Push/pull is a good idea if you can fit it. Shrouds are a good idea in order to avoid having that void spot you mentioned. Shrouds for me start with a 120mm x 25mm fan, and then I remove the motor hub, blades, and support, thus leaving me with just the frame. Since your radiator is on the thick side with high FPI, you'll want higher performance fans. You have a number of good options though.
San Ace 9G1212H1011 - 120mm x 38mm, specs are listed, commonly used on thick/high FPI radiators in demanding loops
San Ace 9G1212H4011 - 120mm x 25mm, smaller version of the fan listed above. CFM is slightly less, static pressure is slightly less (6.5278 mmH2O vs 7.1882 mmH2O), and noise level is 1 DBA higher. Not a huge difference, and it's a great choice if you can't fit 120mm x 38mm fans.
Scythe GentleTyphoon AP-30 - This 120mm x 25mm fan is a great option if you feel the need for even more cooling power. At 44 DBA it isn't THAT much louder, but CFM is higher as well as static pressure (9.652 mmH2O).
The reason I recommended those fans is due to their noise profiles, how well they undervolt (no clicks or whines will be present), and of course their performance. I have experience with San Ace, but not GentleTyphoons so those remarks are just going by reviews and other comments on them by highly experienced people.
n9zn-extra
I would suspect in a pull configuration the direction of exhausted fan air would yield no benefit unless obstructed. When a push/pull or push only configuration is used this is a little less clear. Having air flow directed straight ahead when entering the radiator would seem to provide little or no airflow around the edges of the round fan openings and cause the radiator surfaces in those areas outside the fan blade opening to be under cooled. Possibly this is only a concern in push configuration assuming a push / pull configuration would have the pulling fans drawing air from the entire radiator surface, this is something I may be very wrong on. Any ideas on this topic (I also need to digest the fan article more throughly)?
From my testing, and from most others' results, push vs pull hardly makes a difference in any scenario, but pull is said to be a tiny bit better (or results are the same).
Running a push/pull setup usually lowers temperatures, either by a few degress or by a bit more (my reduction was 3-5 degrees Celsius after adding an identical fan, more on this later). This is very dependent on the setup you're running - radiator, fans, and what's on your loop (how much heat). If you can't fit push/pull, you shouldn't worry too much. Maybe then you can go for the more powerful AP-30s.
n9zn-extra
Have you developed any understanding in this area where cooling increase dramatically tapers and noise increase becomes more significant?
I may have developed some vague understanding, and I've noticed that this depends highly on your setup. If you're running your radiator close to its limits, then you may notice a reduction in temperatures even when you get to very high performance fans. On the contrary, if your radiator is overkill (lets say you're only using that rad for a CPU loop), then very high performance fans will not help temperatures much at all.
I'll use my PC as en example. I run a Prolimatech Megahalems CPU cooler and an Intel Core i5 760 running at 4.2 GHz, with 1.30625 VCore. This CPU with this overclock now runs closer to 190W (up from the 95W standard TDP category, more info on this matter can be found
here). A heat sink of this size with an average fan can only dissipate the heat effectively from a 200W CPU or so. I used to run a pair of Scythe Slipstream (terrible) 120mm x 25mm and my CPU cores would exceed 80 degrees Celsius under load. I swapped to one Panaflo U1 120mm x 38mm fan (2705 RPM/114.7 CFM/45.5 DBA, about 9 mmH2O static pressure), and my core temps lowered to the mid 70s. I got a second Panaflo, and temps were in the lower 70s. Now with my Nidec VA450DC (5300 RPM/220 CFM/64 DBA, about 21 mmH2O static pressure) my core temps barely exceed 60 degrees Celsius (core 1 or 2 might not even exceed it, depending on the ambient temperature, dust buildup and other things).
On the other hand, when I ran an Intel Core 2 DUO E8400 and some AIO water cooler, swapping the stock fan (no idea on performance) to a pair of Yate Loon high speeds did not make much of a difference because of how much headroom was left.
n9zn-extra
Lastly there is the issue of fan header maximum amperage, I thought I read somewhere the motherboard headers were limited at 2 amps which is well below what many server fans require at full load. It would be nice if I could tie the motherboard fan headers to the radiator cooling fans and increase fan speed when temps climbed on motherboard components. Do you know if there are other ways of coordinating fan speed with component temperatures ( via software) without placing additional temperature sensors in the cooling system near components and purchasing additional hardware?
Most fan headers are rated for 1A or 12W. All of the fans I listed above are well below this rating (highest is the GentleTyphoon AP-30 at 0.56A or 6.72W), so you can use them with any fan header. As for coordinating fan speeds with component temperatures, try Speedfan.
post edited by boredgunner - 2011/11/13 09:11:34