Ethernet cable categories: Cat5e through Cat8, and what you need

Leave the Cat5e already in your walls alone. It carries Gigabit to 100 meters by design, and the IEEE project behind 2.5GBASE-T set out to hit 100 meters on that same Cat5e. For anything new inside a wall, buy Cat6A: the last category that reaches a full 100 meters at 10 Gbps on an ordinary RJ45. Cat8 is a 30 meter data center part, and in a house it buys a stiffer cable for speeds nothing there can produce.

Two standards bodies, two sets of names

Categories come from ANSI/TIA-568 in North America and ISO/IEC 11801 internationally, two tracks that are close but not identical. TIA names the cable ("Category 6A"). ISO names the finished channel ("Class EA"). The gap is not purely cosmetic: ISO's Class EA limits run about 3 dB tighter than TIA's at 500 MHz. Vendor copy mixes both freely, which is where the confusion starts. IEEE writes neither; its 802.3bz objectives sheet simply tells the task force to "Select copper media from ISO/IEC 11801:2002."

The current TIA document is ANSI/TIA-568.2-E, which revised 568.2-D and absorbed amendments on Category 8 balun testing and on power delivery. TIA's announcement is dated November 5, 2024; Siemon dates the release to October 24, 2024. One change reaches ordinary installations: DC resistance unbalance testing, previously specified only for Category 8, now applies to Category 5e, 6 and 6A cables, links and channels. Cabling "claiming compliance or certification to ANSI/TIA-568.2-D, may not comply with the current ANSI/TIA-568.2-E requirements," Siemon warns. Check which revision a contractor's test report was run against.

What each category is actually rated for

Cat5e does more than its reputation suggests

Gigabit over Cat5e at 100 meters is not luck. That is what 1000BASE-T was written against. IEEE's 802.3bz project existed because Wi-Fi outgrew its uplink: the project authorisation request calls for more than 1 Gb/s over structured twisted pair to serve "IEEE Std 802.11ac-2013 based enterprise wireless access points that approach 2Gb/s and 4Gb/s." The buildings they hung in were already wired, so the fix had to run over "Category 5e or better twisted pair cabling." The NBASE-T Alliance, founded in October 2014 by Cisco, Aquantia, Freescale and Xilinx, pushed those speeds into IEEE 802.3bz-2016, published October 18, 2016.

On 2.5 Gb/s the objectives sheet is unambiguous: define a PHY for "Up to at least 100m on four-pair Class D (Cat5e) balanced copper cabling." Class D is the ISO channel built from Category 5e components, so that is a 100 meter target on cable pulled twenty years ago.

The 5 Gb/s line is where the internet gets sloppy, in both directions. The same sheet lists two media for it: "Up to at least 100m on four-pair Class E (Cat6)" and "Up to 100m on four-pair Class D (Cat5e)." Cat5e is genuinely in scope, so articles calling 5GBASE-T a Cat6 only application are wrong. But read the wording. Cat6 gets "at least 100m." Cat5e gets "100m" flat. Both sit behind the qualifier "on defined use cases and deployment configurations," mostly a question of alien crosstalk and how tightly your old cable is bundled. A 5 Gb/s port dropping to 2.5 on a twenty year old bundle is that qualifier showing up, not a fault.

Cat6 and the 10 Gbps distance trap

The most repeated error about Cat6 is that it does 10 Gbps to 100 meters. IEEE 802.3an-2006, published September 1, 2006, does say "10GBASE-T specifies a LAN interconnect for up to 100 m of balanced twisted-pair structured cabling systems." That sentence assumes cabling with the alien crosstalk performance to back it, which plain unshielded Cat6 lacks.

Alien crosstalk is leakage between adjacent cables in a bundle, not between the pairs inside one cable, and it is the limiting impairment at 10GBASE-T's frequency. The figure usually quoted for Cat6 is 55 meters; TIA's bulletin on it, TSB-155, is more conservative, generally summarised as drawing the line nearer 37 meters once the crosstalk environment turns unfavourable. Treat it as a range, roughly 37 to 55 meters, the top of it assuming separated runs, small bundles, and Cat6 rated jacks and panels end to end. A standards body does not publish a mitigation bulletin for a problem that is not there.

Cat6A fixes it. The 500 MHz rating and added alien crosstalk parameters buy back the full 100 meters at 10 Gbps, which is why it is right for a run you never want to reopen a wall for.

