I was able to build a really cheap and quiet HTPC/720p system, and I wanted to share details on the experience. The main thing that made it all possible was cutting the middle out of a 120mm fan, then mounting the ring/frame at the back of the case to duct the HSF exhaust out of the case back.

This duct meant that the CPU heatsink 120mm fan was only about 20 mm away from a 120mm duct connected to the back of the case. A large amount of the CPU exhaust exited the case and a small amount created turbulence inside the case. Case is an Antec knock-off from 2008; a full-tower of thick steel with a 120mm case fan behind CPU, and no other back or side vents. No case fan is used.

Corsair vx450w power supply (21 db says our Mike C)
XIGMATEK Gaia SD1283 with included "16-24 db" fan
Intel E5200 dual-core wolfdale 2mb (Overclocked 2.5-> 3.33ghz)
WD green 750 GB
iStarUSA T-5-SA Trayless Anti-vibration SATA Rack
Crucial 4GB DDR2 800
Gigabyte GA-G41M-ES2L
HIS Silence H557HR1G HD5570 1GB DDR3 (passive graphics card)

$330 +s/h (as of Nov 2010)
(+$45 for case, back in 2008)

Only two fans (CPU heat sink & power supply) that are both 120mm and temp-controlled are used. System is quiet without the use of any expensive parts. It streams HD and plays games at 720p just fine, exhaust air is never more then warm.

I say do the pictures.

Pictures or it didn't happen.

Ok, I have pics, but need to crop & post them. I have taken a few measurements and done some calculations that I will post before I forget.

Using a kill-a-watt, the system uses about:
65 watts at idle
90 playing videos
130 watts in 3D games

The fan is about 18 mm from the duct. If we use an idealized system, we should be able to figure out how much air that passes through the CPU cooler would be forced out of the case. The whole area that air is forced into can be thought of as a cylinder (AKA 'can of soup' shape). See wikipedia picture.

The bottom of the cylinder is the fan pushing air in. The top of the cylinder is the mesh-covered 120mm hole in the back of the case. The 'sides' of the cylinder are the open area between the fan and the duct, where air can escape back into the case. By bringing the duct and the CPU fan closer, we reduce the open 'sides' of the cylinder ('h' in the picture), and we should be able to decrease the amount of air that escapes into the case. This means more air would be ejected out of the back of the case.



If air escapes in any direction it can from the HSF, the current system should eject about 53% of the air that is pulls through the CPU cooler. If the empty 120x20mm fan frame (AKA duct) was replaced with an empty 120x35mm fan frame the ejection fraction would increase to 87%. The concern is that objects close to a spinning fan can increase noise. I will only make the switch if summer temps require.

Ejection math:
through-duct escape area / total escape area
(120mm circle * mesh resistance) / ((120mm circle * mesh resistance) + (cylinder-sides area))

Please clarify... maybe explain again using different words?

Makes sense. Have you tried it with the fan just mounted to the back panel, so it acts as a duct to the heatsink by itself? Or do you have a grill on the back panel still and this would therefore cause turbulence with the fan up against the grill?

Could always use tape (electrical tape or, ideally, self-amalg) or something to seal the duct onto the heatsink, thus removing the gap? Just thoughts for improvement, sounds like it's working pretty well already .

Have you looked at the Thermalright CPU cooler ducts:
http://www.thermalright.com/new_a_page/ ... _duct.html

They might be useful for similar situations.

Drive mount pics. Note that two drives in lower cages are not hooked up.