I'm intended to install two GPU's in my system, to use cheaper one for OS display and more powerful one for rendering (Octane and other), since in that case I can use the whole available RAM of the more expensive GPU, but I'm concerned about viewport performance in modelling software. So the question is - can I setup the most powerful GPU in my system to take care not only of GPU accelerated calculations, but of displaying viewports as well? Sorry for poor english.
In that case I'm losin about 100-150 mb of RAM on OS duties, and it matters, since even for GTX 580 max available RAM on the market is 3Gb. I can't afford top Tesla or Maximus with lots of RAM, but I can't sacrifice my viewport performance either. So my only option here is to lose RAM anyway?