So my question is, do you think with those graphics cards we would be able to run your program smoothly?
Please refer to the guide on hardware for Lumion. LUMION 3: Minimum hardware requirements
The Passmark: Videocard Benchmarks
web site provides the guide as to performance of GPU's.
As you would appreciate, we are not able to provide definitive recommendations about running Lumion with your different scenes.
The complexity of scenes, referred to in minimum requirements is important to take note of.
Quadro cards that you currently use are not supported for Lumion, although some users do use them and get by OK. It depends on the actual scene requirements.
GTX cards (or equally performing Radeons) provide the best GPU platform. Apart from the number of cores in the card, the more VRAM there is the better, as this allows for more or higher resolution textures.
In regard the 660Ti, it has a Passmark of 4663 and that is sufficiently above the passmark score guide for moderately complex scenes.
The GTX? 480 is still near top of the Passmark list at 4306, and I see there are still plenty available on eBay etc, however they only have 1 or 1.5GB and are now 'old' in regard support for real-time shaders.
Budgets are always a constraint, but sometimes it is worth spending some marginal extra $'s so that it helps in possible future projects. You may find that the 680 that supports 4GB (some 660's do also) will be a better medium term option. Of course, NVIDIA are now in process of releasing the next generation 700 series to follow on from the Titan.
There are some cases where CPU and other hardware can cause bottlenecks for the throughput needed for the GPU Windows 7 +, definitely as 64bit, lots of RAM and a fast CPU can all help. I have an old Q6600 which is noticeably causing problems and so is being replaced soon
Lumion v3 is more demanding on PC's due to features such as real-time GI, Reflection Planes, 3D Clouds, etc. Your scene descriptions do not sound too heavy, but by the time you add in the entourage, and then effects and try to run that in real-time as fly-through during client presentations, then I would suggest trying to go for as much bang for your $ as possible (and then some