You also can't assume that just because memory is utilized, it cannot work with less unless you have visibility in to exactly how memory is utilized.
Take a win7 system for example. The more memory I throw at it, the more will be utilized for caching (which at a certain point becomes an imperceptible performance benefit). Without visibility in to that mechanism, I may assume that I need 8+GB of memory to do anything, as the most simplest of tasks leave me with so little unused memory. The truth is, though, that if I demand more memory for user processes, that memory will be freed up from the space currently used as cache.
Memory management is far more complex than "Oh, I'm using 95% of my available memory, I need more!" Not in the least bit. A well designed system dealing with more potential data than it can fit in RAM will always use as much of that memory as it can (leaving enough free to handle things that may need to occur faster than the cache space can be freed) to cache data such that it doesn't need to fetch it from another location. That cache can always be freed up if needed.
So these rudimentary measures are practically useless in the case of video cards where we have absolutely no visibility in to how memory management is being performed.