My options in cmake are:
cmake -DBUNDLE=cudaminer -DCOMPUTE=30 ..
And I also tried:
cmake -DBUNDLE=cudaminer -DCOMPUTE=52 ..
Well as a test, had the 108 (current default git) been working for you? I had an issue with 110 (compiles as 1.1.2) crashing, Genoil advised me to go into this folder ~/ccp-ethereum/libethash-cuda and modify the CMakeLists.txt to remove references to 61.
Example, mine now looks like:
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};-gencode arch=compute_35,code=sm_35;-gencode arch=compute_50,code=sm_50)
But if you only have 750ti, you should be able to just use:
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};-gencode arch=compute_50,code=sm_50)
Then try your cmake, with just:
cmake -DBUNDLE=cudaminer ..
I don't think the other options will be needed anymore, I didn't need them since that is what I edited out.
If none of that works and I'm not helping at all, can you try starting with the -G and see if opencl is working? At least that way you know the drivers are at least loaded and seen. Maybe we can go from there. Also, what flavor of Ubuntu? I'm using 14.04 as I could not get it working on 16.04.