Hello, I’m building a machine specifically for CUDA use. I’ve read on the toolkit’s release notes that it’s recommended that 2 GPUs are installed: one for computation and another one as display adapter.
Now, will this work for an onboard Intel chip and a recent NVIDIA GPU?
This thread seems to say it’s not a good idea:
Does this still apply, or is there some trick to use CUDA without needing 2 NVIDIA graphics adapters?
The operating system is, as the forum would give away, Windows XP.
Edit: Also, if indeed 2 NVIDIA cards are required, is it needed for them to be in SLI mode (thus using a SLI capable motherboard)?