Router vs Switch vs Hub: What is the Real Difference?

 The standard answer is a single line: hubs work at Layer 1, switches at Layer 2, routers at Layer 3. It is correct, it is useless, and it explains nothing about why anyone should care.

Here is the version that actually helps. These three devices represent three separate historical solutions to three separate problems. A hub repeats electricity. A switch makes delivery decisions. A router makes destination decisions. The layer numbers are a consequence of those jobs, not an explanation of them.

This article breaks down router vs switch vs hub by what each one physically does to a signal, why two of them are still essential in 2026, and why the third disappeared entirely.

The Hub: A Device That Thinks About Nothing

A hub is an electrical repeater with multiple ports. That is the complete description.

When a hub receives a signal on one port, it retransmits that signal out every other port. It does not read addresses. It does not know what a MAC address is. It cannot distinguish a data frame from electrical noise. It amplifies and repeats, and nothing else.

Three consequences follow immediately:

1. Every device receives every transmission. Your neighbour's traffic arrives at your network card, which examines the destination address and discards it. The data still crossed the wire.
2. All devices share one collision domain. If two devices transmit simultaneously, the signals collide, both are corrupted, and both must wait a random interval and retry.
3. Bandwidth is divided, not multiplied. A 100 Mbps hub with ten devices does not provide 100 Mbps each. It provides 100 Mbps total, shared.

Half duplex vs full duplex

This is the detail that finished the hub permanently.

A hub operates in half duplex. A device can transmit or receive, but never both at once. A switch operates in full duplex. A device can transmit and receive simultaneously.

The practical difference is larger than doubled throughput. In half duplex, every device must listen before transmitting and detect collisions while transmitting. As device count rises, collisions rise, retransmissions rise, and usable throughput collapses. A busy hub network can lose a substantial fraction of its capacity to collisions alone.

Hubs are obsolete. They are not sold, not deployed, and not recommended. They appear on certification exams because understanding why they failed explains why switches were designed the way they are.

The Switch: A Device That Learns

A switch also has multiple ports. What it does with a frame is completely different.

When a switch receives a frame:

1. It reads the source MAC address and records which port that device is on, in its MAC address table.
2. It reads the destination MAC address and searches that table.
3. If found, it forwards the frame out that one port only. No other device sees it.
4. If not found, it floods the frame out every port except the source. This is the one moment a switch behaves like a hub.
5. When the destination replies, the switch learns its location and never needs to flood for that address again.

Two properties result:

• Every port is its own collision domain. Devices cannot interfere with each other.
• Full duplex operation on every port simultaneously.

And one important limitation: a switch forwards broadcast frames out every port. All devices on a switch remain in one broadcast domain. Connecting five switches together does not create five networks; it creates one larger network.

Three ways switches forward

Worth knowing because it affects latency:

1. Store-and-forward — receives the entire frame, verifies the checksum, then forwards. Slowest, but discards corrupted frames. This is the standard method today.
2. Cut-through — reads only the destination address, then forwards immediately. Fastest, but propagates damaged frames.
3. Fragment-free — reads the first 64 bytes before forwarding, a compromise that catches most collision fragments.

The Router: A Device That Chooses

A router works with IP addresses, which are logical and assigned, rather than MAC addresses, which are physical and permanent.

When a packet arrives:

1. The router reads the destination IP address.
2. It searches its routing table for the most specific matching entry.
3. It decrements the TTL field. If TTL reaches zero, the packet is dropped and an error returned. This is what stops packets circulating forever in a misconfigured network.
4. It builds an entirely new Layer 2 frame with new source and destination MAC addresses for the next hop.
5. It forwards the packet.

Step four is the conceptual centre of routing. The IP addresses remain constant from origin to destination. The MAC addresses are rewritten at every single hop.

And the defining behaviour: a router does not forward broadcasts. Each router interface is a separate broadcast domain. This is what makes large networks possible.

