Originally posted by: soccerballtux
My TR-Ultra-120 and processor did not benefit 1C from lapping both, solidy and to a nice finish that was nearly mirror. Lapping is overrated in my opinion. Still was fun though, and I might do it again, because trying to get better temps is fun
Well, soccerballtux, I appreciate your input here.
Even so, let me observe this. First, in the best tradition of scientific tradition beginning with Bacon, Descartes, and the 17th-century contribution to the "history of ideas," there is a difference between "Knowing," "suspecting," "believing that you know," and [God forbid] -- "believing."
Let me summarize my "method."
1) I ran Coretemp while testing stability and temperatures after taking each successive approach to cooling.
2) I turned Coretemp's logging feature "on," setting it to sample at 8-second intervals and log at 8-second intervals.
3) I carefully controlled room-ambient, using a digital thermometer that I had also calibrated with a bowl of ice and ice-water for an hour.
Here are my results:
1) Switching from TR-Ultra-120-[original] to TR-Ultra-120-Extreme: an improvement in load temperatures, same room-ambient, same over-clock setting, of -5C degrees to -8C degrees. The range of difference there was a comparison of my results with review results, and the difference between lapping the U-120-Ext and installing it without lapping.
2) Diamond thermal paste: measured improvement, at same over-clock setting as (1), shows an improvement between -2C and -4C degrees, same room ambient, same over-clock setting.
3) Lapping the processor cap to bare copper: Tested on a Q6600, an E2140 and an E2180 respectively, a -4C to -5C improvement in load temperatures at same and/or normalized room ambients -- across the board.
Room ambient is very important in being able to see and measure these changes. From my experience, still following "the method," I observed -- and others have probably observed -- that changes to room-ambient affect all component temperatures linearly, and degree-for-degree. Thus a 1-degree increase in room-ambient means a 1-degree increase in load-temperature -- with absolute certainty. The only way this relationship becomes "an approximation," or varies slightly in accuracy, arises from the operational ranges of certain heatpipe coolers. Those coolers may be more effective, for example, at room-ambients of 55F and above. But any loss in effectiveness at lower room-ambients is completely obscured because the cooler doesn't need to be quite as effective at those lower temperatures.
In any case, we measure the impact of these incremental cooling refinements in a range between 68F and 80F. In the event our room-ambient is lower by some few degrees than this range, the linear relationships still remain overwhelmingly effective, so normalizing the results under these assumptions produces pretty accurate extrapolations of load values at higher room ambients.