Cat7 is a trap, and Cat6e does not exist

Cat7 (600 MHz, Class F) and Cat7A (1000 MHz, Class FA) are ISO/IEC classes. The TIA-568 series has no Category 7; the North American track goes from Cat6A straight to Cat8. Cable Matters, which sells patch cable, notes Cat7 lacks approval from IEEE and the EIA.

The connector kills it. Class F performance is specified around GG45 or TERA, neither of which found a market. XDA states the retail consequence directly: "the majority of Cat7 cables sold on Amazon use a standard RJ45 plastic tip," which makes it "structurally impossible to maintain true Cat7 criteria over a standard RJ45 layout." What arrived in the box is a heavily jacketed cable delivering, at best, what its RJ45 ends can deliver. Cat6A won that market instead: RJ45 native, with TIA and IEEE behind it.

"Cat6e" is worse: it appears in neither TIA-568 nor the ISO/IEC class list, which runs D, E, EA, F, FA, I, II with nothing between E and EA. No test limit exists, so a field certifier passes such a cable as Cat6 or Cat6A and nothing else. The label means a manufacturer wanted a number in a gap that has none.

Shielding is a system, not a cable feature

A shield (foil is "F", braid is "S") carries EMI-induced noise currents to ground and off the data pairs. Siemon's grounding whitepaper states the payoff narrowly: screened and fully shielded 10 Gb/s systems such as Category 6A F/UTP and Category 7 S/FTP are "all but immune to the alien crosstalk that presents problems for category 6A UTP cabling," the comparison there being 6A UTP, not Cat6.

The catch is grounding, and the honest version cuts both ways. The same whitepaper cites ANSI-J-STD-607-A: a standards based UTP system "requires no path to ground," while a screened or shielded one must be bonded to the Telecommunications Grounding Busbar in the telecommunications room. The outlet self grounds to the patch panel on insertion, a 12 AWG wire on the panel's ground lug grounds it to the rack, and a 6 AWG wire connects that rack to the TGB. Siemon then calls the fear of all this "unfounded," arguing that a facility already bonded to safety code "more than satisfies" what shielded cabling needs. Inside a telecommunications room, that is fair.

Your house is not one. No TGB, no bonded rack, usually no metal panel with a ground lug. Siemon describes no residential path, so it says nothing about an unbonded shield in a bedroom wall. The common account, widely repeated rather than measured, is that a shield with no defined return path couples noise in rather than carrying it away. Continuity is the firmer point: one unshielded jack or panel breaks the chain. Buy unshielded Cat6A for a house unless you can name the interference source, a long parallel run beside mains cabling, say, or a motor. If you buy shielded, buy shielded jacks, panels and cords in the same order.

Cat8 is the wrong purchase for a house

Cat8 is real and fully published, not draft: TIA ratified it under 568-C.2-1 in 2016 at 2000 MHz, four times Cat6A's 500 MHz. ISO/IEC splits it in two. Category 8.1 is Class I, minimum construction U/FTP or F/UTP, backward compatible with Class EA on the 8P8C connector you own. Category 8.2 is Class II, minimum F/FTP or S/FTP, built around TERA or GG45. Neither has an unshielded form, and that mandatory foil is part of why Cat8 is stiffer to route than Cat6A.

Thirty meters is the whole problem, and it was a design input, not a discovered limit. The task force wrote it into the objectives before the PHYs existed: a link segment of 4-pair balanced twisted pair copper, "up to 2 connectors," "up to at least 30 m." IEEE 802.3bq-2016 repeats it: 25GBASE-T and 40GBASE-T "specify LAN interconnects for up to 30 m of balanced twisted-pair structured cabling." That is a rack to rack budget. Basement panel to a far upstairs bedroom, measured along the route the cable really takes, often spends it before you plug anything in.

Nothing in a house can use the speed either. 25G and 40G over twisted pair are switch to server products. Consumer motherboards sit at 1GbE and 2.5GbE, with 10GbE priced as a step up. Ten gigabit is the ceiling worth cabling for. IEEE has published no BASE-T amendment above 40GBASE-T, so there is no successor category for Cat8 to get you ahead of. Cable Matters, which sells the stuff, says it on its own blog: "For most home applications, Cat6a is more than sufficient and provides better value than Cat8."

Before ordering anything, read the print on the jacket of what you already have. If it says Cat5e and the run is under 100 meters, a 2.5GbE switch will light it up at full rate without opening a wall.