.5 mm thick copper plate. This was "Lapped" (eg progressively sanded , 250 -> 600 -> 1200 grit ) until my arms gave up on the idea. Unfortunately .. the piece of copper i had is slightly (about 0.1 mm) depressed in the middle as you can see . When i can be bothered i will redo that plate again .. just ran out of time as it was all done in the space of 4 hrs .Hell thats what grease is for .. fill the slight inadequacies of the surface :)) To the left of that we have the back brace, which has a curved lip to put pressure downwards on the asus slocket. The slocket is bottom left, again i strived for a nice flat copper finish for the celeron 400 , as this transfers heat better than the standard tin surface. To the right of that we have the 4 60watt TECs , the bottom one needed a silicon repair on one of the corners. Each one draws 5 watts at 14 v and makes about 9 C temperature difference when the cpu is 100 % busy @ 744 mhz @ 2.4v . Finally bottom right we have the front brace, this time the curved lip holds the aluminium water block upwards. Effectively its a sandwitch design with a lower and upper lipped braces holding aluminium heatexchanger up and slocket down under tension .
Heat exchanger semi-assembled.
As you can see from above there are no drilled holes in either the waterblock or coldplate for tension bolts. After some consideration i decided these would only inhibit my cooling effect, eg carry heat through from the aluminium heatexchanger through to the copper plate. So what holds it on ? thermal grease and tension . When the unit is fully sandwitched by those two lipped plates there is a huge amount of pressure on the unit .. its not going to even budge :) . There are two gutter bolts (60 mm) in the front plate (bottom left) which hold the whole unit together . I considered 4, but then decided on two , right in the middle vertically of the cpu core. that way i can easily get even pressure right where i want it !
Heat Exchanger insulated and ready to be sandwitched together ! (two lipped brace plates not shown)
Here we have the 10 mm polystyrene insulation been glued in place on the aluminium waterblock, with transparent silicone. Bottom left we have the Asus slocket nicely fittied out with polystyrene as well. Bottom right we have some closed cell foam insulation which is going slap bang in the middle of the waterblock and the slocketed CPU. This is used here just to seal the unit on the edges better than two rigid pieces of polystryne could do. As its soft and conforms to uneven surfaces.
Stable power for PeltiersTECs
Solution 1. 4 x AT power-supply-units (PSU).
A few months ago , when my cooling consisted of one water-cooled 60watt peltier, I ran the peltier off my cases 230 watt ATX power supply (PSU). This worked good for the single peltier, my power supply at 12v supplies 8.5 amps of current. The peltier at 12v drew 6 amps of that current. I had one HD and CD-ROM at the time, so I'm not sure how it would have coped with more than 2 IDE devices, as they all draw current to some extent. My advice is if you want to use a single peltier with your cooling it should be ok from the main system PSU, but to be safe you may want to use an old AT power supply. I have 3 in the picture above + 1 ATX PSU. These AT PSUs came from throwaway 386/486 machines. I have found 1x12v 8-9amp power supply for each peltier produces the least amount of power ripple, I use my peltiers with no power filters currently so that is important, peltiers can lose efficiency when the power ripple is more than 10 %.
The nice advantage of one PSU per peltier is easy options for having 1-2-3-4 peltiers on or off. That way you can reduce the heat load in windows by turning peltiers off, and running RAIN or CPU IDLE to get back to approximately the same temperature. This only works if you aren't doing anything intensive, as RAIN or CPU IDLE will use the halt command to turn the cpu off on idle clock cycles, if there are no idle cycles your temperature will increase !
The above arrangement of 4 AT PSUs did the job well .. although it was rather messy with all those cables, but it was cheap .. and supercooling in my book is about doing it cheaply .. otherwise you may as well go and buy yourself that top end cpu. Anyhow ... for this project and others i decided to get myself a dedicated TEC PSU, which i picked up for a bargain price anyhow.
Note the water temperature (bottom left reading, top is Air temperature) and AT PSU switches here. There is 3 switches here for 3 peltiers, the other peltier is still powered from the systems ATX PSU. I have tried 2 PSUs for 4 peltiers, the temperature fluctuates by 4 C constantly, so I moved to 4 PSUs which supplied current much more steadily.
14V PSU wired up
A Closer look at the setup. Note the 1.9C on the dual thermister LED (white thing which says temp alert) is the water temperature. The 23.8C is the air temperature. Here you can clearly see the size of the IEC PSU in comparison to my case.
Power drain from the 4 TECs
The pic above shows the amperage draw of the 4 60watt TECs (thermo electric coolers). Each one on startup draws about 5.5 watts, as the water block temperature rises (hot side of TECs) the current drawn settles down to 20amps total load, about 5 watts per peltier. At 14.4v the TECs are shifting ~70 watts. They are rated 60watts @ 12v , 4.5 amps.
PSU Cooling
Note a very good photo of it , but two GlobalTech 240v AC cabinet fans (aluminum finish) are keeping the PSU's heatsink cool, with out them you could burn you hand on that heatsink !
Peltier Wiring
Cleaned up the cables quite a lot as you can see in this shot. I used bullet connections as they were tidier than wire screws (those blue caps I used before). The main trunk wire is rated for 17 amps @ 50v. Its lukewarm to touch after a while at 20 amps @ 14.4v
Full System Shot
My very packed computer desk, the other system is an Abit BP6 dual celeron 400 @ 570 mhz.
Result achieved with WC2 - 744 mhz from a celeron 400 !!!!!
System Boot Up
Sisoft Sandra @ 744
3D mark 99, yes folks ... 744 mhz stable in games also
WC2 was a labour of love for me last year ... since then I have tried many things with peltiers / water and freon. Stay tuned to hexus.com for the next installments !!
Author: Orthello
