I am watching a video course from Brocade. The instructor says that "all devices on a single collision domain must operate with the same speed parameters — for example, 100 megabits; full duplex. This lowers the speed of your network to the slowest device on the segment. When a bridge is used to separate collision domains, each segment can operate a different speeds." In the slide is a picture of two bus topologies separated by a bridge. Is what he said accurate? If I connected two PCs to a hub — one with a 10 Mbps NIC and another with a 100 Mbps NIC — would the switch force the 100 Mbps NIC to operate at 10 Mbps? Or is it possible he's referring to really old coaxial Ethernet setups rather than the twisted-pair hubs?
Ethernet Hubs – Speed and Shared Collision Domains
ethernetspeed
Best Answer
An ethernet hub is nothing but a single cable inside a box with multiple RJ-45 connectors just like this:
The devices connected to the hub are connected among them using a single cable (technically it's called a single domain).
They must be on the same speed, otherwise the network won't work. Network cards nowadays can use 10/100/100 Mbps per second (through autosensing they can use the same speed as the network), but there are older cards that are fixed to 10 Mbps or 100 Mbps so in that case the card must be in concordance with the desired speed on the network.
Devices on a hub must use half duplex because they need to be aware of possible collisions on the cable. Full duplex can be used on switched environments where there is no need of collision-hearing.
If you use a bridge to connect two single domains then each domain can be in a different speed and the bridge controls the speed difference, like this: