So if you look at the temp values in the chart above your post you can see that indeed I perform a warm-up cycle. Thus far I have not been able seen active down-throttling on the boost values with long term usage. Typically imho that would/could happen with the founder editions as these cards have a temperature limiter set at roughly 80 degrees C, and it in fact that cooler reaches that number. At that moment the hardware will try and protect itself by lowering clock/voltage value to get closer to that 'safe' 80 degrees C again. The AIB card however all have cooling well below that thermal threshold of 80 degrees C, since these cards run say 10 degrees C lower, they will (should) not run into that issue. That boost clock will only go down if the graphics card measures a valid reason for it. But it is complicated with GPU Boost 3.0, I'll give you that. Nvidia has so many sensors up these days. In short - Thus far I have seen very little boost variation at the end of a two hour benchmark stretch. I am not saying it can't happen. But the chances with properly cooled AIB card pretty much should prohibit that behavior. Also and let me add this, you need to keep things in perspective. If the card did reach a thermal limit and adapts by downclocking 50 MHz on the boost clock ... honestly you are not going to notice it in performance. You need more orders of magnitude for that say 100~150 MHz would result into maybe a 2 FPS lower framerate on a 60 FPS average.