I had this setup working previously on macOS 10.12 and some earlier version of CUDA, but I have since upgraded to 10.13.1, and am struggling to set it up again.
I have a Thunderbolt 3 eGPU housing a GTX 980 connected to my Mac. It is successfully enumerated and listed in the system’s hardware tree. I also have the following drivers installed:
- CUDA Driver Version: 9.0.222
- GPU Driver Version: 10.27.6 378.10.10.10.20.107
Along with that, I installed the CUDA SDK 9.0.176, which is the latest version, and built deviceQuery with Xcode 8.2. When I run it, I get:
CUDA Driver Version / Runtime Version 9.0 / 0.0
cudaGetDeviceCount returned 38
-> no CUDA-capable device is detected
Result = FAIL
This device should definitely be CUDA-capable, so I’m not sure where to proceed from here.