Hey everyone, firstly i'd like to introduce myself. I'm new to these forums but not overclocking. There's no such thing as too fast. Started waaay back when, went from air to water to a vapo finally. Anyway, due to some interesting circumstances my computer is destroyed and i've got my insurance company paying for a new one. The figure they ended up settling is they will reinburse me somewhere in the neighborhood of $4500. So, i've got a good wad of cash to blow and the MyGueyver in my is itching to do something fun with it.
Anyway, the idea I thought about was doing some submersion cooling on an SMP + SLI system in a few gallons of flourinert Fc-72, but what would be the point of that if I couldn't cool the hell out of it for some overclocking. I've been reading about different things people have done with mineral oil and flourinert ect... Dry ice or Ln2 is too transient for me, so I was thinking waterchill. However, I thought that instead of chilling water and pumping it in why don't I just chill the flourinert directly? The pour point of fc-72 flourinert is -90C so I don't see a problem with it getting frozen, so i'm wondering what experience or ideas people might have on the subject. What kinda temps could I achieve putting the evap directly into the flourinert? I'm guessing i'd need some way to circulate the flourinert so that no hot spots develop.
Anyway, once I get a better idea of precisely what i'll be doing i'm going to build this sucker, with updates in this thread. Any help from some people experienced with waterchill or anything else helpful please lend me a hand.
Last edited by ArkhAngelos; 09-30-2004 at 12:07 AM.
sounds like he had a pretty sweet system....so the price of that system is prolly well over half that money...I can tell you honestly that you wont see the best performance going that route..you could always build a really powerful waterchiller which would make cooling everything doable..or just go phase...choice is really yours but performance will be as follows...phase(best)...chilledwater(good)...subme rsion(extremely hard to equal temps of a chiller)... GL and definatly make a log on which way you go
Can't use alchohol, i think most alcohols are around .05ish u/cm conductivity, not to mention the possibility of fire/explosion in submersion. Flourinert really is the ideal liquid for this, aside from it's expense but thats not an issue for me at the moment.
I did consider a good ol' vapor change or a water chiller, but a couple things kept my mind going back to submersion.
Firstly, a dual cascade system would mostly likely make my electrical wiring break down and cry, I think whenever i turn on a light somewhere else in the house one goes off.
Secondly, i'm going to be building a SMP system and getting some SLI Geforces once that double nf4 mobo comes out (Soon I hope!) Vapor change is probably just too much of a hassle or even an impossibility with 2 procs and 2 gpu's to cool.
Water chill is still a possibility but this just seems like too much damned fun. I also live in a super humid enviroment so condensation can be very troublesome. In addition, I do some sound recording and being able to have a near 0db system would be a very nice plus to me (I could temporarily turn off the compressor when needing to record)
Anyway, it may not be the absolute coldest or fastest system conceivable but it would suit my purposes perfectly as well as be really goddamn cool. In theory it's pretty simple but what i'd really love is some ideas on how to get the flourinert as cold as possible without ultra extreme measures. I'm not too familiar with waterchill, so I dunno what kinda methods are used to get the lowerst possible temps. The coolant you guys chill and then pump is basically the same idea except it the coolant would stay in place. Can a higher BTU translate into lower temps? Or does it only effect the capacity to cool? What is the lowest temps people have their waterchills running at?
Thanks for the help.
sounds like a nice idear, but i wouldnt go as low as -70C as the bios chip and some other parts dont like low temps... i dont know if -70C is ok, i just know that -100C is too low for bios chips and other components to work properly.
if you really want to go for this def get a nice acryllic waterproof "case"!
will def look sweet
i would get some good waterblocks and mount a good watercooling loop on the board and videocard, a nice heatsink on the nb and sb, and then submerge the entire thing in flouri-nerd
i would use a normal watercooling system with blocks and pump the coolant or maybe water through it as just a heatsink on the cpu and gpu will def cool the parts "enough" but pumping a chilled coolant through a waterblock attached to the cpu and gpu is def more efficient and gets you lower temps on those parts. the only advantage of the flouri-nerd wuld be that all system components are chilled as well and that it looks very sweet of course
Can't use alchohol, i think most alcohols are around .05ish u/cm conductivity, not to mention the possibility of fire/explosion in submersion. Flourinert really is the ideal liquid for this, aside from it's expense but thats not an issue for me at the moment.
I'm not sure what you mean by u/cm, but assumeing its thermal conductivity, Flourinert is really quite poor. Several times worse then methanol. It has worse heat capacity too.
Quote:
Firstly, a dual cascade system would mostly likely make my electrical wiring break down and cry, I think whenever i turn on a light somewhere else in the house one goes off.
Just keep in mind a DD system is going to be more effcient then useing a secondary cooling fluid for a given amount of cooling.
Quote:
Water chill is still a possibility but this just seems like too much damned fun. I also live in a super humid enviroment so condensation can be very troublesome. In addition, I do some sound recording and being able to have a near 0db system would be a very nice plus to me (I could temporarily turn off the compressor when needing to record)
Useing a different coolant isn't going to affect condensation or noise, so I don't really follow you here.
Quote:
Anyway, it may not be the absolute coldest or fastest system conceivable but it would suit my purposes perfectly as well as be really goddamn cool. In theory it's pretty simple but what i'd really love is some ideas on how to get the flourinert as cold as possible without ultra extreme measures. I'm not too familiar with waterchill, so I dunno what kinda methods are used to get the lowerst possible temps. The coolant you guys chill and then pump is basically the same idea except it the coolant would stay in place. Can a higher BTU translate into lower temps? Or does it only effect the capacity to cool? What is the lowest temps people have their waterchills running at?
