Search around all these are simple TEC Qs that you can find, try different forums though there isnt anything along the lines of a tutorial sticky around these parts.

My design used an undervolted TEC so you will need one much bigger than a 170w, your system will also have to be adapted to the heat load you are planning as well(437w at 7-9v seems about right to me). Keep looking and LMK what you come up with, I am waiting for someone to get back to me on some P4 results with this setup.

If you're going to run without a pump (so using water heat differential to "drive" the water around the loop) IMHO you'd want the radiator "inlet" higher than the "outlet".
Try putting the rad on its side, so one pipe is higher than the other and see if there's any change. If there is, maybe think about something long-ish and single pass, so you can have the inlet 'way higher than the outlet. Like maybe the fedco 2-766 (14.188x3.25x1.625) single pass. Or if you don't want to figure out how to move air through something that odd a shape (I'd just cut a vertical slot in my case to fit and block all other inlets so whatever exhaust fannage you've got will pull air through the rad, the case, then out), then maybe a more square fedco 2-149 (7.75x6.125x2) - also single pass and lots easier to fit - although I'd guess you'd want the inlet higher than your heat source, so maybe the long, narrow one would actually be easier...
Oh - I have a little experience with using "no pump" cooling systems - tried cooling 2-stroke cylinder heads on a race bike long ago - worked fine so long as the "rad" (was a car heater core) inlet and outlet were at top and bottom - well enough that we could run leaner (2-strokes are also somewhat gas-evaporation-cooled) which meant fewer pit stops. Airflow through the rad was, as you might imagine, not an issue...

BTW, anyone who missed Cathar's Experiments in TEC based water chilling - go read that. Great stuff - got me interested in TECs again - thanks Cathar!

Re: LS7Corvette's passive peltier setup
 

Just a lil FYI from messing around with the new PSU I got. The TEC @5v seems to be pulling about 7-8amps instead of the 10a that I predicted earlier.

more testing to follow...

Re: LS7Corvette's passive peltier setup
 

for all those people who recommend gallium and liquid metals, it will never work.

Natural convection velocity is a function of 1/viscosity^2.

A big heatsink with the fins hanging vertically would work too but it is no where near as fun

Re: LS7Corvette's passive peltier setup
 

Good advice, I saw this first hand messing with flourinert. I will have to try messing with fluids again after I get done testing the TECs.

Any ideas for a natural convection fluid?

Re: LS7Corvette's passive peltier setup
 

water oddly enough is near as good as it gets.

I would suggest that for your loop the correlation with fluids goes something like

h is directly proportional to (beta/ viscosity^2)^n * k

maybe

h is directly proportional to (beta/ (viscosity*alpha))^n * k

where symbols are for fluids being their normal things
0.25<n<0.333

EDIT - had so many questions on natural convection of late that i may have to write an artical with a program for it. Had to do a beasty paper (Reached 8500, still could of added more) on it so i've read up.

So if you have a fluid plug the numbers and that is how it will perform relative to water. Should be easy to do for something like FC77 (flourinert). There are other realtions that i could post if your interested.

You could get hard number if you wish by pluging the numbers for your setup with a known liquid (near pure water) and then seeing what changing the numbers does.

Re: LS7Corvette's passive peltier setup
 

Would be interesting to see what comes out of the equations, I have wondered if other fluids might work better in this setup.

Re: LS7Corvette's passive peltier setup
 

doubt there is anything. something lower viscosity maybe an it s abit hit or miss that this is how your system is working.

I'm fairly sure water (maye with some special additives is your best bet unless there is some superuper fluid on the web.