Do I need an integrated graphics card for cuda or opengl & GLSL debugging?

I have a laptop with quadro 2000m, but it’s still quite slow when running some cuda programs. Now I want to buy a desktop with 2 * GTX 780 , for the CPU part, I would like to choose between i3770 and e3 1230 v2(this cpu ships without integrated graphics card), by the way, I have two Dell U2312HM. From the Parallel Nsight user manual, I know that it’s recommended to debug CUDA programs under headless configuration. So here is my question:

  1. Do I need an integrated graphics card to drive my monitors and let two GTX 780s run under headless configuration when debugging CUDA code?
  2. I will also write cuda programs together with opengl & GLSL, so if I want to debug in local setting, is it necessary to use the integrated graphics card for driving the monitor?
  3. which debugging model is better? local debugging or remote debugging ? if remote debugging is better and also convenient, I’d like to buy e3 1230 v2 which is cheaper and has nearly the same performance as the i3770 does.

Thank you in advance.

  1. Integrated graphics won’t be necessary. Not plugging in a monitor to either card will make the card headless. However, if you need 2 headless cards, you will need a 3rd adapter.

  2. It shouldn’t be necessary, however please note CUDA and OpenGL interop debugging is currently unavailable and will only be working in a future version of Nsight.

  3. It really is up to your needs. Neither way is necessarily “better” than the other.

Thank you for your suggestion. It helped a lot.