An article that was ported over from Dr Surlyjoe's old site (which IMO is one of his best ideas) was direct die cooling on a celeron 533. The great thing about the celeron, is there is just a die on the top of the chip, so it's easy to submerge. When it comes to the AMD Duron, which I wanted to tackle - there are many resistors and bridges that could possibly short out. This made a small problem, but not a huge one. As well as my own idea for a block, Dr Surlyjoe sent over his original direct die cooler for me to test.

I knabbed these pictures from the original article, to give you an idea of what his cooler is.

What the cooler basically is - is a cube of plexiglass with a large drilled out hole in the middle. The water comes in the bottom near the die, runs over the die (which is covered in a thin layer of epoxy) and is taken out the top. From the jam jar waterblock I attempted, I can see that the Doctors main mistake was the epoxy. Epoxy is an insulator and doesn't conduct heat very well.

My first plan was to cover the resistors in epoxy, but NOT the core, then attatch the block with epoxy putty (this can be chipped away easily for later removal). You may be thinking I'm mad submerging the die directly in water but I have reasons for doing it.

1) the core already has that blue layer of glue on it, so if the chip is new (i.e not chipped) then this layer of glue should be watertight.

2) I once saw an article over at OcShoot where a lapped PIII had become permeable and taken in condensation - this was cured by putting it in the oven.

Basically, I think the chances of the water actually getting into the core are minimal and even if it does - just stick it in the oven! If this did happen to become a problem, my 2nd plan was to cover the core in Artic Silver Epoxy - at least it transfers heat better than standard epoxy.

This is Dr. Surlyjoe's cooler after I have epoxy puttied it to the chip. It holds solid enough, but I used the clip to attatch anyway. The chip in question is a Duron 700. The board I am using is the AM-35 EC (KM133 Chipet). This board has no multiplier adjustment but will run at 133mhz - so I have two speeds to test with, 700 and 933 - both with 1.95v. I used an old heater core for a radiator with an 120mm 138cfm fan blowing on it. I am using the Senfu water pump with 10mm hosing.

<< Back | 1 | | 2 | Next >>