The full native support for Volta architecture cards exists in CUDA versions 9.0 and higher. This version of CUDA will be integrated by MATLAB in future release.
If you are intending to use an NVIDIA Volta card with Deep Neural Networks in MATLAB (including in the Computer Vision System Toolbox) please be aware of the following bug report
where Volta cards produce an error for all MATLAB versions.This error will have the identifier 'nnet:cnngpu:CuDNNInitFailed'; and the message
ERROR: 'Failed to initialize the cuDNN handle. Return code was CUDNN_STATUS_INTERNAL_ERROR.'
All other MATLAB GPU code is forward-compatible to future architectures using a byte-code representation that allows it to be compiled by the graphics driver at runtime. No NVIDIA CUDA 9.0 compiler was available when R2017b was released for native compilation, however this runtime compilation offers limited support with possible performance impact.
When MATLAB encounters a card with an architecture the GPU code was not originally compiled for there will be a one-time delay for GPU computing commands whilst the byte code is compiled and cached. Users may find the need to increase the CUDA cache size to prevent this delay recurring.
We recommend achieving this by setting the environment variable:
CUDA_CACHE_MAXSIZE 536870912
This sets the CUDA cache size to 512MB which is the minimum we would suggest. The default for this is 32MB. For further instructions on how to change environment variables on your system please consult here.
If you run into a specific limitation using GPU computing in MATLAB with a Volta card not detailed here please contact MathWorks Support: https://www.mathworks.com/support/contact_us.html
