I am new to CUDA programming, and I am working on a problem that requires multiple GPUs in one machine. I understand that for better graphics programming multiple GPUs need to be combined via SLI. However, for CUDA programming do I need to combine GPUs via SLI as well?