I'm aware toothpaste didn't score too highly.
Crest did in the test in the above video...
Thinking about it, though, there may be a flaw in the apparatus. If the mating of the 2 surfaces is too perfect, it can create conditions ideal enough to minimize variance between TIMs that would be misleading. Ignoring the lack of longevity, for example, one might as well use Crest given the result vs the cost.
I don't remember the wattage used or the heat produced, but the variance was so small as to make it useless to buy the expensive products. That testing ignored potential differences in results at different degrees/watt. Not all TIMs perform consistently from the lowest to highest temps. Additionally, although he controlled quite precisely for temp, he limited the test to 12 minutes, which I suppose can be used to calculate certain specifications (I hesitate to suggest which since my math skills have deteriorated severely since I was in school 4 decades ago), but it doesn't give the ability to determine at what point the TIM will become ineffective, dry out, improve, or otherwise indicate any different characteristics that would become evident through more thorough testing. This sort of complexity has always been what makes me skeptical of lab test results - sometimes we introduce new variables while eliminating other ones, and sometimes the results are achieved in an environment so divorced from the reality of the situation (in this case, PCs) that they can cease to be meaningful.
When testing in a PC, we see results that fluctuate and that may be influenced by updates and other background tasks (including malware), effects from other parts of the PC, and so on, which can end up confounding things, but that is exactly the reality faced by computer users. There is a line between the real and lab worlds that I think testers need to walk on, making sure also not to produce results that are irrelevant because of being too sterile and remote from reality. Some results are unlikely to be easily replicated unless the methodology is followed and the same equipment is used (assuming the condition and calibration are the same), but sometimes the most important factors are actually the least important because the difference between the best and worst products is so small that it won't affect most people, except for extremely heavy loads like 3D rendering, AI, saving & loading exceedingly large files, etc. Does it really matter if one TIM only rose to 31 C and the best to 29.7? Not really, if both can handle the same max temp for the same duration, and have similar lifespans, there are few practical considerations left other than the risk of pump-out and price. Why pay for Arctic Silver when you can get almost the same results with MX-4, right?
Ranking without being able to tell the range can be problematic unless there is some sort of indication that shows it is large or small.
Any thoughts?