I'll be installing the cuda toolkit 7.5 on debian stretch, currently without an nvidia card. I only intend to do remote development (sync projects to create remote builds on targets) on an nvidia server with two titan blacks.
The getting started guide says I need a cuda-capable gpu, but I'm wondering if it's really needed if I never build locally.
Has anyone tried this kind of install scenario?
If so, what do I need to do to get nsight for eclipse working to create, debug, and profile remote builds?