Developing code in CUDA in a virtual GPU or on HPC

Accelerated Computing CUDA CUDA Programming and Performance
1

I have DGX station and develop accelerated computing code using CUDA.

I am considering using virtual or cloud hardware in addition to my DGX with a perpetual license.

I’d like to know if it is possible to develop on such virtual platforms? for instance, can I open Eclipse and write and compile code on a virtual GPU environment?

Can you list some possible platforms/solution providers for this?

thanks
Dag

2

The entire CUDA toolkit (minus the GPU driver) can be installed on any compatible platform, with or without a GPU. This enables code development and other activities (such as usage of a profiler GUI on profiler files captured elsewhere in-situ).

This should be possible on any x86_64 cloud instance on any of the major CSPs, such as GCP, AWS, Azure, etc. I won’t be able to give you a recipe for this, and running a GUI on a cloud instance may require specific setup steps and/or software packages. The treatment of that is not unique or specific to CUDA.