Test setup:      Asus P5VDC-MX
                        Celeron D 391 @ 2.93 Ghz
                        One SATA drive
                        512MB pc 2100
                        No other fans or peripherals installed

                        Room temp during testing ranged from 18C up to 22C

To truly test the performance of this heatsink it was installed as the only operating cooling method other than the fan in the PSU. There are no other intake or exhaust fans installed. The only configuration tried here is the one where it is blowing the air out the back of the case. First I tried it with the included thermal grease from Zalman. The temps here were quite impressive, only a (degree) difference between the motherboard sensor and the cpu. Next I tried a different cooler with some Artic Silver 5. The cooler resembled the stock cooler that usually comes with the Celeron chips, but was actually a aluminum/copper sink made by Masscool.

This is where I saw where the reputation for the hot Prescott cores came from, I had problems keeping this chip from setting off alarms. I thought it might not be set properly so I removed it, cleaned it and reset it, and still was getting high temps. Letting this one stay on there a week for the AS5 to cure was a hard decision and I was glad to take it off and put the Zalman 9500 back on, but this time with the AS5. Immediately I saw a huge drop and even before it cured the temps were lower than before by 2-3 degrees. Once fully cured the temps dropped another 2 degrees. I wanted to see how this performed at max speed so I turned up the fanmate to top speed and let it run on full load for 3 days and recorded the temps again. The difference in sound between medium and top speed is a noticeable difference (silent vs a slight hum).

Since it wasn’t being heard at the mid-range setting I didn’t see a need to test it out at the lowest speed. No benchmarks were run with this overclocked, nor did I run any video benchmarks . Even though the temps show a warmer temp at full speed, it is still only 1 degree difference between the motherboard and cpu, so I’d say the temps stayed within the margin of error.
When testing this in a warmer room (standard summer conditions ~75 F/ 24 C) the temps increase on full load to 43 C