Unless I'm missing something, your system would work the exact same as a waterchilled system, so the info in there forums should be pretty useful.
moosturdsoed25, yes, was time for something new going to take some new pics of my cats soon
and yeah saratoga, the way i got it its a chilled water cooling system submerged in flouri-n3rd to cool the rest of the system (sb, mosfets, memory, voltage regulators, etc, etc)
I suppose the title was a bit unclear, what I was interested in building was submerging a system (in fluorinert) and then cooling the fluorinert directly (computer and evap in the same container, no pumping). After giving it some thought, I decided that if i'm going to go as gung-ho as possible I might as well build or purchase a dual cascade system as suggested and then submerge. The obvious benefits of this are no worries about condensation and some additional cooling benefit for the mosfets, ram ect... Plus it would look damned cool
Oh, and for those interested in the thermal properties of fluorinert I decided to stop being lazy and just look up the datasheet from 3m. Flourinert comes in several different formulations, each with different properties.
The main difference between the different types is the pour point and boiling point. For computer submersion the ideal type is FC-72, with a boiling point of 56C and a pour point of -90C. The thermal conductivity of this fluorinert measured at 25c is comparable to water (well, very slightly below) The specific heat is approx. 1/4 that of water.
You can buy this stuff by the barrel (5 gallons) for about $750US; which, while significant, might be worth it to some to get rid of condensation worries and make things purdy.
Now I gotta start reading up on cascades, thanks everyone for the help.
After giving it some thought, I decided that if i'm going to go as gung-ho as possible I might as well build or purchase a dual cascade system as suggested and then submerge. The obvious benefits of this are no worries about condensation and some additional cooling benefit for the mosfets, ram ect... Plus it would look damned cool
thats what we were telling you all the time
just submerging it into flouri-nerd and not pumping it and just cooling it would have gotten you really bad results!
750$? wtf?
thats really realyl expensive, and if i were you i would look for some clear motor oil that doesnt conduct electricity (i know there are many and they dont cost a lot) and make sure that they dont attack the pcb and electronics.
tbh i had the idear once to just attach heatsinks to all parts of the mobo that get hot and then use clear electronics paint and cover the entire board with it. then attach the watercooling parts and everything and seal all the pci and memory slots and all the other small parts with seal string or other stuff, and then submerge the whole board in water+antifreeze....
you could do that but instead of antifreeze submerge it in oil with would not short the board even if it was not sealed perfectly.
yeah but i think theres some machine oil that can do the same... the only difference is that it doesnt conduct heat as well as flouri-nerd, but thats not that important imo. we are talking about memory and southbridge and the mosfets. with heat/ramsinks the heat conductivity of non conductive oil should be big enough to cool it all very well, just like it was fouri-nerd...
I suppose the title was a bit unclear, what I was interested in building was submerging a system (in fluorinert) and then cooling the fluorinert directly (computer and evap in the same container, no pumping). After giving it some thought, I decided that if i'm going to go as gung-ho as possible I might as well build or purchase a dual cascade system as suggested and then submerge. The obvious benefits of this are no worries about condensation and some additional cooling benefit for the mosfets, ram ect... Plus it would look damned cool
I don't think that would work too well with a CPU. Turn off the pump in a watercooled system and at best you'll crash the system. At worst, you'll boil off the water. Theres picks of blocks with melted tops because a pump died.
In this case you're going to use a fluid with 1/4 the heat capacity of water, and much less conductivity so the delta T will be even greater. You can help this out by picking a large heatsink to help out convection, but thats only going to do so much. I think it will work, but I think you'll be very dissipointed with your CPU temps. Passive convection works, but only at very high delta Ts. You could compensate by introduceing a pump to direct capacity where its needed, but that raises the obvious question of why even bother submergeing it if you're going to run pipes and tube anyway . . .
Quote:
The thermal conductivity of this fluorinert measured at 25c is comparable to water (well, very slightly below)
rofl, should work with strong fans like a delta ROFL but i doubt they will move a lot of water. they will spin really slow and die because of too much pressure holding them back i think
FC-77 costs 1875.00 for a 5 gallon pal if you can get it for 750.00 I'd like to know where??
__________________
i7 920 @4.305 ghz (205*21 HT on) Prime stable@ 1.312v (Batch# 3844A717) your results may vary
i7 920 @4.410 ghz (210*21 HT on) 1hr Prime stable@ 1.39v (Batch# 3844A717) could push more but no reason too
Enzotech Heat Killer SCW-REV.A (very nice block 73c / 79c per above voltages)
GA-EX58-UD4P (Bios ????),, 6GB- OCZ3P1600LV6GK (7-7-7-20 @1640 mhz) 1:1
Powercolor 4870x2 with EK nickel block
Water Cooling - TFC X-Changer 480 and 6-Scythe SFF21E fans @1000 rpm , Vario 655
Raid 0 velociraptors 600GB,, ZM1000-HP (powerfull and silent)
Mitsubishi 22" CRT (old school ,,no input lag , no slow response times , true color reproduction & blacker blacks) its deep but so what
750 for 5 gallons sounds about right..the stuff is completely nonconductive..plus it can take some low temps....
There was a test done, cant find it now, but when in complete submersion, a PSU did die. It is not really conductive, but natural impurities make it so. It is good for occasional splashes, not submersion.
Submersion cooling is messy and inefficient and for the same price as those liquids you can just build 2 single stage phase-change units. cooling the entire motherboard and components will not raise the performance higher than having your gpu and cpu at -30 die temp.
Get chilly1 to build you a dual evap dual cascade, and cool CPU and GPU submerged in that lesser expensive crap to prvent condensation. Problem solved and 1 kick ass system. Go for the benches!
Flourinert + submersion + chiller?
going to take some new pics of my cats soon 
wtf?
Linear Mode
