650MHz Duron @ 850MHz

All the heatsinks I have that are suitable for Socket462 processors fit on my A7V - there are no capacitors close to the socket stopping a clip from going on. Having said that, I proceeded to test the Arcticooler, in three different ways (I have two of them, which made this easier) - using the TIM, using the TIM after putting it through a "phase change", and without a TIM, just using thermal paste. To put the TIM on the Arcticooler through "phase change" it needs to be heated - to at least 60ºC, preferably more, and held there for a while. To do this, I simply unplugged the fan on the Arcticooler from the CPU fan header on the A7V, and let the temperature climb to 70ºC, at which stage I plugged the fan back into the motherboard - the CPU temperature read at least 65ºC for a good 5 minutes - which should be plenty to put the TIM through the phase change that Agilent recommend.

Here are the result for the Acticooler CA in the three stages I have mentioned.

As you can see, putting the TIM through its phase change did help - but the Arcticooler works even better with the TIM removed, using just plain thermal paste. For the final comparison, I've used the best results for this heatsink, but the other results can easily be compared if they need to be.

Next I tried the Alpha PH6035 - it certainly fit on the socket (as I mentioned before), and the pads on top of the Duron help keep it stable.

As you can see, these temperatures are high - a LOT higher than the Arcticooler. I think the fact that the PH6035 is a few years old, and the fact that it's designed for PPGA processors with large CPU-heatsink contact areas, means it's simply not able to remove enough heat through the small contact area it has with the top of the Duron core. It can also be seen that the Alpha heatsink performs better when air is being drawn through the pins of the heatsink and exhausted out the top, rather than the other way around, which is how most heatsinks (other than Alphas) are designed to operate.

Finally, I tested the Alpha PAL6035, Coolermaster heatsink and the Chrome Orb, and produced a final graph, showing the performance of all the heatsinks.

As you can see, out of these coolers, the Arcticooler CA is the best - 6ºC cooler than it's nearest competitor at idle, and 3ºC cooler under load. The Coolermaster DP5 6H51 heatsink is the surprise package here - while costing around the same as the Golden Orb, it outperforms it considerably. The Alpha PH6035 is an old heatsink, designed for different processors, and it shows. The really surprising result here (to me at least) is the Alpha PAL6035 - yes, it's equal second here, but I really thought that it may have competed more favourably with the Arcticooler CA, and I certainly thought it would perform better than the Coolermaster heatsink.