Yes, the title can be read in many ways, and all are correct.

For the past week it has been quite hot over here, and I've been having problems with my computer overheating. So, I decided to mess around with the fans on the case to try to get it cooler. I tried out all sorts of patterns to put the fans in, I have:

1 x 120 mm fan as output in the rear of the case, this can't be fitted anywhere else.
1 x 120 mm fan as part of the PSU that sucks air upwards right above the CPU.

2 x 80 mm fans with possible placement in 4 ports, 2 on the side near the rear of the case, and 2 on the opposite side near the front of the case.

Duct tape - to cover holes if needed.

So I tried pretty much every combination of input, output with these 2 fans and 4 ports and duct tape. I know, I'm crazy, but it just pissed me off that something I assembled was not working properly.

Findings:

All the times that I put these fans as input fans (air blowing into the case) they make a lot of noise, especially if I place it right in front of the CPU fan. I think it's because the CPU fan spins clockwise while the case fan spins ccw, and so it generates some weird turbulence that sounds like a high-pitched buzz, like a giant wasp or something, absolutely unbearable (it will drive you mad).

The best and quietest solution, which is the one I am using now is to put both of them in the front 2 ports and output fans, then cover up all holes except or the other 2 ports near the rear, that way air is draw in there and across the components on its way to the front of the case. Much much quieter than the other solutions, and the one that produces the lowest temps.

The thing is, before I did this I look at lots of different sites online on how to place the fans, and they all say to have a rear output fan, and the other fans as input. Well, that didn't work for me.

I think the best thing to do is open up the case and look at how all the fans move air, then place fans so as to draw in air where it is needed. It also takes some experimentation.

I'm also wondering, wouldn't it be a lot easier and better to just to use water cooling ? I was thinking of buying something like:

http://www.google.com/products/catal...CCUQ8wIwBjgA#p

Anyone used anything like this ? Does it work better, any flaws ? I've heard mixed feeling about these things, some fear, some pros, some cons. I think I'm willing to try it if it isn't too dangerous. The principle behind it seems to be similar to car radiator ... should work, right ?

I'm thinking about this because I had to clean out a lot of dust, and it just sucks doing that.

IMHO water cooling is not to be screwed with.
Expensive, hard to maintain and can leak.

yes, a leak is what I fear, but is it really that likely ? And how is it hard to maintain ? Expensive, maybe, but not really that expensive.

Why use ordinary liquid cooling, when surely Peltier cooling will be even better?

What about condensate ?

You know I think I got the temperatures down low enough now, so maybe I shouldn't risk it. I'll mark it solved.

Thermal design is harder and more important than people acknowledge. And one of the major deficiencies of build it yourself systems is that thermal management is all ad-hoc; a case of "try this and that until it works". The result is that home built systems often scream like jet aircraft taking off.

And it is hard to do it any other way; we purchase this part and that part and stick 'em in a case. We don't do ANY thermal design, or airflow design; we just apply the basic principles: in low in front, out high in back, and hope it works.

I did a major upgrade about 3 months ago, and I actually gave thermal design some thought at that time. It is something I had been working on for awhile, to keep the predecessor system in the same case cooled down and to keep the dust out, without being too loud.

This last upgrade, I seem to have it working well. I can't say that will continue through the NEXT upgrade, but presently my Antec P-182 case (a noise suppressing case with input air filters) with the ASUS mobo and the great big (and very quiet) Evercool Transformer 4 processor heat sink are playing very nicely. If I didn't live in Phoenix, where my interior temp is often above 80 degrees F, I could keep all the fans turned clear down, which would make the system almost silent yet still retain good cooling. Given that I AM in Phoenix, I run most of the fans on high to keep it cool, and the system still makes less noise than the ceiling fan overhead or the portable fan in the hallway outside the office. There is very little turbulence - or, at least, there is no AUDIBLE turbulence - and the heat sink fans direct air the proper direction to work with the case fans while the fins act as flow straighteners. No hot spots inside the box - even though I have 5 SCSI hard drives in there.

As for water cooling, I've thought about it but fear leaks. How likely are leaks? Well, that depends on the technology used to build the water cooling system, now doesn't it. Anyplace there is a joint and a clamp, there can be a leak. If there is a chemical mismatch between the coolant that is inside the pipes and the pipes themselves, there can be a leak. I have seen copper pipes spring pinhole leaks. I've seen plastic split. I've even seen plastic crumble. I've seen clamps fail - and none of this with any exotic fluids, just water and antifreeze. None of these on computer systems, mind, but it's all in my memory...

I can't tell you if any of the commercial water cooling systems on the market are properly engineered and tested, and using appropriate components, or not. I can tell you that it seems to me that a lot of these systems started life as hobbyist systems and I rather suspect that the reliability engineering consists of putting them out in the field to find out what breaks.

To address the leak fear, why not mount the motherboard "upside down" - attached to the top of a desktop case, with the processor and expansion cards hanging down. That way any leak or drip goes away from the sensitive electronics.

I found that the best solution for quiet cooling is to use BIG fans. My current desktop has a 140mm intake fan on the front, and a 120mm out the back. The power supply also has a 140mm fan.

And one unusual thing about my case is that the power supply is laying on the floor of the case with the fan on top, instead of hanging near the ceiling of the case with the fan on the bottom. Not sure whether this is good or not, but these might be advantages:

It lets me put a big fan right atthe top to pull out all the hottest air, and the power supply gets the cool air from the bottom of the case.

But one issue I had was CPU temp. Now it's not high, but I didn't change anything. Just one thing that I did change was the video card. The original one was big and got incredibly hot, the new one isn't very hot. And, of course, it runs Linux now.

That is certainly a big part of it. My Antec has 2-120mm 3-speed fans in top and top back for exhaust, 1-120mm 3-speed fan low in front for intake, and 1-120mm 3-speed fan in the bottom power supply tunnel (which has drives in front and power supply in back) located in the middle of the tunnel and pulling air across the drives and to the PS.

My PS has a 120mm fan in it with a speed control; I usually have it turned way down. That Evercool heat sink has two 120mm fans on it in a push-pull arrangement that spin at 1000 RPM. This makes it quiet but, since the fan speeds can't be adjusted, affects its performance if you overclock heavily.

Alright, thanks for the responses, I think I'll stay away from water cooling for now, and from TEC.

You know, I think maybe I should replace the thermal paste with something better, I'm using the one that came with the CPU, but one of the cores seems to be significantly hotter than the others, like up to 10 C hotter, this might be because the paste is not set properly.

Could be the paste, could be some problem with the processor's thermal management system, could be a problem with processor affinity settings so that the workload isn't balanced.

My quad core phenom-II cycles through the cores to balance the workload (and hence the temperature distribution). You should be able to see that happening with your system. I use gkrellm and it shows the effect very clearly.

Yeah, I see that it is cycling like you say, but when it's idling, like right now here are the temps:

46, 44, 51, 46 C and from my calculations the real temperature is 15 C below these.

It's not a lot off right now but it has gotten to like 41 vs 50, so almost 10 degrees difference for that one core, always the same core.

I noticed that in my Intel Core i7 920.

Assuming the temp sensors are all uniform, then that does indeed seem like the paste.

Wouldn't be a big problem to pull the heat sink off and see if it is uniformly distributed. Of course, whether it was or wasn't, you'd still have to replace the paste then.

My CPU is tending to stay around 40C in an 80 - 83 F (26-28 C) environment.