pipeline input - performance advantage using $_ in PROCESS block vs $input in END block?

Hello,

Are there performance reasons for using the pipeline input and $_ in a PROCESS code block as opposed to using pipeline input but just using the $input variable in the END code block?

My assumption is that that's the point, you should use the PROCESS block so the next link in the pipeline can start processing data as it comes in.. this conveys a speed thing to me. I've also seen examples where $input is just used in the END block, which I assume means all pipeline data has to get there first, then be processed...

does it matter? is this just a design preference?

Thanks!