Almost always a broken or wrongly wired pair. A gigabit link uses all four pairs and a 100 Mbps link uses two, so a channel with a fault in one pair falls back to the rate that does not need it and works perfectly at that rate.
A link at 100 Mbps between two gigabit ports is a link that has negotiated downwards, and negotiation is a fair process. Something told it the higher rate was not available.
The usual cause, in one paragraph
Gigabit operation over copper uses all four pairs. The 100 Mbps rate uses two. So a channel with a broken conductor, a badly terminated pair, or two conductors taken from different twisted pairs can fail at the higher rate and work faultlessly at the lower one — which is why the link is up, stable and slow rather than absent.
What to check on a link that fell back
Swap the patch leads. Both of them, one at a time, for leads known to work. This costs a minute and finds the cause a large proportion of the time.
Try a different port at each end, and a different switch if there is one, to eliminate a single failed port.
Suspect the terminations if the run is a fixed one you or somebody else made. Untwisting too much pair before a module or a plug is the most common installation fault there is, and a split pair is its most confusing outcome.
Look for physical damage. A cable stapled too tightly, crushed by a cable tie, trapped in a door, pulled by its plug, or bent tighter than its published minimum radius. None of these leave much of a mark.
Check the length. The standards are built around a hundred-metre channel including the patch leads. A run beyond it is outside anything guaranteed, and behaving well at one rate and not another is exactly how that shows up.
The things that are not usually the cause
A forced speed setting on one end. It happens, and it is worth a look on a managed switch or in a machine’s adapter properties, but it is far rarer than a cable fault.
The category of the cable. Category 5e carries gigabit over the full channel. If a link is falling back to 100, the grade is not the reason; a fault in the channel is.
The switch being old. A gigabit switch that has been forwarding gigabit links for years does not selectively forget one port.
The instrument worth owning
A wiremap tester is inexpensive and answers this question directly. It shows which conductor goes to which pin, and it shows the fault that a continuity check cannot: a split pair, where every pin reaches the other end and two of them are travelling in the wrong twisted pairs.
Anything that certifies a channel against its grade is a different class of instrument entirely, and it is not what is needed here.
Why it is worth fixing rather than tolerating
A link at 100 Mbps is doing a tenth of what the port can do, and the cause is a fault that is still there. A marginal channel that has fallen back once tends to be a channel that reports errors later, and finding it now — with a spare lead and ten minutes — is much easier than finding it during a large transfer that will not finish.
The questions that come up before an order
Could it be the cable even though it is brand new?
Yes. New leads are made in vast numbers and a proportion of them are poor, and a lead that has been trapped in a door, run over by a chair or pulled hard by its plug can have a broken conductor with no visible damage. Swapping in a known-good lead is the fastest test there is.
How would a fault only appear at the higher rate?
Because the higher rate uses conductors the lower one does not, and because it is more demanding of the pairs it uses. A split pair — where continuity is correct but two conductors were taken from different twisted pairs — is the classic case: it passes a simple continuity test and fails as soon as anything fast is asked of it.
Last reviewed 10 September 2026