Current CPUs develop a lot of heat in exchange for multi-gigahertz processing power. A little thought in the application of cooling measures should pay dividends by avoiding the 'fan assisted oven' trap which 'sounds like a Helicopter'!

Firstly, a few basics so that we are all singing from the same song-sheet! When a volume of material, air or metal etc., absorbs an amount of heat, its temperature rises by a proportionate amount. This means that if a volume of air, at say 25ºC, absorbs an input of 'X' Watts of heat, its temperature may rise to, say, 30ºC. But what happens if the air was already at 30ºC? With the same 'X' Watts heat input, its temperature will rise to 35ºC. And if it was at 35ºC it will go up to 40ºC.

If the air was absorbing its 'X' Watts of heat from a heatsink then the heatsink's temperature also follows the rule. That is, if the heatsink was a number of degrees warmer than the air and it warmed the 25ºC air to 30ºC then air at 35ºC will cause the heatsink to rise to a higher temperature in order to dump the same 'X' Watts of heat into the same volume of air. Since the cooling air now is 10ºC warmer, the heatsink will reach a temperature 10ºC higher than before.

So, recapping, if the heatsink needed to be at say 35ºC to dump those 'X' Watts of heat into 25ºC air then, when the air is at 35ºC, the heatsink will go up to 45ºC. The rise in the temperature of the heatsink exactly tracks the rise in the temperature of the cooling air.

If we increase the coupling of the heatsink to the cooling air, somehow doubling it and thereby halving its thermal resistance, it will be able to dump the same amount of heat into the air at half the temperature premium previously needed. That is, the heatsink will be cooler, only rising by half the previous amount.

That is the rule that shows why bigger heatsinks work better than small ones. They have better coupling to the cooling air by means of larger fins, more fins, grooved fins, etc. They may also have improvements on the base like a polished surface and/or a copper insert which serve to reduce the thermal resistance of the CPU interface.

What this illustrates is that the temperature of the 'hot' thing is directly linked to the temperature of the thing that is trying to cool it and if this latter goes up a number of degrees then so will the former.

There are two simple guidelines to come out of this:

• Cool cooling air lends to lower temperatures
• A low resistance thermal path lends to lower temperatures

One means of getting cool air to the CPU heatsink is to cut a hole in a side panel and fit a duct to guide the outside air directly to the CPU Fan. This works quite well but doesn't greatly improve the lot of the other heat sources in the computer like the Voltage Regulators, the Northbridge chip and the graphics card because the warm exhaust air from the CPU heatsink is allowed to circulate around the case.

A better way is to dispose of the heated air outside the case. The PSU already does this by exhausting its heated air directly through the PSU fan grill at the rear of the case. The case Rear fan should also be mounted to blow outwards since we don't want it to suck the hot PSU air back into the computer!

Naturally, recirculating the hot air doesn't help the cooling. Blowing cool air in at the front and hot air out of the back of the case is a partial solution which relies on the heated air to co-operate and be expelled through the PSU and Case Rear fans and not become entrapped in pockets by the ribbon cables. An improvement to the airflow can be made here by removing any sheet metal hole patterns to leave a clear open aperture and fitting wire fan grills to protect fingers where needed. Using wire grills also reduces the noise from the fan especially if raised 6mm or so above the sheet metal aperture as illustrated below.

Now, if the CPU fan was mounted the other way up so that it sucked air into the sink from the sides and expelled it from the 'top' then it would be easy to duct it away to the outside. [I am using 'top' here to mean the face to which the fan is attached.]

This close-up shows the CPU end of the duct which is made from two tins cut at an angle and soldered together. The end which is over the CPU fan is also cut at an angle to be more certain of collecting the heated air by fitting more closely.

This annotated view taken through an empty front panel 5.25" slot shows the disposition of the various parts. The mounting shroud is part of the heatsink and covers the top of the fins around the edges. [Note that the photo has been rotated by 90 degrees to make it easier to annotate and understand.]

This photo shows the 80mm Panaflow Rear fan with the duct attached. The duct is surprisingly light! Once the top and bottom is cut off a food tin and the label removed, what is left is only half of the 42 grammes starting weight! That makes it about 3/4 oz in English. This duct is very much a prototype which had little time spent on it since it was intended to prove the principle of the thing and find out what the problems are. Consequently, I got the fixing positions of the four feet a little wrong but was still able to catch the edges of the feet under the mounting screws.

Noise from fan vibration can be reduced by fixing the fan on soft mountings. To this end, I cut four short pieces of 5mm dia silicon rubber model glow-fuel tubing and inserted them into the corner holes of the fan so that about 1mm of it stuck out at the face that would mount on the case. Being soft material, the standard case 6-32 or 3mm screws will drive into this sleeve quite successfully and provide a very effective noise reducing mounting. In the photo above, the two light-grey 'buttons' in the fan's corners are the inside view of the sleeves.

The performance of these cooling measures can be described as excellent! With a room temperature of 24ºC the motherboard system temperature is 28ºC and the CPU temperature is 39ºC. Now here is the really good part: by using VCool and SpeedFan the CPU is halted when it has nothing to do and this reduces the amount of heat generated. It reduces it so much that I have controlled the CPU fan and the Front fan not run at all until the CPU temperature rises to 50ºC. Since I don't run background tasks it almost never gets to 50ºC unless playing games or running Prime95 Torture Test which infact can't quite get it to 50ºC most days! Therefore to further reduce the fan noises, I have put a 100 Ohm rheostat in series with the Rear Fan which now runs at about 2000 rpm and I run the PSU fan from a 6v DC regulater which, again, is around 2000 rpm. I have removed the Northbridge fan altogether and added a 47 ohm 1/4 Watt metal film resistor in series with the fan on the GF2 MX 400 close to the central motor so that it lives in the airstream.

SpeedFan's home page is at: http://www.almico.com/speedfan.php and the page for supported boards: http://www.almico.com/sfhardware.php It isn't the easiest thing to set up, but the first thing to do is check out the working range of the CPU Fan by reducing the controller on the readings sheet to see where your fan responds. Mine is right at the bottom of the range and runs full chat 5500 rpm at a controller value of 5% and then slows down to 3400 at 1%. At this speed, the 60mm CPU Fan is virtually silent.

I have setup the 80mm Evercool Front Fan on the other controllable channel (my Winbond chip has two channels) and this runs across a similar range of values. The CPU Temperature sensor controls both of these Fans and I have the 'set-point' set at 50ºC. As the CPU temp reaches this value, the SpeedFan turns up the 'go' to the the fans and they start to run up. VCool is from http://vcool.occludo.net/ and works by allowing the CPU to halt between doing things. Something in the Via chipset gets in the way of the earlier utilities that did this like 'Rain' and 'Waterfall' which don't seem to work on AMD XP CPUs in Via chipset boards.