So, you're calculating the thermal resistance by the power consumption of the CPU& VRM? Isn't that going to give us results that don't really make any sense? The heat sink is removing heat from the cpu, not the WRMs.
Yes, we are calculating based on the combined power of the CPU & VRMs. And, yes, I suppose this method will inevitably overestimate a bit. I'm not claiming it's a perfect method, but it's better than the alternative. Would you rather I used the 130W TDP when the
whole system is only drawing 120W? Using the measured number is a clear improvement.
Besides, I don't think things are as cut and dried as you make it seem. It's not as simple as "the heatsink cools the CPU and nothing else". VRM heat has to be dissipated somehow, and the heatsink plays a role in doing so. On the other hand, no heatsink will remove
all heat from the CPU; some heat will always be dissipated through other avenues, such through the motherboard. Motherboards are designed to conduct heat for just this reason.
Because of this, getting a completely accurate number for the amount of heat in the system is unfeasible. And, to be honest, it doesn't really matter that much, for the same reasons that the absolute CPU temperature doesn't matter that much. We're not looking for 100% accurate numbers, because we've accepted that we can't obtain them. However, we
can get numbers that will be valid for comparisons made
within our system.
Switching to the measured heat may not be perfect, but there's no question that it's more accurate than the method we used previously. It's also more repeatable, because it eliminates a previously unknown variable: The amount of "error" between the rated TDP and the actual power dissipation of our chip. It will now be easier for people to replicate our work because that variable will no longer be a source of error.
Measuring the CPU heat has further benefit: It allows us to tell when the power consumed is abnormally high, as can happen when the system is beginning to overheat. An increase in consumed power could potentially skew the results near that point, and measuring the power eliminates this source of error.
I'm still very confused. In the Ultra 120 review, the CPU cooled that chip slightly better/on par with the Ninja, yet in this situation, it does not. What was different in that test bed? If every variable was the same in that test station for the testing of the Ninja and Ultra 120, the temperature measurements should still be valid, regardless of how you were calculating thermal resistance.
I'm quite aware of how the test
should have turned out, but if I could report that, there would be no point in re-testing. All I can do is report how the test
did turn out. I'm in the same boat as you: I can't explain it, but neither can I ignore it, no matter how much more consistent it would make my work seem. If you can help me put my finger on what went wrong, I'd be happy to listen. However, unless you can explain it, all I can do is report what I measure, and write off unexpected variation as random variation.