These temperatures where taken after it had stabilised using folding@home for 100% cpu load. The temperatures were taken from the socket thermister as I didn't feel putting a probe inside the cooler would give good results. The CPU wattage is estimated using a peice of software called radiate.

Not brilliant eh? A lot of HSF's perform as well as that.

So how can we improve on this design? My idea was more volume of water (as it could be possible it's getting saturated very quickly esp. with my rather poor pump). Running the water nearer the core might also help to bring the cold water straight to where it is needed. Extra surface area would definetley help, but using artic silver epoxy and a heatsink is more likely to hinder the performance becuase of the inefficiency of transfering from the core to the sink.

So, basically I set out to make a slighty larger volume cooler with water running going STRAIGHT onto the core. Me being the "on the cheap, home made" sorta guy - I went looking around the house.

I know what your thinking, what the HELL has Spode gone and found this time, another jam jar?? No, I found a sunny delight bottle..... you'll be the ones laughing later THANKYOU!

I chopped the top third of the bottle off to get what you see below. I cut a couple of holes and epoxied the plastic barbs in. Because the block is so light, it doesn't need a clip - instead it will just be held on by epoxy putty. The top barb has a small amount of 6mm hosing coming out the bottom which directs the water directly to the core.

Here the cooler is in action. It's not the prettiest thing in the world, but lets see how it performs!!

For some reason (possibly due to not so good pencil markings) the voltage dropped to 1.90v.

Hows THAT for a Sunny Delight Bottle!! I think you can safely say that the peformance is quite good. Now, for something to compare to - here is the Socket AHO cooler from 3dfxcool (just one I had lying around waiting to be reviewed).

So, my cooler is definetly good (either that or the Socket AHO is not that good).

Conclusion

For a block that cost basically nothing to make, it performs a treat. The downside to it is that no peltier can be installed, BUT what will be really good is when a chiller unit is installed to this system to lower the water temps (as good as using a peltier). As the Duron actually has less surface area than the thunderbird, I can see the results from the thunderbird chips actually being much more impressive - someone give me a 1.5 tbird!! Unlike using normal epoxy which is pretty permant, epoxy putty can be chipped off - I managed to take the cooler off 3 times while experimenting with different ideas. All in all, I think direct die cooling may be a good way forward for a lot of us.

Spode

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