You pulled Cat6a through the house because it was only a little more expensive than Cat5e and it seemed daft to install the old thing. Then you ran a speed test, and it matched exactly what the old cable did. Nothing is faulty and nothing was wasted, necessarily — but the test could not have come out any other way, because the cable was never the thing setting the number.
The negotiation happens at the ends
An Ethernet link runs at a speed the two ports agree on. A one-gigabit port at each end produces a one-gigabit link, and it does so over Cat5e as readily as over Cat6a, because gigabit needs a fraction of what either cable can carry.
The categories differ in bandwidth, not in speed: Cat5e is specified to 100 MHz, Cat6 to 250 MHz, Cat6a to 500 MHz. That extra headroom becomes speed only when both ports ask for something that needs it, and the ports in most houses do not ask.
What each cable actually carries, and how far
The distances are where this gets concrete, because every category has a speed it supports over a full 100-metre channel and a longer speed it supports over less.
- Gigabit. All three categories, full 100 metres. No argument to be had.
- 2.5 gigabit. Specified over Cat5e for the full 100 metres under IEEE 802.3bz — the standard exists precisely so that installed Cat5e need not be replaced.
- 5 gigabit. Cat6 to 100 metres. Cat5e is not supported at the full length, and gets there only on shorter runs.
- 10 gigabit. Cat6a to the full 100 metres. Cat6 is guided to roughly 37 metres, and may reach 55 depending on how much interference comes from neighbouring cables in the same bundle. Cat5e is not recognised for 10 gigabit at all.
Notice what the 10-gigabit row is really about: alien crosstalk — the noise cables induce in each other when they run together in a bundle. That is why the limit is stated as a range rather than a number, and why it is a serious constraint in a comms room with forty cables in a tray, and a mild one in a house where a single run crosses the loft on its own.
The arithmetic for a house
Now put a domestic installation into those rows. A typical run from the router to a bedroom is 10 to 30 metres. The UK's median household broadband speed is about 80 Mbit/s, with the average near 157.
Two things follow, and they are our reckoning rather than anyone's specification. First, at those lengths the difference between the categories is almost entirely theoretical: every one of them carries gigabit, and Cat5e carries 2.5 gigabit as well. Second, for the cable to become the limit, you need both ends to run at 10 gigabit — and a pair of 10-gigabit ports costs more than the entire cable run they would sit at either end of.
So the honest sequence is the reverse of the one people follow. The cable is the last thing to upgrade, not the first, because it is the cheapest component in the chain and the one least likely to be the bottleneck. The internet connection sets the speed to the outside world; the ports set the speed inside the house; the cable sets a ceiling that neither of them is currently approaching.
When the better cable is the right buy anyway
None of this makes Cat6a a mistake, and there is one case where it is plainly correct: cable that goes inside a wall or under a floor. The cost of the cable is trivial next to the cost of pulling it again, and installed cabling outlives three generations of the equipment attached to it. If the run is buried, buy the headroom.
Two other cases favour the thicker cable, and both are physical rather than about speed. Power over Ethernet puts current down the same pairs, and thicker conductors run cooler, which matters in a bundle or a hot loft. And a run that shares a trunk with mains cable or sits near fluorescent fittings benefits from the tighter noise specification, whatever it is carrying.
Against that, Cat6a is stiffer, has a larger bend radius, is harder to terminate correctly, and is more likely to be terminated badly by someone in a loft with a crimp tool — and a poorly terminated Cat6a run performs worse than a properly made Cat5e one. The category on the jacket is a ceiling, not a promise.
The purchase this does not justify
Do not re-pull working cable that is already in a wall in order to gain a speed nothing at either end can request. If the existing run is Cat5e and it links at gigabit, it is doing its job, and the upgrade that would actually change your speed test is a faster internet line or a pair of multi-gigabit ports — in that order.
Equally, do not buy patch leads by category as though the number were a performance tier. A three-metre lead between a router and a console is not going to be the limiting element in any household network, at any category, and the premium ones on the shelf are priced against a belief rather than a measurement.
The check that takes a minute
Open your router or switch's status page and look at what the port actually negotiated: 100, 1000, 2500. That number is the honest description of your link, and it does not change when you swap the cable unless the cable was faulty or the ports could always have gone faster.
Then measure the length of the run. If it is under thirty metres and the ports say 1000, you have all the evidence you need: the cable has capacity to spare, and the next useful pound goes on something at the ends.
How this was put together
Five independent sources sit under the figures above: the category bandwidth specifications of 100, 250 and 500 MHz for Cat5e, Cat6 and Cat6a; IEEE 802.3bz, which specifies 2.5GBASE-T over Cat5e at the full 100-metre channel while placing 5GBASE-T on Cat6 for that length; TIA guidance on 10GBASE-T over Cat6, giving roughly 37 metres with operation possible to 55 depending on the alien-crosstalk environment; the recognition of Cat6a for 10GBASE-T across the full 100-metre channel, and the non-recognition of Cat5e for that speed; and Ofcom's UK home broadband figures, with a median near 80 Mbit/s and an average around 157.
The derived reading is ours: that in a 10-to-30-metre domestic run every category on sale carries what the household's ports can negotiate, that the cable becomes the limit only when 10-gigabit ports exist at both ends, and that those ports cost more than the run they would terminate.
