Hi guys. Yeah, the state of the Mac GPGPU ecosystem is frustrating for all of us. I am a Mac guy, and miss my 2013 Macbook w/ 650M where Octane ran fine.
For newer Macs, we are suggesting users either work with an expander box (there are threads on this you can look up on the forums), or just wait for headless rendering, which is coming at latest in the 3.1 release (3.1 is next on our roadmap after 3.0.x). HR will be easier and hopefully cheaper than an expander, and will use a PC Nvidia GPU anyway on the LAN to run Octane as if it were connected to your Mac.
In the long run we are working on bringing our CUDA cross compiler framework ( http://venturebeat.com/2016/03/09/otoy- ... platforms/
) to Metal on MacOS. Today Metal it's not mature enough to support this well. OpenCL and low level GPGPU APIs is DOA on MacOS (and not great on Linux either).
We are working to implement headless rendering ASAP. We won't have a date to share publicly until we are within two weeks from alpha and know exactly when it will drop.