There are many conclusions we could draw from this info, but most would be assumptions!
But for one thing, x64 will not make any difference over x32 for this issue by virtue of the 64bits alone - this is for the CPU not the GPU, so only if one version of the driver happened to have a bug that causes this issue and the other didn't - but that seems unlikely as the code for each is likely to be the same or extremely similar. Even then, there is no reason to assume that the x64 would be the driver without the bug!
I would like to think that the but is in the nVidia driver - but logic doesn't necessarily support that. If it was, then most OpenGL/SDL apps would suffer the same issues, and they do not. For instance, I have run Savage 2 which is an extremely graphics intensive application - vastly moreso than the Orbiter, with no issue at all. The frame rate on a particular application may slow right down under extreme workloads, but that still shouldn't introduce tearing. Tearing really ought to be a functional timing issue, rather than a performance issue.
Again, that doesn't necessarily point the finger at the Orbiter either. My guess (based on nothing whatsoever) is it is something to do with the way the Orbiter code interacts with the driver - so a combination of the two. But that would require someone with a really deep level understanding both of the Orbiter code and graphics generally. These people are very thin on the ground here, and otherwise occupied at the moment!