The One Sentence That Separates All Three

Memorize this, because it answers most exam questions and most real design questions:

• A hub creates one collision domain and one broadcast domain.
• A switch creates one collision domain per port, and one broadcast domain total.
• A router creates one collision domain and one broadcast domain per interface.

Everything else follows from that.

Speed, Intelligence, and Cost

Speed: A hub is fastest per frame because it makes no decisions — it simply repeats. But its effective throughput is worst because of collisions and half duplex. A switch is extremely fast, with forwarding handled in dedicated hardware. A router is slowest per packet, because inspecting IP headers, consulting a routing table, and rebuilding frames takes real processing.

Intelligence: A hub has none. A switch maintains a MAC table and can support VLANs, port security, and traffic filtering. A router maintains a routing table, runs routing protocols, performs NAT, enforces access control lists, and terminates VPNs.

Cost: Unmanaged switches are inexpensive. Managed switches cost considerably more. Routers are the most expensive per port, which is exactly why networks use many switch ports and few router ports.

When to Use a Switch and When to Use a Router

When to use a switch:

1. Connecting multiple devices within one network. This is the default case.
2. You need more ports. Adding a switch is the standard way to expand.
3. You want VLANs to segment a network logically without buying more routers.
4. Performance matters between local devices — file servers, printers, workstations.

When you need a router:

1. Connecting to the internet. Non-negotiable.
2. Connecting two different IP networks.
3. You need NAT so many private addresses can share one public address.
4. You need firewall rules or access control between network segments.
5. You need to connect remote sites over a WAN or VPN.

And the honest answer for most networks: you need both. A router at the edge, and switches providing the ports.

The Layer 3 Switch: Where the Line Blurs

Modern networks include a device that confuses the neat categories.

A Layer 3 switch is a switch that can also route. It has many ports like a switch, forwards at switch speeds, but can route between VLANs internally.

How it differs from a router:

• It routes between internal networks extremely fast, using hardware rather than software.
• It generally lacks WAN interfaces, NAT, and advanced security features.
• It is designed for the inside of a network, not the edge.

The practical division in a real deployment: a Layer 3 switch handles routing between internal VLANs, and a router handles the connection to the outside world.

Your Home Router Is Not Just a Router

Worth stating plainly, because it causes genuine confusion.

The box from your provider contains at least four devices:

• A router — the single WAN or Internet port.
• A switch — the four numbered LAN ports.
• A wireless access point — the antennas.
• A modem — the connection to the phone line, cable, or fibre.

This explains a common observation: when your internet fails, devices at home can still reach each other. The internal switch is working. Only the router's path outward has failed.

Common Misconceptions

1. "A switch is just a faster hub." They make fundamentally different decisions. A hub repeats electricity; a switch reads addresses and forwards selectively.
2. "More switches mean more networks." Connected switches form one larger network with one broadcast domain. Only routers or VLANs create separate networks.
3. "A router makes the internet faster." A router chooses paths. It cannot exceed the bandwidth your provider supplies.
4. "Switches assign IP addresses." That is DHCP, which on a home network runs on the router.
5. "Hubs are still fine for small networks." They cost the same as basic switches and perform worse in every measurable way.
6. "A Layer 3 switch replaces the router." It handles internal routing well but is not built for internet edge duties.

A Quick Decision Guide

Ask three questions:

1. Do the devices belong to the same IP network? Yes — use a switch.
2. Do I need to connect different IP networks, or reach the internet? Yes — you need a router.
3. Am I considering a hub? No. There is no scenario in 2026 where a hub is the correct choice.

Conclusion

The difference is not about layer numbers. A hub repeats blindly. A switch learns where devices are and delivers precisely. A router decides which network a packet belongs to next. Each was invented to solve a problem the previous one could not, and two of the three still solve those problems today.

Try one practical exercise: run arp -a on your computer and look at the MAC addresses listed. Those entries exist because your switch delivered frames to the right ports and your device resolved addresses on its local network. Seeing that table populate with real hardware addresses makes the distinction in router vs switch vs hub concrete in a way no diagram manages.