Must agree - a better metric for performance is simply deltaT (using a constant thermal load). A lot of words below, just pointing out the obvious reasons to not use absolute temp values.

1) Change in T(ambient) (which is usually small) does not change the heatsink degree/watt efficiency relationship, so just baseline graphs at T(ambient) to show deltaT. We know the /W output is constant (same processor each test), so it's fast to visually identify performance/efficiency.

2) The way/rate efficiency decreases when W increases (heatsink saturates/overloads) can be measured using deltaW. Rate of change in deltaT for W is then a convenient metric.

3) Plotting from any systems arbitary '0' degrees (and what systems zero point? 0 K is different to 0 F is different to 0 C) means that any visual ratio estimates made from the graphs are incorrect, as the bars are far too long. The T(zero) to T(ambient) region is not part of the 'performance' of the heatsink. Imagine if you used a 0 K baseline - huge real performance differences would appear to be only a few percent.

4) Heatsink performance between reviews with the same test system are directly comparable, since data is independent of T(ambient) for every test.


Minor point really - all up, as usual, nice review.