Is it possible to build a water cooling setup with no pump? A closed system with only temperature differences to get some type of flow going. Not for very hot components of course, say a 45W TDP CPU, NB and perhaps a GPU.

I doubt that. Water is too viscous for that to work.

not withing a closed system, using an open system you could use evaporation to pull the water through the system, but within a closed system the water simply wouldn't flow, as the water would not know which way around the system to flow.

Best bet is to get a quite pump, but with such a low TDP cpu, is there any need to water cool ? you might be better off with an air cooling solution such as a fanless Ninja

A few people have done it.
I forget the details, but you'll find examples if you search.
It's essential to have the heat-generating components lower than the radiator.

http://en.wikipedia.org/wiki/Ground-cou ... _exchanger

http://en.wikipedia.org/wiki/Geothermal_heat_pump

I remembered reading about a 'thermosiphon' at overclockers, did a search, and came up with something different than I remember. ( Looks like that site has changed a great deal since I read there regularly. )

http://www.overclockers.com/cpu-cooling ... for-water/

The idea I remebered was to use the cpu to boil ether and use that boiling as a pump. ( I started watercooling for the silence and wonder: how 'quiet' is a boiling ether pump? )

Sterling engines can also be made that operate on a temperature differential well below boiling a medium, be it either water or ether.
http://en.wikipedia.org/wiki/Fluidyne_engine

There's no absolute reason against this. When water heats up, it will decrease in density, and this can be used to move it around. It would be rather difficult to estimate how well it would work, but it's probably an interesting thing to attempt.

Radiator would need to be horizontal, and physically above the water block. I'm reasonably sure you want one short, vertical tube connecting the rad to block together, and a longer loop on the other side (preferably with a 90 degree barb just above the return to the cpu block).

The reasoning is that in a symmetrical loop, there's no motivation for the water to move through the loop so it wont. Limited to conduction through 10cm of water will lead to terrible temperatures. An asymmetric loop stands a better chance of establishing a flow, though I'm not convinced the forces involved would be enough to significantly overcome friction. I'd be more inclined to set it up and test than to calculate things in this case.

Have you given it a try yet, or just thinking out loud?

As far as I know, pump-less solutions has been used to cool old car engines. The old Saab 92 (1949-56) had a passive watercooling system for its 2-stroke engine. It doesn't seem not physically impossible, but I guess cooling a computer is quite different from a car engine.

Also found this on Wiki:
http://en.wikipedia.org/wiki/Thermosiphon

Hmm the thermosyphon idea is very cool! I was thinking about something roughly similar, but I was thinking big heat pipes.

EDIT
I found this:
http://www.benchtest.com/heat_pipe1.html
http://www.benchtest.com/heat_pipe2.html
http://www.benchtest.com/heat_pipe3.html

Yes it can be done, theoretically. As per above, the driving force would be the difference in density, caused by the difference in temperature. It's not needed to have an assymetrical loop though, more important is to make sure the friction loss in the tubes is kept at a minimum, i.e. wide tubes (1" maybe?).

This has been used in buildings for central heating, so it works. Problem with a PC would be relation between power (W) and resultant water flow - you might not get enough heat dissipation.

I guess you could argue against this idea much the same as against the "going fanless" idea: even the slightest mechanical help (pump/fan) will increase the effectiveness tremendously, so if you can bear (or drown out) the pump/fan noise, there's no real need for trying this (but it would be fun, of course).

/d

Another gem. Thanks.

Bultaco road racer with no water pump:

http://books.google.com/books?id=7bGuud ... &q&f=false

The difference between cooling a CPU and an internal combustion engine is there is a much lower temp differential with a CPU. The flow rate won't be high, but it should flow through tubes.

It should be easy if you have a change to setup water hoses high above your system. Even better if you have a cool place up there which water must pass during it's journey. I'm using same idea with air cooling and I have been able to remove all fans from main chamber of P182.
http://www.silentpcreview.com/forums/vi ... hp?t=57320
http://www.silentpcreview.com/forums/vi ... hp?t=57619

no, is not possible. Always need to pump.

Yes it is possible, this guy did it http://www.vonslatt.com/proj-cc.shtml

As other people have mentioned you need the radiator at the top and use nice big tubing to reduce impedance.