This is the Spode's Abode archive. The old articles and forum have been kept here purely for historical purposes and are no longer updated.

Not all portions of this archive may work as expected.

What you will need:

This method of applying thermal paste will hopefully reduce your processor temperature a bit - although I can see it making more difference on large core processors like the PPGA Celerons or classic Athlons with a heat plate.

For a CPU to be cooled properly, there needs to be a good contact between the surface of your CPU and the heatsink base. Some cheaper heatsinks come with a thermal pad, which is a thin pad of modelling-clay like material (I doubt it is modelling clay!). What it does is fill the imperfections between the bottom of the heatsink and your CPU. Most older CPU's have slightly concave surfaces (like a dish) as I observed on my PPGA Celerons. Only the corners of the processor touched the heatsink. Obviously this was wasting most of the surface for thermal conduction between the heatsink and CPU. The thermal pad fills this gap.

Most high-end heatsinks come with a small sachet of thermal paste, which is similar to a thermal pad - but requires you to apply it to the surfaces. As most of you will know, thermal paste works much better than a thermal pad in filling the imperfections and bridging the air gap between the surfaces. I think this is mainly because of the consistency of the materials. The thermal pad is quite hard, whilst the paste is well, paste! This allows it to run into the tiny gaps and fill them easier than the thermal pad. Then I had an idea! If the thermal paste were thinner it would work better, but if it were too thin it would run out of the place it was needed and be rendered useless. This is probably why the pastes on the market have the consistency they do. The reason I think that it will work better on large contact areas is becuase the pressure per square CM on a Duron with a core of about 1.5cm² is probably about 7.5 pounds per cm², while on a classic Athlons, with perhaps 60 cm² of surface will have around 0.167 pounds per cm² (pressure=force/area). The extra pressure of a Duron helps to squeeze out the excess paste. Both these values were worked out assuming the clip made 10 pounds of overall pressure.

This had got my mind going and I was thinking of different ways to thin the paste down enough. I knew that water would not dissolve the paste at all, and if it ran onto something electrical could do some damage! Then I hit on the idea of alcohol - I knew that it dissolved the paste and evaporated quickly - leaving no trace! So began my search around the house for some alcohol. I found a squirty can of 99.5% pure Isopropyl alcohol (used for cleaning keyboards and the like). I started off by testing it on an old 486-66mhz CPU and mainboard I had lying around, to see if the alcohol damaged anything. I put quite a copious quantity of thermal paste onto the processor, and smearing it around the entire surface. I then sprayed a squirt of alcohol onto it, and clipped the heatsink down. I left this for 20 minutes while I had a snack and came back, fired the machine up and it worked fine! The heatsink I used was quite small and had no fan on, so got quite hot after a while, in fact too hot to touch, (50+°C) this is normal for this setup. I turned the system off, left it for 5 minutes so I could remove the sink without injury, and unclipped the cooler. I was surprised to find the heatsink very hard to remove - the thermal paste was creating suction between the CPU and the heatsink. This made me think that it had actually thinned out like I wanted it to! Also, when I removed the sink, I found that the alcohol had evaporated, leaving thermal paste in its original form again, just in a VERY fine layer!

Now for the real test - A Duron 800@1050 (1.85v) cooled by a HUGE heatsink, (larger in all dimensions than my FOP38 - think its 80x65x55mm!) with a YS tech 60mm 26cfm fan on it. I repeated the applying of thermal paste to the processor, and sprayed some of the alcohol onto it, then mounted it all back in my system. The case is a full tower, and the internal temps were 21c for all tests.

	Idle	Full Load
Normal Application	27c	41c
Special Alcohol Method	27c	39c

Conclusion

Well, as you can see in the results, it shaved 2c from my load temperature - which isn't bad for 20 minutes messing about. It didn't allow me to overclock any higher though, but as some people say - a 10c drop in processor temperature lengthens its lifetime by a factor of 2! Well, its going to last a bit longer (although I do not expect to own this processor by the time it becomes even near the end of its lifespan, as I will have sold it, or more likely broken it by then!)

This will not work with all thermal pastes - I had a small sachet of paste that came with a heatsink that wouldn't dissolve in the alcohol, it just beaded up.

 

David Artscan
Feel free to e-mail me for questions and maybe drop by my web site.

Note from Spode: Using alchohol near a heat source could be dangerous as it is flammable - so make sure it is all evaporated before switching on!

Update: (from feedback)

"Rubbing alcohol contains oil isopropol has only grain alcohol & water thus it will evaporate completly. From my time as a tv/vcr repairman I can tell you this simple mistake has killed literaly thousands of vcr's from attempts of well meaning but misunderstanding do-it yourselfer's. Do not use it!"