Hi everyone,
I have liked to used cuda for a while now , i want to build a mini supercomputer nothing major but just something thats fairing good. I am planning on using a combination of nvidia graphics cards and amd phenom x4 , but my question is , is it possible to run cuda across multiple system so having a master node and computing nodes and the master node send the data out to the computing node and then the computing node sends the data thats be process back to the master node in other words, cuda working with mpi to use all the power of the cpus and gpus on each computing node and returning the data once been completed.
Any information will be great