The source and the sink negotiate a link rate that allows error free communication, based on the number of errors seen on a known pattern over a certain amount of time during the link training phase.
When the established link rate is too low for a requested video timing, the source requests a less demanding video rate.
It’s why the control panel of a GPU might now show the video timing that your monitor officially supports.
In modern GPUs and monitors, an alternative is that DSC (display stream compression) is used, and that the compression ratio is cranked up to still fit the requested timing in a reduced availabile bandwidth. The image quality impact is usually unnoticeable.
Normal DP cables in particular are entirely passive, there is no chip or coding in them that would tell source/sink what kind of cable it is. IME the most likely result of a marginal cable will be that the screen sometimes goes black for 1-2 seconds or shows an artifact here or there. With DVI, which just directly blasted red/green/blue bits on separate wires at the pixel clock (bad for EMI) it was pretty easy to end up with a cable that had issues just on the twisted pair for a particular color.
That’s why the source and sink need to determine the maximum rate literally based on trial and error.
The source tries a link speed (and other parameters such as drive strength and pre-emphasis), the sink reports back home many symbol errors it detected. Rinse repeat until a stable configuration is found.
For marginal connections that’s often not sufficient. After the introduction of DSC, DP has mandatory support for Reed-Solomon forward error correction, which is a great help in fixing occasional error burst. It can be enabled when DSC is off, but I don’t know that happens in practice.
