don't quote me on this but i think the VRM's are something like 95% efficient. so that means for a 170w GPU, you are losing something like 8.5w of heat to the VRM. you still need to cool that down, but if you are doing active cooling, it's nothing at all. same goes with RAM it's only a small fraction of what the GPU itself pulls. so the amount of actual cooling required is VERY small in relative terms. also the temp tolerances for VRM is pretty high. on high quality components that would be used on a modern GPU, you are looking at a temp tolerance of 105 degrees C. so it can get HOT and still be well within spec of the components used.