Quick Answer: 10GBASE-T over unshielded Cat6 is limited to 55 m, not 100 m, and that figure assumes low alien crosstalk from neighbouring cables. Cat6 is rated to 250 MHz while the 10GBASE-T line code puts energy up to roughly 500 MHz, so the link works on margin rather than on headroom. Shielded Cat6 and Cat6a at 500 MHz are specified to the full 100 m channel. Plan for 10GBASE-T ports dissipating about 2 to 4 W each, and for a permanent link test that includes alien crosstalk, not just continuity and wire map.
10GBASE-T is the only common 10G Ethernet that keeps the RJ45 and reuses installed copper, which is exactly why it gets designed badly. The PHY has to push 10 Gbps down four pairs in a channel that was qualified for 1 Gbps, so every decibel counts: insertion loss, return loss, and crosstalk from the cable lying against yours in the tray.
Where the 55 Metre Rule Comes From
IEEE 802.3an defines 10GBASE-T with a 400 MHz signalling band and a 500 MHz spectral mask, while Cat6 is characterised to 250 MHz. Above 250 MHz the cable still passes signal, but attenuation and pair-to-pair crosstalk rise steeply and manufacturing tolerance, not the CAT label, decides the outcome. That is the origin of the 55 m allowance for Cat6 UTP: it is a statistical budget for a single link in reasonable conditions, not a guarantee for every installed run.
Alien Crosstalk Is a Bundle Problem
Alien crosstalk is coupling between pairs in different cables, so no amount of care at the termination fixes it. Ten Cat6 runs bundled tight and all carrying 10GBASE-T couple into each other, and the link that passes on the bench fails in the rack. The practical countermeasures are spacing, loose bundles and avoiding long parallel runs against power conduit.
Shielded Cat6 changes the limit
An overall foil shield with a drain wire blocks most alien coupling, which is why F/UTP Cat6 is commonly qualified to 100 m for 10GBASE-T. The shield must be bonded at both ends to a properly earthed patch panel, otherwise it becomes an antenna. If your installer cannot show a bonded earth at each end, buy Cat6a instead of shielded Cat6.
Distance by Cable Type
The table is the part worth copying into a quote. Distances are channel limits, so they include patch cords at both ends.
| Cable | Rated bandwidth | 10GBASE-T channel | Field note |
|---|
| Cat5e | 100 MHz | Not supported | 1 Gbps only |
| Cat6 UTP | 250 MHz | 55 m | Sensitive to bundling |
| Cat6 F/UTP | 250 MHz | 100 m | Requires bonding |
| Cat6a U/FTP | 500 MHz | 100 m | Larger diameter |
| Cat6a S/FTP | 500 MHz | 100 m | Data centre trays |
If the run is 40 m and lightly loaded, standard Cat6 cable line is a rational choice. If it is 70 m, or shares a tray with twenty other 10G links, move to Cat6a 10G cable and stop arguing with the tester.
Heat, Port Power and Uplinks
A 10GBASE-T PHY is not free. Depending on generation it dissipates roughly 2 to 4 W per port, so a 48 port switch dumps well over 100 W of heat into the closet before the optics or the power supply are counted. That is why some 10GBASE-T switches ship with loud fans and why dense copper closets run hotter than the room design assumed. It also affects the cable: bundled Cat6 in a warm ceiling does not improve with age.
When copper is the wrong answer
For switch to switch links inside one rack, a direct attach cable or fibre costs less power per port and sidesteps alien crosstalk completely. Keep 10GBASE-T for the last drop to a workstation, an access point or a server that already has an RJ45 port.
Practical Scenarios
- Office refit with existing Cat6: measure the longest channel before promising 10G. Anything past 50 m should be re-run with Cat6a Ethernet cable.
- Dense 48 port bundle: assume alien crosstalk and test with a unit that reports it. A wire map pass proves almost nothing at 10G.
- Ceiling run above a hot plant room: derate. Heat raises insertion loss, and the 55 m budget already assumes a benign route.
- Two switches 3 m apart: use a DAC, not copper RJ45 ports. Lower power, lower latency, no crosstalk argument.
FAQ
Will 10GBASE-T negotiate at 100 m on Cat6?
Often it will, because the PHY downshifts or retries, but that is not a design. Beyond 55 m the link has no specified margin, so it may run for months and drop when a new cable is added to the tray. Treat any Cat6 10G run over 55 m as out of specification and re-cable it.
Does Cat6a need different connectors?
Yes in practice. Cat6a is thicker with a larger conductor and often a foil, so it needs RJ45 plugs and keystones rated for that diameter and for 500 MHz. Standard Cat6 plugs will not close on the jacket, and a plug that only grips the pairs gives you a link that fails return loss.
Can I mix Cat6 and Cat6a in one channel?
You can, but the channel is only as good as its worst section. Two Cat6a patch cords on a long Cat6 permanent link do not extend the 55 m limit upward. Keep the whole channel at one category and test the installed link, since the patch cords are part of the 100 m budget.
