ED NOTE: Nevin over at Arctic Silver sent this in:
I bought a few new heatsinks at the Fresno computer show to test out how well in-socket thermistors compare to CPU temp readings using Intel’s internal diode.
When I looked closely at the Chrome ORB mounted on the Celeron I used for testing, I saw that it was not contacting the pads I use to stabilize the heatsinks and was actually canted on the core. The problem is that the center round portion of the base extends approximately 0.01″ beyond the rectangular portion that should contact the pads. The attachment clip on the Chrome ORB does not center the pressure on the core so the heatsink tilts toward one side.
By playing with a second layer of padding (Bigger pieces on the low side, smaller pieces on the high side to compensate for the uneven pressure.), I was able to achieve a mounting that was much more flush to the core. This dropped the thermal diode temperature 4.5 C, but did not change the in-socket thermistor reading.
It should be noted that the Super Orb does not have this 0.01″ shelf and sits flush to the core despite having very a very similar attachment clip.
The internal diode was calibrated to reflect actual core temperatures (-3 C correction applied to all diode measurements).
I placed a thermistor in the socket to measure CPU back temps per 99.999999% of the heatsink reviews on the web. I removed the PCTC prior to first use to avoid as much contamination as possible. Arctic Silver was used on all sinks.
Stock clip pressure was used (Since these are all Socket 462 heatsinks, the force is actually high for an Intel CPU). I used
custom corner pads on the CPU to insure level mounting of the heatsinks. Ambient temp was 21 C, ambient case temp (heatsink fan inlet) 25 C. Each heatsink was mounted and tested 3 times with RC5 running and using Motherboard Monitor 4.17.
Here are the results including the first tests with the Chrome Orb not flush to the core (Chrome Orb1) and with the mounting pads modified to level the Chrome Orb. (Chrome Orb2).
The in-socket thermistor shows a 4 C spread vs. the internal diode temp’s 9 C difference – a very different perspective on heatsink performance.