Yes, you can run any CUDA and OpenCL applications on the eGPU.
Most CUDA libraries and frameworks like Theano, Torch, Caffe, and CUDNN are supported.
Note: As of version 1.2, TensorFlow no longer provides GPU support on Mac OS X.
Thensorflow with GPU support can be used via Boot Camp and Windows 8.1 or 10.
Here is a feedback e-mail from one of our customers who is using the Titan X eGPU bundle:
"For future queries, we found that we can get >30x acceleration over CPU for our deep convolutional neural nets. Common is to train and test deep learning models on GPU’s because it is much faster.
We used CUDA-TORCH libraries and had no problems getting the same speeds with the eGPU as if we were connected to our GPU server."
The optimal configuration for GPGPU is the Titan Xp Sonnet eGFX Breakaway Box bundle: https://bizon-tech.com/us/sonnet-egfx-breakaway-box.html/#470:1617;472:1622;481:1660;483:1665