I currently have an old GPU card 8800GTX in my system. If I add one more card to the system for accelerating computation,
Do I have to switch my current system to a motherboard and system which SLI supports? Or SLI is irrelevant to multiple GPUs for CUDA?
I have only one PCI-E x16 slot now, can I use a extender card to parallel the two cards? Will it harm the performance a lot? Or I have to use a motherboard with at least two PCI-E x16 slots?
Since the current information about multiple GPUs inside a system are targeting around SLI, I am a little bit confused if GPUs for computation have to follow SLI’s specification or not. Thanks for any suggestions.