I read about this topic in several forums, and I have tried to read them all and followed all the weblinks of information I can find about it so far. It is just weird to me that I am having this issue.
I have a Dual Xeon x64-based system with Win 8.1 Pro 64-bit and two Nvidia Quadro K5200 (compute value 3.5) with the latest drivers (416.78). I have Blender 2.78 64-bit and Blender 2.80 beta 64-bit installed and neither will show my graphics cards under the CUDA Cycles Devices, just “No compatible GPUs found”.
At this point, I am not sure what else to do. I have read about (in the GPU Rendering section of the Manual) deleting the local Cuda include host config file (but I think I would need to create my on Blender build?) or create a build with a newer CUDA developer toolkit but these are steps I am not sure I should pursue at this point. Somewhere else I read that installing the CUDA toolkit could help but my C drive is nearly full and the toolkit is HUGE and meant for developers only.
Looking in the Nvidia Control Panel, I see where I could Maximize 3D performance by enabling SLI and PhysX(?) between the two cards. I have ordered an SLI connector and will give that a try Monday.
These Quadro cards several thousands of dollars when purchased years ago and they should still be good to go for Blender GPU rendering it seems.
Does anyone know how to get Blender to “see” these cards? Should I combine the cards with SLI or not? What is Nvidia’s PhysX and does it help?
Thanks for any answers!