No fans no pump at the moment, but im adding a rad tomorrow and will test closed loop just like in case. Idle CPU 5W (so asus tells me) so im pretty confident.
Printable View
No fans no pump at the moment, but im adding a rad tomorrow and will test closed loop just like in case. Idle CPU 5W (so asus tells me) so im pretty confident.
Central heating header tank...big plastic reservoir 18x12x18highQuote:
Originally Posted by Eldonko
I've thought about this several times... I knew it would work just not how well.
Once the water in the block gets up to a high temperature differential, it is just dying to move and make way for cooler water, thereby jump-starting the convection flow. Once it is flowing and has inertia then it should take little force to continue movement through the loop, so the reduced heating still provides enough convective force to keep things moving.
I bet that loop design/block design could be optimized to make this work quite well. I think the location of the loop components in relation to the heat source should have an at least semi-signifigant effect.
Tony i tried with pump and no rad, and its horrible if your loop is not large like you have it.
Also when you do add more heat wattage, it fails fast. :T
The key is the large reservoir. But with a large reservoir you could even run it overclocked.
I will try with a rad only first...i have actually run something similar a while back with a TEC pumping around 50W and that worked quite well...the TEC jump atarted convection and load temps were reasonable...the main issues was the block i feel...the HF looks to be a lot better for convection than the old DD block i was using.
I think you need to consider the fact the average PC is not on load full time, hence normal use with load fluctuating should not be a problem with the closed loop not being overwhelmed. Of couse folders need better cooling though so pure convection would probably not be a good idea.
option 1 (Top) has no pump and convection only, im thinking option2 is more for the overclocker or someone who runs the PC under load more than normal. Pump is CPU temperature controlled, only pumping when needed.
mad paint skill I know but you should get my point ;) T line is incase the res is sealed, or you just run that part of the loop to top connection on rad and do away with the res.
Coooooool =]
WOW!! this is VERY interesting here......... I wonder what would happen with a smaller res (like an EK 250 or even a Micro Res...) but with a triple rad....
This is something I might look into......
The same as the chances of thermal laws stopping unexpectedly...Quote:
Originally Posted by Eldonko
This is really neat for stock/low-OC setups. I thought it would only be effective if the loop was set up to have hotter water rise, though.
I'm not sure if this is helpful at all regarding radiators, but I did run one experiment. I took my TFC480 and set in on my radiator testing bench without any fans on in the room and let it rip with 100watts applied.
About 3 hours later, temps were still climbing and this is the plot of my sensors. The reservoir I use is about 1/2 gallon, so I think the key is the large reservoir for the additional heat storage, although I never would have thought you could get convection to move the water properly.
http://img377.imageshack.us/img377/5...ivetestob6.png
So with the 480 rad, I was probably going to level off with about a 20C delta using 100 watts as a heat load. So I suspect a really small heat load could be doable, that or any amount of air that's pushed through the case for something like the PSU would likely improve performance over true passive a fair amount.
I did however have an unintentional pumpless test run while testing the Apogee Drive, I accidentally disconnected my GPU loop pump (reaching back for the backplate) and really disfigured my tubing. The tubing at the GPU outlet was all deformed (Flattened). I logged and recorded that little adventure too, suprisingly though it did seem to almost level out in water temps so I must have been convecting some sort of flow rate.
giggleQuote:
Originally Posted by AndrewZorn
Interesting idea for an average computer - no reason it won't work.
I do have a small question about option 2, however.
Is it just me or will the pump loose about half it's flow going around in circles?
Don't you need something to stop the flow from running backward through the other half of the "Y" ...?
It would need a backflow valve, but that would goof with the convection... you would be mostly relying on the water's inertia to carry it through the correct path.Quote:
Originally Posted by QuietIce
There is no need for a valve...water will flow in the path of least resistance...for it to flow around the sharp bend on the Y's the block would need to be massively restrictive.
Quote:
Originally Posted by NaeKuh
this is a very interesting thread although it doesnt apply in my case
if i turn my pump off the cpu heats up very fast and fails within seconds..seeing that convection would need some time, it seems rather impossible with such wattages:confused:
But your pump is stopping/blocking the flow of water altogether. Also, this isn't for well OC'd systems.Quote:
Originally Posted by Giannis86
Axis
I just can't seem to understand how the hot water would know which path to take away from the CPU block to initially kick start the convection process.
A nice easy experiment would be to add a couple of drops of dye into the CPU block and fill the rest of the loop with clear water then fire it up - if convection starts you would see the dye start to circulate.
Might give that a go actually.
This is certainly an interesting experiment.
well hot water rises so you would want the cpu inlet comming pretty horizontal and the outlet going pretty much dead straight upwards.. ;)Quote:
Originally Posted by coolmiester
all based on physics..
True, and the blocks we are using are more of a convection compromise thats for sure.Quote:
Originally Posted by dopestuff
if you have the board mounted in a case with the rad on the top of the case inverted ATX is looking to be key for performance...longer tubes add capacity so you want the CPU as low down as you can me thinks ;)
Also im now testing with just a rad and no case, I had to tip the board around 35Deg from horizontal to get the water moving...i was watching the temps in bios and just tipping the board dropped 5C+ :)
more soon...this is looking very promising...its actually matching my ubber loop which as 12gallons of coolant..LOL
This time I have about 0.5L of coolant which is a lot less ;)
here is how I have the system now:
Closed Loop, No reservoir, 3x Xilence fans 12V
Idle room temp 20C@ radiator height
http://www.ocztechnologyforum.com/st...closedloop.PNG
Just run off the first load test run then let it idle for 15 mins to give a good representation of how it would be in a system and its idling at 23>28C dpenedant on what core you look at ;)
Load was a little higher than with the ubber loop once it settled down...i will run it again a post some further info.
lol martin i didnt even need 100W. :PQuote:
Originally Posted by Martinm210
I had about 25W the heat load off a DDC-3.2
But Tony your ideas are awesome, love to see followup on your theories. :D
nice tony
now i am wondering what are the temps when fully stressed???
room ambient 20C load was 55 thru 59C, Linpack 4th loopQuote:
Originally Posted by dopestuff
Doing an idle test now with just 1 fan and its not moved from 23C >27C depending on which core you look at. Room ambient is 21C now also
It's always cool to see people experimenting in water cooling. I believe what Tony has built is called a Thermosyphon. There was a good article on overclockers.com about this several years ago but I can't find it. The guy built a very tiny loop with nothing but a heater core and a water block and it worked as well as a regular heatsink. Although I don't like their spelling you can read more about it on wikipedia: http://en.wikipedia.org/wiki/Thermosiphon the references section has some good links!
Well 1 fan is coping so well at idle I don't even feel you need it...so idle test will continue with no fans for at least 90mins.
In this age of ever increasing fuel bills this looks to be the way of having your cake and eating it along with having the bread buttered both sides ;), im thinking the loop designed like my option 2 with a small pump controlled by the bios fan header control maybe the way to go. I just wish there was an inexpensive fan/pump controller (sub 20$) that you could control via usb like a Tbalancer and you could add the fans/pump.
More soon.