Suresh used a weird dynamic to explain bandwidth. I’m an advocate for the traffic explanation myself.
car = data
lane = cable or any physical layer
if you have a 3 lane highway it can handle the bandwidth of cars/data but if that car lane merges from 3:2 then if causes traffic and bottlenecks.
The cable may be able to hold a lot of data (three lanes of traffic) but once it hits the small form factor device(the two lane merge) plugged into the switch it is converted back to electric. It is not efficient enough to do this quickly so it bottlenecks.
this is why people use something called ether channel, which is where they virtually combined multiple cables so they are seen as on cable on the switch, but this takes up space, time and power. It also ultimately does not make the device converting the data any quicker, we are supplementing and instead of having 3 lanes of traffic merge into two we now have the capacity to merger 9 lanes into 6.
why would the above work? Because, it is based on the volume of traffic. Just because you have 9:6, theoreticaly your data ran fine on 3 lanes, so even though you have 9:6 it doesn’t mean you are using all that and at the same time your 6 doubles the expected bandwidth.
hope that makes sense!
After this the device it self will be the only thing not able to handle the traffic so ASICs and memory needs to be advanced to que all data being routed/switched/processed.