I thought about buying the Nano to extend my GTX 960 GPU and outsource for faster processing as a small scale alternative to cloud computing. Is the nano made for this or are there better alternatives? Can it be used to suggest to the system there were more gpus at once?
There are folks who use Jetson’s to make mini clusters as a way of getting into HPC-style architectures with lower cost, and also green HPC for energy efficiency. Most commonly they get connected over Ethernet through a gigabit switch, and you can install MPI which is a common multiprocessing framework in HPC.
On the cloud side, you can install IoT frameworks on Nano like AWS Greengrass to extend your cloud capabilities out to the edge.
To compare both, how much more GPU-memory will the Jetson offer and what speed increase should I expect, combined with my current GTX 960 setup?
I’m not sure what variant GTX 960 you have, but the Nano has 4GB of memory and 128 CUDA cores. Since the 960 has 1024 CUDA cores, the Nano is unlikely to be faster than it. At only 10W, it will save a lot of power though!